Perspective

Downloaded from https://academic.oup.com/ptj/article/89/3/267/2737718 by University of Wisconsin-Madison Libraries user on 24 October 2023

Movement Variability and the Use of
Nonlinear Tools: Principles to Guide
Physical Therapist Practice
Regina T Harbourne, Nicholas Stergiou
RT Harbourne, PT, MS, PCS, is As-
sistant Professor, Physical Therapy
Fields studying movement generation, including robotics, psychology, cognitive Department, Munroe-Meyer Insti-
science, and neuroscience, utilize concepts and tools related to the pervasiveness of tute, University of Nebraska Med-
variability in biological systems. The concepts of variability and complexity and the ical Center, 985450 Nebraska
nonlinear tools used to measure these concepts open new vistas for physical therapist Medical Center, Omaha, NE
practice and research in movement dysfunction of all types. Because mounting 68198-5450 (USA). Address all
correspondence to Ms Harbourne
evidence supports the necessity of variability for health and functional movement, at: [email protected].
this perspective article argues for changes in the way therapists view variability, both
in theory and in action. By providing clinical examples, as well as applying existing N Stergiou, PhD, is Director of the
Nebraska Biomechanics Core Fa-
knowledge about complex systems, the aim of this article is to create a springboard cility, College of Education and
for new directions in physical therapist research and practice. College of Public Health, Univer-
sity of Nebraska at Omaha and
University of Nebraska Medical
Center.

[Harbourne RT, Stergiou N. Move-

ment variability and the use of
nonlinear tools: principles to
guide physical therapist practice.
Phys Ther. 2009;89:267–282.]

© 2009 American Physical Therapy

Association

Post a Rapid Response or

find The Bottom Line:
www.ptjournal.org

March 2009 Volume 89 Number 3 Physical Therapy f 267

Movement Variability and the Use of Nonlinear Tools

V
ariability in human perfor- ibility, our patients need adequate features. In general, biological sys-
mance and the nonlinear man- variability of the motor system. It fol- tems, including humans, are com-
ner in which skills and charac- lows that adequate variability should plex, nonlinear systems with inherent
teristics of movement change over be a focal point of examination and variability in all healthy organisms.7
time reflect the complexity of the intervention in order to achieve op-
movement system. As Bernstein1 de- timal function for the individual. In this perspective article, we will

Downloaded from https://academic.oup.com/ptj/article/89/3/267/2737718 by University of Wisconsin-Madison Libraries user on 24 October 2023

scribed, multiple degrees of freedom provide an overview of the con-
of the body, including joints, mus- Clinicians trained in medical fields structs of variability, complexity in
cles, and the nervous system, com- use linear models for prediction and human movement, and nonlinear
bine with external forces during problem solving.3 However, it is be- analysis. This overview focuses on
movement to produce countless pat- coming increasingly clear that linear clinical relevance; therefore, the def-
terns, forms, and strategies. The re- models are limited in many cases and initions and explanations herein are
dundancy of the system allows for are certainly not the optimal model more conceptual than mathematical.
the use of multiple strategies to ac- for function. Professionals in several We first describe variability of move-
complish any given task. Logically, areas of health care, including epi- ment and its relationship to com-
there are multiple performance vari- demiology, infectious disease pro- plexity and the changing view of the
ants for each movement, depending cesses, and biomedicine, are turning importance of variability for success-
on the constraints of each individu- to nonlinear models for solutions ful function. Next, we review non-
al’s system. How do therapists utilize to difficult problems.4 For example, linear measures and the concept of
these redundancies in practice? Al- drug dosages are nonlinear; one can- complexity in movement. Finally,
though many movement science and not increase the dose for more effect we propose principles and examples
neuroscience students are now well because there is generally a thresh- of the use of variability, complexity,
versed in the importance of variabil- old value at which the desired effect and nonlinearity in examination, in-
ity, this information has not been in- occurs and beyond which negative tervention, and research compo-
tegrated into many physical therapy effects occur.5 In our field, thera- nents of physical therapist practice.
interventions and is far from being pists know that a given amount of
embedded in general practice.2 practice cannot ensure the learning The Importance of
of a skill in a linear manner. Our Variability and the
Think about the last time you en- patients usually learn sporadically, Concepts of Nonlinearity
couraged a patient to acquire a new progressing in a nonlinear manner Variability: For Better or Worse?
or more-efficient movement. How over time. In addition, rates and Before discussing the changing per-
did you go about it? Did you demon- paths of progress vary among indi- spective on variability as it pertains
strate the movement and ask the pa- viduals with the same diagnoses who to clinical application, consistent in-
tient to copy your movement pat- are similar in characteristics. Motor terpretation requires definitions of
tern? Did you ask the patient to learning progresses nonlinearly, ex- terminology. First, the definition of
experiment with various movement hibiting nonlinear learning curves variability occurs in behavioral, bio-
strategies to find success? Perhaps depending on the task, conditions, logical, and statistical forms. Behav-
you asked the patient to think about and characteristics of the learner.6 ioral variability describes differ-
the movement, the goal, a similar Linear can be defined as pertaining ences in observed behavior when an
task, or a mental image of perfor- to a straight line, or consisting of entity is placed in the exact same
mance. Was some feature of the only one dimension. Nonlinear, usu- situation.8 Webster’s Dictionary de-
movement a focus for measurement ally used with the term “dynamics,” fines variability in a biological sense
to determine progress or change? Re- as in “nonlinear dynamics,” can be as “the power possessed by living
gardless of the practice setting, we defined as “not in a straight line,” or organisms, both animal and vegeta-
ask patients to move, and we gener- as a system whose output is not pro- ble, of adapting themselves to mod-
ally ask them to move differently portional to its input.* Nonlinear sys- ifications or changes in their environ-
than they moved before they came tems are more complex than linear ment, thus possibly giving rise to
to us. Although our goals often do systems, necessitating the use of sets ultimate variation of structure or
not explicitly target variability in of equations producing unpredict- function.”9 Statistical variability re-
movement, our implied expectation able outcomes that exhibit chaotic fers to measures of centrality around
is that the functional movement that a mean or an average and includes
emerges will be adaptive and flexible measures such the standard devia-
* Terms needed to understand nonlinear con-
enough to meet the everyday goals cepts are underlined and are listed in tion, the range of possible values,
of our patients. To achieve this flex- Appendix 1.

268 f Physical Therapy Volume 89 Number 3 March 2009

 Movement Variability and the Use of Nonlinear Tools

and the variance. All of these defini- Counter to a therapeutic assumption edly fall both in and outside the ac-
tions contribute to our understand- that equilibrium is an indicator of ceptable range in such an individual.
ing of human variability. health, nonlinear theories empha- When one movement falls outside
size disequilibrium as healthy. This the expected range, the next move-
Human movement variability encom- means that the system never quite ment is perturbed. Gait is an exam-
passes the normal variations that oc- settles into a stable state, and con- ple of a continuous, cyclic task,

Downloaded from https://academic.oup.com/ptj/article/89/3/267/2737718 by University of Wisconsin-Madison Libraries user on 24 October 2023

cur in motor performance across stant fluctuations characterize the within which the steps cannot be
multiple repetitions of a task over healthy variability that allows adap- random but also are not completely
time.10 Variability is inherent within tation to environmental change.4 A repeatable or robotic. Thus, we
all biological systems and reflects complex dynamic system is in slight propose that optimal movement
variation in both space and time, disequilibrium with the environment variability lies between too much
which can be illustrated easily in hu- and maintains this disequilibrium variability and complete repeata-
man movement. As a person walks over time.14 Goldberger15 described bility.10 Scientists define this optimal
through sand or snow, his or her complete equilibrium as equivalent range of variability using mathemati-
footprints never repeat exactly, re- to the death of the organism because cal chaos. Furthermore, techniques
flecting variability from step to step it implies a static, nondynamic state. from mathematical chaos describe
in a continuous cycle of movement. Therefore, health is indicated by a important features of variability.19,20
During quiet standing, we sway dynamic equilibrium that is not a
around a central equilibrium point static state. How Nonlinear Tools Advance
without ever remaining exactly still Understanding of Variability
as we maintain orientation to the Rather than being a negative feature, We have established that variability
world. Are these examples of vari- variability reflects important infor- is not a negative feature of functional
ability in movement considered er- mation for the maintenance of the movement or of biological systems
rors in motor performance, or are health of the system. Reduced vari- in general. However, exactly when is
they normal output of a healthy mo- ability is known to cause repetitive variability good? Moreover, how do
tor system? stress injury in a mechanical sense. we determine whether the varia-
Although outwardly this appears to bility we see is good variability or
Traditionally, random error or noise be a mechanical problem, the under- bad variability? The answers to these
within the system was deemed re- lying story describes an information questions lie in understanding dy-
sponsible for the variability mea- problem. The lack of variability in namic models of movement.
sured during repetitions of a skill.11 movement leads to abnormal map-
Motor learning textbooks usually de- ping of the sensory cortex, which The conversation about variability
scribe movement variability as error subsequently disturbs motor func- began with systems theory, a depar-
and skilled movement as movement tion. These neural maps (both sen- ture from linear models. In 1990 at
with decreased variability.12 General- sory and motor) are more complex the II STEP conference, dynamic sys-
ized motor program theory (GMPT) when movement variability is tems theory (DST) was promoted as
considers variation in a given move- present and less complex when vari- a model for the progression of move-
ment pattern to be the result of ability is reduced.16 –18 Movements ment skills and as a way to model
errors in the ability to predict the with an optimal amount of variability change in movement skills for phys-
necessary parameters for using the avoid this abnormal mapping and es- ical therapists.21,22 Dynamic systems
underlying motor program.13 How- sentially contribute to the neuroplas- theory introduced the notions of
ever, there is mounting evidence of ticity needed for maintaining or stability and nonlinearity to explain
the importance of variability in nor- achieving functional skill. Thus, vari- variability. Based on environmental,
mal movement, which reveals varia- ability of the movements used for the biomechanical, and morphological
tion not as error but as a necessary task contributes information to the constraints, any biological system
condition for function. Variability nervous system, which then serves will self-organize to find the most
reflects multiple options for move- to prevent injury. stable solution. An increased amount
ment, providing for flexible, adap- of variability indicates less coopera-
tive strategies that are not reliant on Too much variability also can be a tive behavior among the compo-
rigid programs for each task or for problem, such as in an individual nents of the system, which eventu-
each changing condition encoun- with an ataxic movement disorder. ally can drive the system to new
tered. Optimal variability as a central Movements that usually fall within a attractor states or behaviorally stable
feature of normal movement is con- specific range of variability to accom- solutions. This is a nonlinear system
sistent with a nonlinear approach. plish a task such as gait unexpect- because the input does not lead to a

March 2009 Volume 89 Number 3 Physical Therapy f 269

Movement Variability and the Use of Nonlinear Tools

linear change in output. Input havioral states, with variability challenging terrains. Therefore, it
changes the variability of the system, considered not as error but rather as seems that variability does not de-
which may drive the emergence a source of behavioral change. crease when we develop and refine a
of a new behavior. Importantly, the Through DST, the importance of stable behavioral state but actually
measure of the variability provides a variability attracted the attention of increases. The structure of variability
means to classify the stability of the developmentalists, who recognize (as opposed to amount) can be de-

Downloaded from https://academic.oup.com/ptj/article/89/3/267/2737718 by University of Wisconsin-Madison Libraries user on 24 October 2023

system. Thus, in DST, small amounts DST as a theoretical starting point scribed using nonlinear tools. These
of variability will indicate a highly to study the emergence or self- nonlinear tools best capture varia-
stable behavior. organization of developmental ac- tion in how a motor behavior
tions, perceptions, and cognitive emerges in time, for which the tem-
The GMPT and DST perspectives are skills.23 However, the variability poral organization in the distribution
similar in that both recognize that within an existing state (as opposed of values is of interest. Temporal or-
decreased variability results from to a developing state), behavior, or ganization, or “structure,” is quanti-
the efficient execution of a given established movement function has fied by the degree to which values
movement pattern. Dynamic sys- not been appreciated as important emerge in an orderly (ie, predict-
tems theory differs, however, be- to skillful movement in adults or in able) manner, often across a range of
cause it proposes that at a specific describing pathological conditions. time scales. Nonlinear tools quantify
critical point, variability increases The use of linear measures, such as the nature or the structure of vari-
and dominates the system as new the standard deviation, limits our un- ability and provide the missing abil-
movement emerges. The system be- derstanding of variability as a win- ity to quantify the concept of “stabil-
comes highly unstable and switches dow to view the nature of adaptation ity” from DST.
to a new, stable behavioral state. in functional skills. Consequently,
An example of this process occurs the ideas from DST of “stable state” Figure 1 illustrates these concepts of
when an individual learns to ride a and “attractor” are not intrinsically variability. The figure pictures 4 dif-
bicycle. If a child has been using accepted as part of our therapeutic ferent time series, with the linear
training wheels, pedaling and con- world. This is partially because we measure of range and the nonlinear
trolling the direction of the bicycle have lacked the tools to see the “hid- measure of approximate entropy
are stable behaviors. However, when den” information in the variability of (ApEn) listed beside each signal. The
the training wheels are removed, movement progressing over time. first and third rows show signals that
the system is perturbed. Pedaling Variability and its underlying charac- look messy, seeming to be random,
and directional control are very er- teristics are not completely de- with one signal larger in amplitude
ratic and unstable as the individual scribed or quantified in either GMPT than the other. Beside these rows,
learns the interaction of balance or DST, even though variability has the range values reflect the varying
with the speed of the bicycle and an important role in both theories. amount of the signal between the 2
controlling movement through space. traces, with the larger number by
At the point where the individual Over time, it was recognized that the larger signal; however, the ApEn
understands how the parameters of neither GMPT nor DST accounts for values are equivalent for these 2
speed, balance, and directional con- the observation that some behaviors, signals. This reflects the fact that
trol interact, the behavioral state (in- which appear to be stable, paradox- only the amplitude varies and not the
dependent bicycle riding) becomes ically occur in quite variable ways. structure of the time series. The sec-
stable. The system must cross a crit- This is especially evident when we ond and fourth rows depict time
ical point, where the speed of the observe elite sports players or musi- series that are very regular, a sine
bicycle makes balance easier for the cians performing skillfully. Even wave. Again, comparison of the
behavior to become less variable in though they perform the same skill range shows they are different in am-
its new state. as others, they seem to have devel- plitude but the same in structure, as
oped an infinite number of ways to reflected in the ApEn value. How-
In DST, increased variability in the perform it. Thus, it seems that a very ever, comparison of the first signal
system reveals growing instability, stable behavioral state is supported with the second signal (and the third
which may lead to a shift to a new by a very “rich” behavioral reper- signal with the fourth signal) shows
attractor, or a new behavior. Con- toire. If we consider fundamental that the amplitude, quantified by the
versely, a lack of variability traps a motor skills (ie, gait, posture) and range, is the same (and the standard
behavior in a specific state or attrac- not the skills of an elite athlete, we deviation, a linear measure of vari-
tor. Thus, DST advances our under- are all skillful in our ability to walk ability, also would be the same) but
standing of transitions between be- through crowds or on diverse and that the structure of the series, de-

270 f Physical Therapy Volume 89 Number 3 March 2009

 Movement Variability and the Use of Nonlinear Tools

Downloaded from https://academic.oup.com/ptj/article/89/3/267/2737718 by University of Wisconsin-Madison Libraries user on 24 October 2023

Figure 1.
Comparison of linear and nonlinear measures of several signals. Four signals are displayed, with the respective values for range and
approximate entropy (ApEn).

scribed by the nonlinear ApEn value, rope, imagine your first attempt. You is characteristic of a skilled and prac-
is different. Therefore, the amount of would likely have wide-ranging ex- ticed tightrope walker.
variability measured by the standard cursions of your center of pressure
deviation (linear) and the structure (COP) at the support surface and The overall task is difficult to break
of variability measured by ApEn wide movements of your body seg- into parts and analyze because the
(nonlinear) are not at all the same. In ments as you try to balance. This different components are interde-
fact, as we will discuss later, they reveals large variability according to pendent, and there must be online
often are inversely related. More- many measures, including kinemat- adjustments calibrated to the rest
over, these different facets of vari- ics, COP movement, and center-of- of the system. The overall system is
ability can reveal information that mass movement. The performer tries complex because the analysis of the
may lead to different clinical deci- many different strategies that may in- system or function is inaccurate if
sions, which is illustrated in a clinical clude stiffening or loosening various examined part by part. Although de-
example later. body segments in an attempt to bal- scribing a range of movement op-
ance on the tightrope. The speed of tions quantifies variability, complex-
Nonlinear Measurement the performer’s reactions also may ity is more difficult to measure. This
and Description be varied. However, these early at- is because measuring each part of
How Does Variability Relate to tempts to accomplish the task of bal- the movement separately will not
Complexity in Functional ancing on the tightrope would not give us an overall measure of the
Movement? be complex, even though they were complexity required for success in
In terms of physical therapy, varia- highly variable. Complexity would the function. Complexity is some-
bility describes the behavioral re- arise from fine-tuned adjustments, thing that is “hidden” within the
pertoire possible for a given func- with selected and well-practiced yet time series of a movement sequence
tion. We will use the example of flexible strategies for balance. These or strategy as it emerges over time.
controlling balance in a new task. If strategies utilize specific information Movements that occur at one mo-
you have never walked on a tight- to make the optimal response, which ment affect and are affected by

March 2009 Volume 89 Number 3 Physical Therapy f 271

Movement Variability and the Use of Nonlinear Tools

movements that occur either before vide insight into the amount of vari- “population thinking,” and he used it
or after the movement in the mo- ability (Fig. 1). to describe the complexities of the
ment. Tools for measuring complex- immune system and then expanded
ity come from nonlinear dynamics, Nonlinear tools give us additional in- the principles to neuroscience and
and mathematical models incorpo- formation about the structure of vari- the way the mind works.29 The basic
rate time to describe this complexity. ability, which describes the evolu- idea is that variability allows an or-

Downloaded from https://academic.oup.com/ptj/article/89/3/267/2737718 by University of Wisconsin-Madison Libraries user on 24 October 2023

tion of the movement over time. This ganism or group of organisms to
Linear Measures: Traditional is possibly the reason why previous be more successful. This pertains to
Measures of Dispersion theoretical accounts of variability in antibodies and viral agents, animal
Linear tools to measure variability human movement (ie, GMPT, DST) species, the nervous system, and the
provide information about the quan- supported the notion that small evolutionary progress of plants and
tity of a signal, they but do not tell us amounts of variability characterize a animals. Variability allows choices
about the time-evolving nature of very stable behavioral state and that among options, selection of strate-
the signal. Linear tools include the improved stability links directly to gies, and flexibility to adapt to varia-
statistics of range, standard devia- decreasing variability. If we measure tions in the environment. If an
tion, and coefficient of variation and improved performance linearly, this animal rigidly performs limited be-
are limited in their explanation of conclusion is very reasonable. How- haviors or functions only within a
human movement variability for sev- ever, nonlinear measures provide ad- restricted environment, it is chal-
eral reasons. One reason is that ditional information and allow un- lenged for survival by a more-
data from several averaged trials gen- derstanding of complexity. adaptive animal. This Darwinian
erate a “mean” picture of an individ- principle describing the advantage
ual’s movement pattern. The mean In Figure 2, it is evident that the time of variability lies within many levels
removes the temporal variations of series signals on the left are different of organisms and is explicitly de-
the movement and masks the true from each other; the first is very reg- scribed for the growth of genetic
structure of variability present in the ular, the second seems to have some complexity in an information-rich
movement pattern. In addition, the type of pattern that is difficult to de- environment.30 Variability exists at
valid use of linear tools to study vari- scribe verbally, and the third seems many levels and within interacting
ability assumes that variations be- to have no pattern. However, when components of a system, often oper-
tween repetitions of a task are ran- these signals are plotted versus their ating at different time scales. Thus,
dom and independent (of past velocities (phase plane), it becomes variability may not be obvious at one
and future repetitions), which has clear that the first signal is com- level but can be revealed at another
been shown to be false.24 –26 Finally, pletely regular, with no variation level. The variability inherent in bio-
linear tools provide different an- from the first cycle to the last cycle. logical systems from genes to behav-
swers when compared with nonlin- The second signal forms a complex, ior cannot be considered error if it is
ear tools regarding the way that they yet organized, pattern, with similar pervasive from one species to an-
evaluate variability. For example, tra- paths for each cycle, but not repeat- other and is linked to survival.
ditional linear measures of postural ing the same path and with each
sway quantify only the magnitude of path dependent on a previous path Why Isn’t Movement
sway and not the temporally evolv- and influencing the next path. The Variability Just Error?
ing dynamics (or disequilibrium) of third is a random signal, where the If movement variability is equivalent
postural control. Despite their use in paths are not similar and not depen- to error, we can reasonably assume
many studies, it is becoming evident dent on each other. We propose that that more-skillful individuals would
that linear measures, such as the functional and healthy movement re- have less error (or variability) at the
range and standard deviation of sembles chaos, the complex center outset of learning and then quickly
the COP, do not quantify stability of picture. Nonlinear tools can deter- drop to zero error. In fact, the oppo-
the postural control system because mine whether a chaotic structure, or site is true. Individuals who use a
it is possible to have a large area of complexity, is present in movement. high degree of variability in cognitive
the COP path while having a stable strategies at the beginning of task
posture or an unstable posture.27,28 Why Is Variability Inherent in development have greater learning
Therefore, linear measures of vari- Biological Systems? and eventual success in performing
ability do not accurately define con- Variability is inherent in biological the task.31 Movement researchers
structs important in movement, such systems because it ensures survival. have started to understand the im-
as stability, because they only pro- Gerald Edelman, a Nobel laureate, portance of variability in motor skill
described this pervasive rule as learning and examine performance

272 f Physical Therapy Volume 89 Number 3 March 2009

 Movement Variability and the Use of Nonlinear Tools

Downloaded from https://academic.oup.com/ptj/article/89/3/267/2737718 by University of Wisconsin-Madison Libraries user on 24 October 2023

Figure 2.
Time series and corresponding phase-plane plot. (A) periodic function; (B) time series from a chaotic system, the Lorenz attractor;
(C) random time series.

differently. For example, a study of Thus, the variability at the beginning of variability. If therapists consider
coordination variability in jumpers of task learning may seem like error variability to be error, it is seen as
revealed a U-shaped curve in the pro- because the task is not performed an impairment. However, if thera-
gression of variability.32 Initial high efficiently or accurately. However, pists consider variability necessary
variability occurs as different strate- this initial variability also can be seen for skill acquisition, they will exam-
gies are attempted. Subsequently, as necessary to map the possibilities ine the structure of variability to help
the learner moves toward a reduc- of movement for the task. It then is build skill. Nonlinear measures can
tion in variability as he or she per- refined into a different type of vari- unmask the hidden structure inher-
forms more successfully. Then, sur- ability when the performer is skillful. ent in variability so that intervention
prisingly, as the learner becomes an Although variability typically is can successfully address different
expert, the variability increases known to decrease as a skill is ac- features of variability as necessary
again. This skillful variability indi- quired, think about how our notions during skill development. Although
cates increasing flexibility of skill to of the mechanisms of skill acquisi- clinicians can easily understand the
allow adaptation to perturbations. tion change if we consider the role behavioral variability we have been

March 2009 Volume 89 Number 3 Physical Therapy f 273

Movement Variability and the Use of Nonlinear Tools

discussing, such as the number of As a person’s range of sway in- rate or the COP time series, corre-
strategies for accomplishing a func- creases, the standard deviation in- spond to a rich behavioral state with
tional task, it is more difficult to creases, indicating a greater amount high complexity.10 This state is de-
understand what underlies that of variability. However, if we use a fined as highly variable fluctuations
variability. nonlinear tool to examine the fea- in physiological processes resem-
tures of the variability in postural bling mathematical chaos. This al-

Downloaded from https://academic.oup.com/ptj/article/89/3/267/2737718 by University of Wisconsin-Madison Libraries user on 24 October 2023

Motor skills researchers increasingly sway, we may note that sway has lows the system to have a relatively
find nonlinear tools useful in reveal- become more regular (with more- predictable course, which can adapt
ing information through time series repeatable movement patterns). This if a change in the environment oc-
analysis. The ApEn measure revealed makes sense, because the individual curs. Low levels of complexity cor-
significant differences between ath- must have some specific strategy of respond to states that resemble ran-
letes who had a concussion and con- control to make the appropriate ad- dom, noisy, and erratic behavior or
trols who were healthy by analysis of justments for balance maintenance rigid, periodic, and regular behavior.
the COP time series, even after other when the range of movement is Therefore, with low complexity,
standard linear measures indicated large; otherwise, he or she would adaptability suffers. Nonlinear mea-
that the athletes had fully recovered fall. The relationship between linear sures allow us to extract information
from the concussion.33,34 Moraiti et and nonlinear tools as described hidden in a time series and to evalu-
al35 used the Lyapunov exponent above can further our understanding ate complexity. Some examples of
(LyE) measure to show that a group of the emergence of functional, possible tools follow.
of patients with anterior cruciate lig- adaptive movement.
ament deficiencies exhibited signifi- Approximate entropy is a measure
cantly more rigid and predictable Nonlinear measures always describe that can quantify the regularity or
walking patterns than controls who a time series, or a series of measure- predictability of a time series
were healthy, suggesting a decrease ments taken at specific intervals over (Fig. 1).40 – 44 Increasing ApEn values
in system complexity and narrowed uninterrupted time. For example, reveal greater irregularity. Con-
functional responsiveness. Kurz and the range of motion of a joint during versely, lower values reveal a more
Stergiou36 used an entropy measure each step as a person walks, taken regular or periodic behavior. Ap-
to show that neurophysiological with an electrogoniometer or mo- proximate entropy has been useful
changes associated with aging may tion analysis equipment sampled at in the identification of differences
result in less certainty of the neuro- 30 times per second over a 2-minute between young and old people in
muscular system in selecting a stable period of time, can be presented as a the COP time series in standing,45 in
gait. Therefore, utilizing a nonlinear time series. Looking at a measure in revealing deficits in athletes with a
perspective to examine variability the context of time because it en- concussion compared with athletes
can assist in differentiating between ables us to understand the ability of who were healthy,33,34 and in detect-
health and nonhealth. the system to adapt as conditions ing developmental changes in sitting
change. The period of time may vary postural control.25 Approximate en-
Nonlinear Measures: from seconds to days, but the impor- tropy measures the probability that
New Ways to Describe the tant concept is that a behavior the configuration of one segment of
Nature of Variability emerging from the complex system data in a time series will allow the
We have established that the amount can be described over time with spe- prediction of the configuration of
and structure of variability are 2 dif- cific mathematical nonlinear tools another segment of the time series a
ferent things (Fig. 1). As a result, that are used to quantify order, pre- certain distance apart.41 Data seg-
changes in measures of the amount dictability, regularity, and complex- ments with a greater likelihood of
of variability may be in a completely ity. A time series also is valuable be- having the same configuration or pat-
different direction than changes in cause information that is important tern upon comparison will result in
measures that evaluate the structure to understand the health or function lower approximate entropy values,
of variability. Similarly, in studies of of the individual may be revealed at and data segments with a low likeli-
postural control, gait, and force pro- different time scales. hood of similarity between segments
duction, researchers found that as will result in higher values. Values
measures of the amount of variability Characterizing the nature of the closer to zero are consistent with
increased, measures of the structure complexity present in a time series is maximum regularity, and values
or organization of variability de- of great interest in many scientific nearing 2 represent maximum irreg-
creased.37–39 Let us reflect on an ex- domains, including biology. Healthy ularity.
ample of postural sway in standing. systems, whether referring to heart

274 f Physical Therapy Volume 89 Number 3 March 2009

 Movement Variability and the Use of Nonlinear Tools

Clinically, ApEn is useful for under- Using the LyE, Yamada50 reported complex in nature. Harbourne and
standing the predictability of a move- body sway that resembles mathemat- Stergiou25 and Boker et al26 have
ment. For example, Cavanaugh et ical chaos from COP data during used this technique to show that
al46 utilized ApEn to determine the standing in adults who were healthy, variability in the COP time series
predictability of everyday walking thus revealing inherent complexity. from infants during the development
activities, using time series data from Use of the LyE in an ongoing inter- of independent sitting is not just

Downloaded from https://academic.oup.com/ptj/article/89/3/267/2737718 by University of Wisconsin-Madison Libraries user on 24 October 2023

a pedometer. They found that inac- vention study with infants with cere- noise but has a deterministic origin.
tive elderly individuals walked less bral palsy was important as a fine- This means that infants learning to sit
and had more-predictable walking grained measure to detect advancing are not just randomly “wiggling.” For
activity than active elderly individu- postural control in sitting, when lin- the clinician, it is important to rec-
als. This finding allows a more com- ear measures of the COP and clinical ognize that within these outwardly
plete picture of the differences in tools did not detect change.51 A stan- unorganized and noisy-looking
walking between inactive and active dardized test, the sitting section of movements are orderly patterns and
individuals and allows the clinician the Gross Motor Function Measure,52 the beginnings of strategies for pos-
to understand why some elderly peo- did not consistently detect change. tural control. The implication for
ple may have a more-difficult time However, a variety of features of the therapists is that a movement that is
responding to fluctuations in routine child’s movement (more attempts to just beginning to emerge will be un-
or adapting to different walking stay vertical, the ability to turn the organized, but necessarily so. The
demands. This insight suggests inter- head without falling, more attempts variability inherent in this disorgani-
vention that can elicit greater com- to reach while sitting with support) zation may be necessary for ultimate
plexity in the activity, rather than were noted as changes in behavior successful selection of movement
just simply increasing the amount of by the parents and therapist. These and postural strategies. This may re-
activity. small changes in movement, or at- flect the system “mapping” the terri-
tempts at new strategies, were quan- tory around the skill region, allowing
The largest LyE is a nonlinear mea- tified by the nonlinear measures of the individual to “get back” to the
sure that can measure the diver- ApEn and LyE, but were not indi- successful region when perturbed.
gence of the movement trajectories cated in the linear measures. This is
(Fig. 2). The LyE describing purely an example of a nonlinear tool pro- Nonlinear analysis includes several
sinusoidal and completely repeat- viding “hidden” information that additional tools such as detrended
able data with no divergence in the would not be easily measured or doc- fluctuation analysis,54 correlation di-
trajectories is zero because the tra- umented otherwise. mension, mutual information, Hurst
jectories overlap rather than diverg- exponent, symbolic entropy, recur-
ing (Fig. 2A). This shows minimal Surrogation is a technique used to rence quantification analysis, and
change in the structure of the vari- determine whether the source of a others. These methods have a com-
ability over time in the data. The LyE given time series is deterministic mon goal: to evaluate the structure
for random, noisy data indicates (has order) in nature.19,53 The tech- or organization of variability and un-
greater divergence in the data tra- nique compares the actual data and a cover the underlying complexity.
jectories (Fig. 2C). The LyE values random data set that has a similar However, they differ in the math-
for the random data are larger, with structure with the original data set in ematic manipulation of the available
values above 0.4.19 The LyE values question. That is, the deterministic time series. Here we do not provide
from data that are described by structure from the original data set is a complete list of all nonlinear tools
mathematical chaos (ie, Lorenz at- removed by generating a random but describe only a few tools to pro-
tractor; Fig. 2B) are between these equivalent with the same mean, vari- vide the basic concepts of nonlinear
2 extremes. Thus, the values be- ance, and power spectra as the orig- analysis. A more comprehensive re-
tween random and periodic define inal data. Subsequently, the LyE (or view on the topic is available for the
complexity, or highly variable fluc- another nonlinear measure) value of interested reader.19
tuations in physiological processes the surrogate data is compared with
resembling mathematical chaos. the LyE value of the original data. Application of Nonlinear
The LyE has been used with gait Significant differences between the Concepts in Practice
time series data to characterize LyE value of the surrogate data and Complexity in Health
the underlying complexity during the LyE value of the original data Goldberger55 described the use of
movement.19,47– 49 indicate that the original data are complexity at the bedside for physi-
not randomly derived and, therefore, cians by providing examples of peri-
may be deterministic and possibly odic behavior of pathologic systems.

March 2009 Volume 89 Number 3 Physical Therapy f 275

Movement Variability and the Use of Nonlinear Tools

Disease brings about a loss of com- modeled using nonlinear tools.62 Ap- bal guidance. As the patient starts to
plexity, with resulting increased ri- plications in the clinic for physical reach for the walker, the therapist
gidity, such as Cheyne-Stokes breath- therapy intervention are now a real- tells the patient to push up from the
ing in patients with heart failure, istic possibility. arms of the chair. The clinician then
tremors in patients with neurologic has the patient practice the sit-to-
disease, and a sinusoidal appearance Clinical Application of Nonlinear stand activity 5 times, giving feed-

Downloaded from https://academic.oup.com/ptj/article/89/3/267/2737718 by University of Wisconsin-Madison Libraries user on 24 October 2023

of heart rate variability in patients Principles and Use of Complexity back as described each time. The
with congestive heart failure.56,57 in Physical Therapy next day, the patient makes the same
The medical field is beginning to rec- Appendix 2 lists some proposed errors, and the therapist provides
ognize the need for a nonlinear view principles for physical therapy inter- less cueing, hoping to fade cueing
toward complexity, particularly for vention emerging from the theory over the next 2 weeks. In addition to
problems that affect multiple sys- and applications in other fields. An counting errors during the sit-stand-
tems. Ahn et al58 described a tradi- example of clinical application in in- walk practice, the therapist notices
tional reductionist approach as the tervention follows. that the patient’s steps are of un-
antithesis of a complexity-oriented equal length and the affected side
approach, with the reductionist ap- Two clinicians must perform an ini- shuffles forward rather than exhibit-
proach being appropriate for use tial evaluation on an elderly man ing a heel-toe pattern of stepping.
with acute, single-system problems, who has had a stroke. One therapist The therapist uses the same ap-
such as an acute infection. However, will use a traditional, linear ap- proach of using verbal and physical
a disease such as diabetes requires proach, and the other therapist will guidance to point out the errors in
management of a problem affecting use an approach based on principles the patient’s gait pattern, fading the
many systems that interact in various of nonlinearity. As a general rule, the feedback over time and counting er-
ways. Extending the nonlinear view, therapist using a linear approach as- rors within the measured distance of
it could be argued that very few sumes that decreasing the variability the walking track.
problems are truly single-system of movement is equivalent to im-
problems and solvable by linear rea- proving functional skill. Therefore, In contrast, the therapist using a non-
soning because each system inter- the therapist has in mind the “cor- linear approach assumes that the
acts with other systems for optimal rect” movement pattern for various general rule for this patient is to en-
function. As in medicine, many clin- functional skills, which she will hance complexity of movement in
ical problems seen in physical ther- guide her patient toward during in- order to improve gait and functional
apy need a nonlinear approach. tervention. Using principles of mo- mobility skills. This will include the
tor learning, the therapist first gives concept of disequilibrium, or keep-
Clinical uses for nonlinear analysis 100% feedback, and then fades sub- ing the patient in a state of dynamic
appear in a variety of disciplines, in- sequently to 50% feedback for vari- equilibrium (as described earlier)
cluding cardiology, neurology, and ous skills.63 Because the therapist during therapy sessions. Addition-
psychiatry. Heart rate analysis using wants measurable outcomes, she ally, the therapist uses the strategy of
ApEn has been used to evaluate risk uses a standard walking course that providing only information for the
factors for sudden infant death syn- has 7.6-m (25-ft) increments up to patient on how to do a task if the
drome.41,42 Nonlinear analysis has 61 m (200 ft). patient does not have a way to get
been useful in verification of im- the information; the rationale is that
plantable cardiac defibrillator inter- To determine decreasing variability variability is encouraged if the pa-
ventions by using entropy analysis of and increasing accuracy, the thera- tient seeks information indepen-
heart rate variability.59 In addition, pist counts errors in the set of pro- dently, and the patient is kept in a
postural control analysis using sta- cedures to sit-stand-walk over a stan- dynamic state.
bilometry has been improved by the dardized course. This principle of
addition of nonlinear analysis, which fading feedback applies as she guides The therapist first asks the patient
can serve to more accurately identify the patient in transferring out of the whether he would like to go sit in a
features of postural control indicat- wheelchair. She first locks the brakes chair 3.05 m (10 ft) away next to his
ing subtle problems in infants,51 de- of the patient’s wheelchair and tells wife. He agrees, and the therapist
velopmental differences between the patient to scoot to the edge of invites the patient to begin the task,
young and elderly people,43 or the chair. The patient receives assis- assuring the patient that the task is
changes that accompany a disease tance to lean forward and place his safe while the therapist is present.
state such as parkinsonism.60,61 Gait feet under his center of mass, both The patient pushes back in the un-
variability also has been studied and with physical guidance and with ver- locked wheelchair, and the wheel-

276 f Physical Therapy Volume 89 Number 3 March 2009

 Movement Variability and the Use of Nonlinear Tools

chair rolls back, putting the patient the process of coming to a standing parts of the task due to the interjec-
further away from the targeted goal. position but is markedly faster than tion of the therapist’s instructions.
The therapist notes that this is a on the first try. The next day the Therefore, the series of movements
point where the patient is not gath- patient makes only one error, moves and postural adjustments in the task
ering enough information for the from a sitting position to a standing of standing from the wheelchair are
task and addresses this problem by position with guarding, and elects to unrelated to each other and unre-

Downloaded from https://academic.oup.com/ptj/article/89/3/267/2737718 by University of Wisconsin-Madison Libraries user on 24 October 2023

having the patient do some guided sit on a couch for some social con- lated to the overriding environmen-
exploration within the task to in- tact with another patient. The thera- tal context and underlying values of
crease flexibility in terms of availabil- pist notes that during walking the the patient. Most importantly, the
ity of options. The therapist tells the patient had short steps and shuffled therapist using the linear measure
patient that he can roll the wheel- with the affected leg. The therapist prevents response flexibility from
chair in many different directions sees this as a possible problem. The being part of the learning process by
just using his feet and challenges him therapist then uses barefoot tasks, insisting that the patient avoid er-
to find 10 different directions of having the patient identify different rors, precisely because the therapist
wheelchair movement by pushing textures or objects under the feet, considers them undesirable. Thus,
with his feet. During this exercise, place pressure on different parts of this therapist focuses on the absolute
the patient maps the way that foot his feet during walking and standing, “correct” pattern of movement, al-
force affects the wheelchair move- and walk with a variety of patterns lowing little complexity in the inter-
ment. The therapist asks whether through different paths and obstacle vention and preventing the emer-
there is a way to keep the wheel- courses. The therapist is more con- gence of a flexible strategy that
chair from rolling, and the patient cerned about increasing the adaptive works for this individual patient.
remembers the brakes. The patient capacity of walking by increasing
then makes multiple errors in his at- variability at a functional speed than If these therapists had the benefit of
tempts to stand, including incorrect in producing a consistent heel-toe examining the patient’s initial gait by
foot placement, reaching for the pattern. linear and nonlinear analysis, they
walker instead of the wheelchair may have noted high variability of
arms, and leaning back and to one What differs in these approaches? the step length (by looking at the
side instead of leaning forward to get The basic difference is that the ther- standard deviation) but low values
up. However, the therapist does not apist using the linear measure seeks on a nonlinear measure (LyE), some-
provide guidance at this point be- to reduce variability of responses what like the signal shown in the last
cause the patient is not making the within the intervention and the mea- row of Figure 1. If the therapists
same error but rather is exhibiting a surement of the task goals and come wanted to decrease variability, they
variety of strategies, which at this to a state of complete equilibrium, would work toward a time series
stage of skill development is desir- whereas the therapist using the non- with a signal like that shown in the
able. At several points, the therapist linear measure seeks to enhance second row of Figure 1 (linear ap-
asks whether the strategy just used complexity by encouraging slight proach). However, if the goal was
was successful, and when answered disequilibrium, particularly at the ini- to increase the structure of the vari-
in the negative, reminds the patient tial stage of task learning. The thera- ability for greater complexity, they
to try some different strategies, just pist using the nonlinear measure would work in a nonlinear fashion
as they did by pushing the wheel- builds complexity into the task, with toward a time series with a signal
chair in different directions with the use of multiple systems: cogni- like that shown in the third row of
the feet. Occasionally, the therapist tive, social, motor, and sensory. The Figure 1. You can see that looking at
gives light touch cues to suggest an practice space is strategically varied, variability from these differing per-
effective strategy. and multiple movement approaches spectives could lead toward different
are encouraged, as well as having an types of intervention, as we have just
At the end of the trial-and-error ses- expanded environmental context in described.
sion, the patient stands and walks which the practice takes place. In
over to his wife to sit in the chair. addition, the functional task pro- Another example of the use of the
After resting and conversing, the ceeds in such a way that the series of principles of nonlinearity to acquire
therapist asks the patient to walk movements within the task are re- or maintain motor skill is a treatment
back to the wheelchair, without giv- lated to each other and dependent approach for the movement prob-
ing any verbal instructions. The on each other. In the linear measure lems of patients with Parkinson dis-
patient makes a few errors, self- approach, each subtask is a task unto ease (PD). The BIG training program
corrects, and visibly thinks through itself and separated from the other focuses on a single attentional pa-

March 2009 Volume 89 Number 3 Physical Therapy f 277

Movement Variability and the Use of Nonlinear Tools

rameter to drive changes in the particular movement form or strat- 5. Examine pilot data to get an idea
motor system.64 This parameter is egy, but rather let the patient dis- of the nature of the fluctuations
“Think BIG,” emphasizing attention cover that increased complexity of in the variable or in the behavior.
to large-amplitude movements. The various movements has an inherent Plotting position and velocity
principles of treatment include high- value in producing success during against each other and even
intensity, multiple repetitions, sa- daily tasks. 3-dimensionally by incorporating

Downloaded from https://academic.oup.com/ptj/article/89/3/267/2737718 by University of Wisconsin-Madison Libraries user on 24 October 2023

liency, and complexity, leading to acceleration can help examine
neuroplastic changes and functional Use of Nonlinear Tools in Motor the organization in the data.
improvement. Although patients Skill Research
with PD have many movement prob- Setting up methodology for nonlin- 6. Replication of a previously re-
lems, including problems with ear analysis in a research project may ported project, but with the addi-
speed, smoothness, accuracy, and seem to be a daunting task. Here are tion of nonlinear tools, can help
quality of movement patterns such suggestions to keep in mind when with planning the methods and
as step length during gait, this ap- designing such a project: interpreting your results. This is a
proach ignores these other move- new area, and, consequently, it is
ment deficiencies. The focus on in- 1. Carefully design your experimen- difficult to know the “standard”
creasing amplitude changes the tal setup, incorporating a matched approach to common issues.
initial conditions driving the move- control group of healthy partici-
ment and shifts the system into a pants to provide reference points 7. If you are considering an interven-
new state space where the move- for observed changes in the non- tion study, keep in mind that a
ment is more skillful and complex. In linear parameters. focus on increasing or decreasing
this way, the BIG approach ad- variability will need to be deter-
dresses complexity because ampli- 2. Seek partnerships with mathe- mined carefully, using a variety
tude serves as the avenue to provide maticians, neuroscientists, and of tools to measure both linear
enhanced adaptability and increased biomechanists who are knowl- and nonlinear factors. The 2 ap-
responsiveness. edgeable about nonlinear tools. proaches are complementary, and
Technical expertise is needed they do not negate each other.
Several principles of nonlinearity are from start to finish on the project,
inherent in this approach. First, the including sampling frequency, 8. Measure more than one variable;
overall variability of movement be- the length of the time series different variables may reflect
havior for these patients is increased needed, and examination of the component skills or constructs
as a fallout of the increase in ampli- data with appropriate nonlinear differently.
tude. The individuals with PD can tools for your questions. Knowl-
now make large movements as well edge of movement measurement Limitations of
as small movements, and all the in- is a benefit in interpretation of the Nonlinear Measures
crements in between, whereas they results. Remember that you will Because nonlinear measurement
had previously been restricted to not “speak the same language” as tools require the use of mathematical
only small movements. Second, an- your collaborator, and be very pa- equations and software to evaluate
other principle of nonlinearity is ap- tient so that ideas can be ex- time series data, nonlinear analysis is
parent because a change in just one changed comfortably. primarily done in the research set-
movement parameter, amplitude, ting. However, the burgeoning inter-
causes a change in other movement 3. Measure a task, skill, or variable est in nonlinear tools in many scien-
parameters that are difficult to pre- that may show emergence of a tific fields bodes well for clinicians.
dict precisely. Third, the input new level of function. In the future, there will likely be de-
needed for a motor system change vices that have embedded software
can come from a different system 4. Variable and task selection must to calculate important measures of
(eg, cognitive, attentional, percep- incorporate a time series; you variability using nonlinear tools.
tual). The focus on attending to mak- should take as long a time series Another limitation is the lack of un-
ing a “bigger” movement, in this as possible within the constraints derstanding of variability and com-
case, recalibrates the perceptual sys- of your target population and plexity in the field of physical ther-
tem to recognize when a movement considering the tools you are us- apy. Physical therapists are taught to
is actually big versus the small move- ing. Many nonlinear tools need use a reductionist approach, as in
ments common to individuals with thousands of data points for accu- most medical fields. This lack of in-
PD. The therapist would not teach a rate use. troduction to nonlinear principles

278 f Physical Therapy Volume 89 Number 3 March 2009

 Movement Variability and the Use of Nonlinear Tools

early in the education of physical 2 Larin H. Quantifying instructional inter- 17 Nudo RJ, Milliken GW, Jenkins WM, Mer-
ventions in pediatric physical therapy zenich MM. Use-dependent alterations
therapists biases the field against the with the Motor Teaching Strategies Cod- of movement representations in primary
productive use of variability and ing Instrument (MTSCI-1): a pilot study. motor cortex of adult squirrel monkeys.
Internet J Allied Health Sci Prac. 2007;5: J Neurosci. 1996;16:785– 807.
complexity. 1–9. Available at: http://ijahsp.nova.edu/ 18 Merzenich MM, Jenkins WM. Reorganiza-
articles/vol5num1/Larin.pdf. Accessed tion of cortical representations of the hand
March 16, 2008.
Additional limitations of the tech- following alterations of skin inputs in-

Downloaded from https://academic.oup.com/ptj/article/89/3/267/2737718 by University of Wisconsin-Madison Libraries user on 24 October 2023

3 Katerndahl DA. Is your practice really that duced by nerve injury, skin island trans-
nique itself create challenges for clin- predictable? Nonlinearity principles in fers, and experience. J Hand Ther.
ical use. Translation of nonlinear family medicine. J Fam Prac. 2005;54: 1993;6:89 –104.
970 –977.
measures to clinical problems re- 19 Stergiou N, Buzzi UH, Kurz MJ, Heidel J.
4 Rickles D, Hawe P, Shiell A. A simple guide Nonlinear tools in human movement. In:
quires concurrent use of linear tools to chaos and complexity. J Epidemiol Stergiou N, ed. Innovative Analyses for
to make associations and determine Community Health. 2007;61:933–937. Human Movement. Champaign, IL: Hu-
man Kinetics Publishers; 2004:63–90.
clinical meaning. The lengthy time 5 Klonowski W. From conformons to hu-
man brains: an informal overview of non- 20 Shelhamer M. Nonlinear Dynamics in
series required for analysis prohibits linear dynamics and its application in Physiology: A State Space Approach.
the use of nonlinear tools for move- biomedicine. Nonlinear Biomed Phys. Hackensack, NJ: World Scientific Publish-
2007;1:5–18. Available at: http://www. ing Co; 2007.
ment that is extremely limited. Lastly, nonlinearbiomedphys.com/content/1/1/5. 21 Heriza C. Designing practice for motor
these measures require multiple rep- Accessed March 16, 2008. learning: clinical implications. In: Lister
etitions or cycles of a movement. 6 Newell KM, Liu YT, Mayer-Kress G. Time MJ, ed. Contemporary Management of
scales in motor learning and development. Motor Control Problems: Proceedings of
Psychol Rev. 2001;108:57– 82. the II STEP Conference. Alexandria, VA:
Conclusions 7 Walleczek J. Self-organized Biological Dy- Foundation for Physical Therapy; 1991.
Optimal variability in human move- namics and Nonlinear Control: Toward 22 Giuliani CA. Designing practice for motor
Understanding Complexity, Chaos, and learning: clinical implications. In: Lister
ment is a characteristic of healthy Emergent Function in Living Systems. MJ, ed. Contemporary Management of
functioning. Nonlinear tools reveal Cambridge, United Kingdom: Cambridge Motor Control Problems: Proceedings of
University Press; 2000. the II STEP Conference. Alexandria, VA:
complexity inherent in normal vari- Foundation for Physical Therapy; 1991.
8 Wray RE, Laird JE. Variability in human
ability, indicating features of motor behavior modeling for military simula- 23 Thelen E, Smith LB. A Dynamic Systems
control that are important for physi- tions. Presented at: Behavior Represen- Approach to the Development of Cogni-
tation in Modeling and Simulation tion and Action. Cambridge, MA: MIT
cal therapists to measure and imple- Conference (BRIMS); 2003. Available at: Press; 1999.
ment in intervention. The applica- http://www.speakeasy.org/⬃wrayre/pubs/ 24 Hausdorff JM, Purdon PL, Peng CK, et al.
VariabilityinHumanBehaviorModeling_Wray
tion of principles based on nonlinear Fractal dynamics of human gait: stability of
Laird_BRIMS2003.pdf. Accessed March 16, long-range correlations in stride interval
dynamics and use of nonlinear tools 2008. fluctuations. J Appl Physiol. 1996;80:
for analysis can provide innovations 9 Definition of variability. Available at: http: 1448 –1457.
//www.webster-dictionary.net/definition/
to guide physical therapist practice 25 Harbourne RT, Stergiou N. Nonlinear anal-
variability. Accessed March 16, 2008. ysis of the development of sitting postural
and research. 10 Stergiou N, Harbourne R, Cavanaugh J. control. Dev Psychobiol. 2003;42:368 –377.
Optimal movement variability: a new per- 26 Boker SM, Schreiber T, Pompe B,
spective for neurologic physical therapy. Bertenthal BI. Nonlinear analysis of
Both authors provided concept/idea/project J Neurol Phys Ther. 2006;30:120 –129. perceptual-motor coupling in the develop-
design and writing. 11 Glass L, Mackey MC. From Clocks to Cha- ment of postural control. In: Kantz H,
os. Princeton, NJ: Princeton University Kurths J, Mayer-Kress G, eds. Nonlinear
This work was funded by a National Institute Press; 1988. Techniques in Physiological Time Series
of Child Health and Human Development Analysis. Heidelberg, Germany: Springer;
12 Schmidt RA, Lee TD. Motor Control and 1998:251–270.
grant (NIH 1K25 HD 047194-01A1) to Dr Learning: A Behavioral Emphasis. 4th ed.
Stergiou (primary investigator); a US Depart- Champaign, Ill: Human Kinetics Publish- 27 Palmieri RM, Ingersoll CD, Stone MB,
ers; 2005. Krause BA. Center-of-pressure parameters
ment of Education, National Institute used in the assessment of postural control.
for Disabilities and Rehabilitation Research 13 Schmidt RA. Motor schema theory after 27 J Sport Rehabil. 2002;11:51– 66.
years: reflections and implications for a
grant (USDE FIP 84.1336-1) to Dr Stergiou new theory. Res Q Exerc Sport. 2003;74: 28 Hughes MA, Duncan PW, Rose DK, et al.
(primary investigator); and the Nebraska Re- 366 –375. The relationship of postural sway to sen-
search Initiative. 14 Price I. Complexity, complicatedness and sorimotor function, functional perfor-
mance, and disability in the elderly. Arch
complexity; a new science behind organi-
This article was received May 1, 2008, and zational intervention? E:Co. 2004;6:40 – 48. Phys Med Rehabil. 1996;77:567–572.
was accepted December 12, 2008. 29 Edelman GM. Bright Air, Brilliant Fire:
15 Goldberger AL. Heart rate variability: tech- On the Matter of the Mind. New York,
DOI: 10.2522/ptj.20080130 niques, applications and future directions. NY: Basic Books; 1992.
Available at: http://www.physionet.org/
events/hrv-2006/goldberger-1.pdf. Ac- 30 Adami C, Ofria C, Collier TC. Evolution of
cessed March 16, 2008. biological complexity. Proc Natl Acad Sci
References U S A. 2000;97:4463– 4468.
16 Byl NN, Nagarajan SS, Merzenich MM,
1 Bernstein N. The Coordination and Reg- et al. Correlation of clinical neuromuscu- 31 Siegler RS. Emerging Minds: The Process
ulation of Movements. London, United loskeletal and central somatosensory per- of Change in Children’s Thinking. New
Kingdom: Pergamon Press; 1967. formance: variability in controls and pa- York, NY: Oxford University Press Inc;
tients with severe and mild focal hand 1996.
dystonia. Neural Plast. 2002;9:177–203.

March 2009 Volume 89 Number 3 Physical Therapy f 279

Movement Variability and the Use of Nonlinear Tools

32 Wilson C, Simpson SE, van Emmerik RE, 44 Georgoulis AD, Moraiti C, Ristanis S, Ster- 55 Goldberger AL. Non-linear dynamics for
Hamill J. Coordination variability and skill giou N. A novel approach to measure vari- clinicians: chaos theory, fractals, and com-
development in expert triple jumpers. ability in the anterior cruciate ligament de- plexity at the bedside. Lancet. 1996;347:
Sports Biomech. 2008;7:2–9. ficient knee during walking: the use of the 1312–1314.
Approximate Entropy in Orthopaedics.
33 Cavanaugh, JT, Guskiewicz KM, Giuliani 56 Goldberger AL, Amaral LA, Hausdorff JM,
J Clin Monit Comput. 2006;20:11–18.
C, et al. Recovery of postural control after et al. Fractal dynamics in physiology:
cerebral concussion: new insights using 45 Newell KM. Degrees of freedom and the alterations with disease and aging. Proc
approximate entropy. J Athl Train. 2006; development of center of pressure pro- National Acad Sci USA. 2002;99:

Downloaded from https://academic.oup.com/ptj/article/89/3/267/2737718 by University of Wisconsin-Madison Libraries user on 24 October 2023

41:305–313. files. In: Newell KM, Molenaar PMC, eds. 2466 –2472.
Applications of Nonlinear Dynamics to
34 Cavanaugh JT, Guskiewicz KM, Stergiou 57 Goldberger AL, Rigney DR, West BJ. Chaos
Developmental Process Modeling. Hills-
N. A nonlinear dynamic approach for eval- and fractals in human physiology. Sci Am.
dale, NJ: Erlbaum;1997:63– 84.
uating postural control: new directions for 1990;262:42– 49.
the management of sport-related cerebral 46 Cavanaugh JT, Coleman KL, Gaines JM, 58 Ahn AC, Tewari M, Poon C, Phillips RS.
concussion. Sports Med. 2005;35:935–950. et al. Using step activity monitoring to The limits of reductionism in medicine:
characterize ambulatory activity in
35 Moraiti C, Stergiou N, Ristanis S, Geor- could systems biology offer an alternative?
community-dwelling older adults. J Am
goulis AD. ACL deficiency affects stride-to- Plos Med. 2006;3:e208.
Geriatr Soc. 2007;55:120 –124.
stride variability as measured using non- 59 Przybylskia A, Baranowskia R, Zebrowskib
linear methodology. Knee Surg Sports 47 Buzzi UH, Ulrich BD. Dynamic stability of JJ, Szweda H. Verification of implantable
Traumatol Arthrosc. 2007;15:1406 –1413. gait cycles as a function of speed and sys- cardioverter defibrillator (ICD) interven-
tem constraints. Motor Control. 2004;8:
36 Kurz M, Stergiou N. The aging human neu- tions by nonlinear analysis of heart rate
241–254.
romuscular system expresses less cer- variability: preliminary results. Europace.
tainty for selecting joint kinematics during 48 Kurz M, Stergiou N. Do horizontal propul- 2004;6:617– 624.
gait. Neurosci Lett. 2003;348:155–158. sive forces influence the nonlinear struc- 60 Bartsch R, Plotnik M, Kantelhardt JW,
ture of locomotion? J Neuroeng Rehabil.
37 Slifkin AB, Newell KM. Noise, information et al. Fluctuation and synchronization of
2007;1:4 –30.
transmission, and force variability. J Exp gait intervals and gait force profiles dis-
Psychol Hum Percept Perf. 1999;25: 49 Yoshino K, Motoshige T, Araki T, Mat- tinguish stages of Parkinson’s disease.
837– 851. suoka K. Effect of prolonged free-walking Physica A. 2007;383:455– 465.
fatigue on gait and physiological rhythm.
38 Riley MA. Turvey MT. Variability of deter- 61 Schmit JM, Riley MA, Dalvi A, et al. De-
J Biomech. 2004;37:1271–1280.
minism in motor behavior. J Mot Behav. terministic center of pressure patterns
2002;34:99 –125. 50 Yamada N. Chaotic swaying of the upright characterize postural instability in Par-
posture. Hum Mov Sci. 1995;14:711–726. kinsons disease. Exp Brain Res. 2006;168:
39 Balasubramaniam R, Riley MA, Turvey MT. 357–367.
Specificity of postural sway to the de- 51 Harbourne RT, Deffeyes JE, DeJong SL,
mands of a precision task. Gait Posture. et al. Nonlinear variables can assist in iden- 62 Hausdorff JM. Gait variability: methods,
2000;11:12–24. tifying postural control deficits in infants. modeling, and meaning. J Neuroeng Reha-
J Sport Exerc Psychol (Suppl). 2007;29:S9. bil. 2005;2:1–9.
40 Oppenheim U, Kohen-Raz R, Alex D, et al.
Postural characteristics of diabetic neurop- 52 Russell D, Rosenbaum P, Gowland C, et al. 63 Winstein CJ. Knowledge of results and mo-
athy. Diabetes Care. 1999;22:328 –332. Gross Motor Function Measure. 2nd ed. tor learning: implications for physical ther-
Hamilton, Ontario, Canada: McMaster Uni- apy. Phys Ther. 1991;71:140 –149.
41 Pincus SM, Gladstone IM, Ehrenkranz RA. versity; 1993
A regularity statistic for medical data anal- 64 Farley BG, Koshland GF. Training BIG to
ysis. J Clin Monit. 1991;7:335–345. 53 Miller DJ, Stergiou N, Kurz MJ. An im- move faster: the application of the speed-
proved surrogate method for detecting amplitude relation as a rehabilitation strat-
42 Pincus SM. Approximate entropy as a mea- the presence of chaos in gait. J Biomech. egy for people with Parkinson’s disease.
sure of system complexity. Proc Natl Acad 2006;39:2873–2876. Exp Brain Res. 2005;167:462– 467.
Sci U S A. 1991;88:2297–2301.
54 Hausdorff JM, Peng CK, Ladin Z, et al. Is
43 Newell KM, Van Emmerik REA, Lee D, walking a random walk? evidence for long-
Sprague RL. On postural stability and vari- range correlations in stride interval of
ability. Gait Posture. 1993;1:225–230. human gait. J Appl Physiol. 1995;78:
349 –358.

280 f Physical Therapy Volume 89 Number 3 March 2009

 Movement Variability and the Use of Nonlinear Tools

Appendix 1.
Working Definitions for Terms Needed to Understand Nonlinear Concepts

Attractor, attractor state: Behav- have a deterministic pattern, mean- Sinusoidal: A regular waveform that
iorally, an attractor is a preferred ing that their future dynamics are exactly repeats itself over time.
state (ie, walk, run). Mathematically, fully defined by their initial condi-

Downloaded from https://academic.oup.com/ptj/article/89/3/267/2737718 by University of Wisconsin-Madison Libraries user on 24 October 2023

Stability, stable state: A rich behav-
an attractor is a set to which a dy- tions. This behavior is known as
ioral state characterized by high
namic system evolves after time. mathematical or deterministic chaos,
complexity. The stability of an orbit
Points that get close enough to the or simply chaos.
in a dynamic system determines
attractor remain close even if slightly
Noise: That part of a system descrip- whether nearby orbits remain near
perturbed. Geometrically, an attrac-
tion that is not deterministic. For or are repelled by that orbit.
tor can be a point, a curve, or even a
simplicity, it is usually assumed to
fractal structure known as a strange State space: The set of all possible
have a simple form, such as white
attractor. Describing the attractors of states of a dynamic system. When
noise.
chaotic dynamical systems is the fo- modeling a dynamic system, the
cus of the mathematical theory of Nonlinear measures: Measures number of variables needed to de-
chaos. that quantify the relationship, or de- scribe the system is called the dimen-
pendency, of the numbers through- sion of the state space.
Chaos: One subject in the field of
out the time series. Nonlinear mea-
nonlinear dynamics, which is part Systems theory: A field of science
sures describe the patterns or
of the broader field of dynamic studying the nature of complex
structure within the time series, not
systems. systems.
simply the quantity.
Complexity: Highly variable fluctu- Time series: Time series data are a
Periodic: Happening or appearing
ations in physiological processes re- specific example of an ordered list of
at regular intervals like a sine wave.
sembling mathematical chaos. numbers, where time is the parame-
Phase plane: A representation of ter that gives order to the list.
Constraints: Variability is intro-
the behavior of the dynamic system
duced into the system from the
in state space. Typically, it takes the Suggested Readings
constraints that dictate the system’s
form of a 2-dimensional plot of the Gleick J. Chaos: Making a New Science. New
behavior. The constraints are mor- York, NY: Penguin Books; 1987.
position X of the time series (on the
phological, biomechanical, environ- Smith LB, Thelen E. Development as a dy-
horizontal-axis) versus the first deriv- namic system. Trends Cogn Sci. 2003:7;
mental, and task specific.
ative X⬘ (on the vertical axis). 343–348.
Deterministic: For a given starting Stergiou N, Buzzi UH, Kurz MJ, Heidel J. Non-
Random: Lack of pattern or order; linear tools in human movement. In: Stergiou
condition, the future state of the N, ed. Innovative Analyses for Human Move-
lack of a relationship between points
system is determined; randomness is ment. Champaign, IL: Human Kinetics Pub-
in a time series or parts of a system. lishers; 2004:63–90.
not present.
Regularity: The repeatability of a Stergiou N, Harbourne R, Cavanaugh J. Opti-
Dynamic system: A system that mal movement variability: a new perspective
pattern. for neurologic physical therapy. J Neurol Phys
evolves over time. Ther. 2006;30:120 –129.
Self-organization: The formation Strogatz S. Sync: How Order Emerges From
Dynamic systems theory (DST):
of moving patterns is a function of Chaos in the Universe, Nature, and Daily
An area of applied mathematics used Life. New York, NY: Hyperion; 2004.
the cooperation of all of the sub-
to describe complex dynamic systems.
systems and their interaction with
Linear measures: Measures that de- the environment; it is not centrally
scribe the central tendency or dis- coded or commanded.
persion of the values within a set
of numbers, such as the mean, the
range, and the standard deviation.
Mathematical chaos: The behavior
of several related systems many
times seems to be erratic, with no
order (ie, random). However, non-
linear measures demonstrate that
such variations are not random but

March 2009 Volume 89 Number 3 Physical Therapy f 281

Movement Variability and the Use of Nonlinear Tools

Appendix 2.
Proposed Principles of Nonlinearity in the Acquisition and Maintenance of Motor Skill

1. An optimal amount of variability 4. Because motor function is sensi- behavior or direct appropriate in-
is necessary for movement to be tive to initial conditions, each per- tervention to allow variability
functional and efficient; normal, son brings a slightly different set changes to affect function.

Downloaded from https://academic.oup.com/ptj/article/89/3/267/2737718 by University of Wisconsin-Madison Libraries user on 24 October 2023

efficient movement includes both of conditions to a motor problem,
deterministic and random char- and the optimal solution to that 7. Traditional measures of variability
acteristics, which can fluctuate problem may be unique to that are not equivalent to measures
within an optimal range. person. Therefore, therapists can- of complexity. For example, as
not “prescribe” the best motor the measure of standard deviation
2. Healthy motor control has charac- pattern or strategy that is com- increases, the measure of ap-
teristics of nonlinearity, including mon to all patients. proximate entropy can decrease.
the spontaneous generation of new Measures of complexity describe
patterns of movement, movement 5. Input into the system can drive the structure of variability in
possibilities that are sensitive to ini- the system into other possibilities new ways that can help quantify
tial conditions, and a limited ability for movement that are not pre- subtle movement changes or
to precisely predict future move- dictable. The input can come characteristics.
ment based on current status. from more than one system (ie,
not only motor but also sensory, 8. Complexity is necessary for sys-
3. If variability increases in a system cognitive, emotional, or social). tems to adapt to changing condi-
without enough variability, new tions; loss of complexity means
movement options can emerge 6. Measures of complexity can help decreased ability for adaptation.
spontaneously. predict the emergence of a new

Invited Commentary Daniela Corbetta

One of the ways in which progress structures. These structures, which ciency. Whether one is beginning to
in movement performance and coor- are not detected by traditional mea- learn a new skill, is an expert per-
dination traditionally has been as- sures of response variation (such as former, is experiencing a decline in
sessed is to measure the amount of standard deviations or coefficients performance, or is being rehabili-
movement variability across repeti- of variation), are critical for under- tated, variability is an inherent prop-
tions of the same movement. In their standing change and progress in erty of the process of movement reg-
perspective article, Harbourne and movement control. They provide di- ulation. In all of those cases,
Stergiou1 argue that progress in rect information on the regulation however, variability may have a dif-
movement coordination should processes used to control move- ferent structure. Again, this is why
rather be indexed by analyzing the ment. Harbourne and Stergiou also the structure of variability should be
hidden structure of movement vari- remark that variability does not al- looked at closely; patterns of variabil-
ability embedded in time series using ways decrease linearly as a function ity can directly inform the practitio-
time-dependent, nonlinear tools of learning time and control. In some ner about the control process in-
rather than by assessing the overall instances, and as further illustrated volved in the movement. In their
amount of variability. By focusing below, it can evolve in a nonlinear examples, Harbourne and Stergiou
their article on the time structure of fashion over learning time. remind us that even within the most
variability, the authors make a point skilled and most highly repeatable
of fundamental importance for our Harbourne and Stergiou tell us that behaviors, there is no stereotypy.
understanding of biological motion variability should not always be con- Rather, ongoing small variations in
and its implications for develop- sidered a reflection of movement er- movement are continuously gener-
ment, learning, and rehabilitation. ror or imperfect control. As they ated to tailor the movement to its
They show through clear examples point out, variability is an omnipres- goal and to respond to the ever-
that patterns with similar ranges of ent characteristic of biological mo- changing action context. Clearly, as
variation can contain very different tion, regardless of the stage of profi- they remark, lack of movement vari-

282 f Physical Therapy Volume 89 Number 3 March 2009