Nonlinear Dynamics, Psychology, and Life Sciences, Vol. 19, No. 4, pp. 489-510.

© 2015 Society for Chaos Theory in Psychology & Life Sciences

Training the Antifragile Athlete: A Preliminary Analysis

of Neuromuscular Training Effects on Muscle Activation
Dynamics
Adam W. Kiefer1 and Gregory D. Myer, Cincinnati Children’s Hospital
Medical Center, University of Cincinnati, Ohio

Abstract: Athletic injuries typically occur when the stable, emergent coor-
dination between behavioral processes breaks down due to external noise, or
variability. A physiological system that operates at an optimal point on a
spectrum of rigidity and flexibility may be better prepared to handle extreme
external variability, and the purpose of the current experiment was to examine
whether targeted neuromuscular training resulted in changes to the rigidity and
flexibility of the gluteal muscle tonus signal as measured with electromyography
prior to the landing phase of a drop vertical jump task. 10 adolescent female
athletes who participated in a targeted 10-week neuromuscular training
program and 6 controls participated, and their tonus dynamics were examined
with recurrence quantification analysis prior to training and after the 10-week
program. The dependent measures, percent laminarity (%LAM) and percent
determinism (%DET) were hypothesized to decrease following training, and
were submitted to a one tailed mixed-model ANOVA. The training group
exhibited a decrease in %LAM and %DET after training compared to pre-
training and controls. The present findings indicate increased metaflexibility
(i.e., greater intermittency and an increase in internal randomness) in tonus
dynamics following neuromuscular training, and have important implications
for the prevention of musculoskeletal injury in sport, specifically within the
context of external noise and antifragility.
Key Words: antifragile, intermittency, piecewise determinism, recurrence
quantification, electromyography
INTRODUCTION
Athletic performance is conditional on the stability and adaptability of
emergent coordination between perceptual-motor and physiological processes
across multiple behavioral scales. It requires the movement system to be
coordinated in a way that is appropriately flexible and adaptive to external
challenges so that behavior can remain stable in the face of a constantly
1
Correspondence address: Adam W. Kiefer, PhD, Cincinnati Children’s Hospital
Medical Center, 3333 Burnet Ave, MLC 10001, Cincinnati, OH 45229, USA. E-mail:
adam.kiefer@cchmc.org
489
490 NDPLS, 19(4), Kiefer & Myer
changing environment. Each external event, or challenge, introduces noise (i.e.,
variability) into the movement system and increasingly tests its flexible and
adaptive nature. Events that lead to unexpected changes in movement are one
such source of external noise, and can occur when a defensive player reacts and
quickly changes direction to shadow an offensive player in football, for
example. Unanticipated challenges (e.g., a basketball player misjudging a
rebound and changing her body posture during a jump that results in an off-
balanced landing), or unpredictable perturbations (e.g., a tipped volleyball
during a block-spike attempt that forces a player to quickly cut in the other
direction to “dig” the ball out before it hits the ground) are also examples of
external noise. In instances such as these a flexible system allows an athlete to
safely and successfully perform myriad complex maneuvers during competition
and to smoothly transition between strings of action states, thousands of times
throughout a career, in the face of an assortment of external fluctuations that
stress the system. When the noise is extreme, such as when an athlete has to
quickly change positions or land from a jump in a misaligned posture, or
fluctuations occur at an inopportune moment during a movement pattern,
however, the global health of the system can be put at risk. As a result, athletic
injuries most often occur when coordination breaks down due to severe external
fluctuations imposed on the movement system.
Unfortunately for athletes the presence and magnitude of external noise
is predominantly unexpected and, thus, the timing of injuries are inherently
unpredictable. This makes injury prevention strategies extremely difficult to
develop and implement with high success. For example, during a perceptually
misjudged basketball rebound the athlete might be forced to transition from a
vertical jump and reach to a vertical jump to the right and reach to the right, and
this could lead to an off-balanced posture. At the same time, an opponent might
charge in from the right to challenge for the ball. This would force a secondary
adaptive response from the rebounder such that as the ball is grabbed the athlete
lands with more force than intended on the right leg due to destabilized trunk
coordination. Disrupted trunk position, combined with altered postural control
due to the external noise incurred by the movement system, drives an
asymmetrical lateral load to the support leg that accelerates geometric increases
in valgus load on the knee. The result could be a rupture of the anterior cruciate
ligament (ACL)—a primary stabilizer of the knee joint (Krosshaug et al., 2007).
In a case such as this, a developed injury-risk profile might indicate that such an
occurrence is more or less likely prior to the injury occurrence due to the
identification of high-risk biomechanics, the presence of limb asymmetries,
limited core strength, and/or reduced neuromuscular control (Hewett et al.,
2005; Myer, Ford, Brent, & Hewett, 2012a; Myer, Ford, Khoury, Succop, &
Hewett, 2011; Sugimoto, Myer, McKeon, & Hewett, 2012). However, the
occurrence of that particular event at that particular moment in time is inherently
unpredictable. Thus, in order to avoid potential injury the system must be
adaptable to unpredictable challenges and fluctuations in order to preserve
health and successful movement performance.
 NDPLS, 19(4), Antifragile Athlete 491
Piecewise Determinism and Intermittency
It is known that adaptable physiologies exhibit particular dynamic
signatures of behavior (Glass & Mackey, 1979; Hausdorff & Peng, 1996;
Mackey & Glass, 1977); see West (2006) for a review). Specifically, the dynam-
ics of healthy systems exhibit a blend of flexibility and rigidity (i.e., piecewise
determinism) that enables the system to transition through optimal local states of
behavior as the global, overt behavior flows through the environment and
responds to any external noise or fluctuations that it faces (Van Orden, Kloos, &
Wallot, 2011; Warren, 2006). Piecewise determinism is behavior that changes
discontinuously and is mathematically characterized through violations of
Lipschitz conditions (cf., Zbilut, Dixon, & Zak, 2002). It is typically observed in
overt behaviors that start, stop and repeat themselves (Van Orden et al., 2011).
For example, an athlete runs down a field, cuts to the left and quickly performs a
vertical jump and landing. These individual, overt behavioral states (the run, cut,
jump and landing) are sewn together with smooth, precisely timed transitions
that produce an overall run-cut-jump-land pattern that appears seamless to the
observer. However, at the local level precisely timed neural activations occur at
various magnitudes in each individual muscle throughout the movement string.
At a broader scale, individual muscles form muscle synergies—multiple,
individual muscles that coordinate with one another to form single, autonomous
structures (Kugler, Kelso, & Turvey, 1982; Turvey, 1990)—and these muscle
groups serve to coordinate the various body segments throughout the time
course of the more overt motor behavior.
The interconnectedness of the various behavioral scales requires
flexible, adaptive behavior at a very local level that permeates into the more
global behavioral scales. In order to maximize adaptability throughout the
system these local dynamics need to live at an optimal point on a spectrum of
metaflexibility (Pincus & Metten, 2010)—a term that we adapt to describe the
ability of a system to respond to a particular challenge through a rapid transition
to a new state via either more flexible or more rigid underlying dynamic
processes. Accordingly, the dynamics should exhibit a blend of flexibility and
rigidity that, in turn, would give rise to impermanency of local behavioral states
prior to the onset of a given movement. This impermanency is a necessary
precursor to observed adaptability at a more global level (as identified through
the presence of intermittency). Intermittency (i.e., 1/f noise) is the imper-
manency of states and is thought to emerge from the nonlinear interactions
among interdependent system processes at multiple scales (i.e., interaction-
dominant dynamics), and may be emblematic of system-wide metaflexibility
(Diniz et al., 2011; Pincus & Metten, 2010). Thus, there is likely an optimal
level of impermanence that a system exhibits as it remains poised for a multi-
tude of behavioral possibilities prior to a given action. For example, an under-
lying muscle physiology that consists of longer, more rigid states of neural firing
behavior (i.e., becomes trapped in a given state) may be less flexible, or slower
to transition to appropriate firing patterns, when faced with external fluctuations.
492 NDPLS, 19(4), Kiefer & Myer
It is possible that these arre all characterristics of an anntifragile systeem
(Taleb,, 2012)—a con ncept within which
w the answ wer to injury pprevention migght
ultimattely lie. Taleb introduces the term antifragiile, instead of robust, as a terrm
m that cannot adapt), and ddiscusses an annti-
antithettical to fragilee (i.e., a system
fragile system as onee that is "…beeyond resiliencce or robustneess. The resilieent
resists shock and streess and stays th he same; the anntifragile gets bbetter" (p. 3). H
He
furthers this argumen nt by offering th hat perfect robbustness is unatttainable and thhat
resilien
nce can only taake a system so far before an outlier, or extreme randoom
fluctuaation, forces thee system into a different (e.gg., negative) sttate (Fig. 1). T The
reason for this is thatt if a fluctuatioon is not predicctable and, theerefore, has nevver
been ex xperienced by the system, th he fluctuation w will fall outsidde of the adaptiive
range ofo the system because the system s has nevver been traineed to respond to
such a rare event.

Fig. 1. A depiction of
o a series of potential torqu ue values that a system mig ght
withstand (white barss) prior to an extreme
e outlierr (black swan)) that occurs a
and
causess a catastrophic
c failure to the system
s (i.e., an
n ACL tear).
Before mov ving forward, it is importannt to delineatee the differencces
betweeen resilience an nd antifragility. While robusttness has been widely used aand
is a faiirly straightforrward concept— —a robust sysstem is one thhat can withstaand
b not necessaarily adapt to it—resilience is not as easilly understood. In
stress but
systemm engineering, resilience
r has been defined as a system thhat has a geneeric
ability to cope with unforeseen
u chaallenges througgh the availabiility of adaptabble
reservees and flexibiility to accom mmodate thosee challenges ((Nemeth, 2008).
Similarrly, Woods and d Wreathall (2008) discuss thhat resilience mmay refer to hoow
well ann organization, or system, can n transition beetween regionss of a state spacce,
and that the transitiions work mo ore efficiently as energy annd resources aare
investeed. Pincus and Metten (2010 0) have discusssed resilience iin the context of
metafleexibility, and ini a way that isi more in linee with the focuus of the curreent
work. They define metaflexibility
m as the abilityy of a system to respond too a
perturbbation by becom ming both morre rigid and robbust, or more fl
flexible and fluuid,
 NDPLS, 19(4), Antifragile Athlete 493
while preserving the whole, integrated system. They further argue that resilience
plays a role in this process such that it is the property that allows the system to
loosen back up or come back together following external fluctuations. Another
way to think about this is that resilience allows for the reorganization of
components, or a change in the interaction processes, within and between the
various behavioral scales. This is similar to the way it has been considered in
catastrophe models (Karwowski, Ostaszewski, & Zurada, 1992), as the response
surfaces can be considered to be malleable or, as Pincus and Metten put it,
rubberized: that is in the context of bio-psycho-social modeled processes, “…the
response surface of the cusp could shift to include bifurcation thresholds that
were so minor that the two point attractor regions at the high end of the
bifurcation parameter would lose their distinct separation from one another.
Such a system would be high in “global resilience” (Pincus & Metten, 2010, p.
373; see also Guastello & Liebovitch, 2009).
Taleb (2012) essentially treats robustness and resilience interchange-
ably, perhaps less because he equates the two concepts, but more because
neither do enough work for him and his theory of antifragility, so they are equal
in their limited utility. This is because an antifragile system goes beyond resil-
ience in that it not only allows the system to adapt to fluctuations (e.g., reorgan-
izes the way in which the system components are linked), but more importantly
the system actually gains strength from them (e.g., the links between the system
components become stronger, overall): an important consideration when making
an antifragile argument, and one that has critical implications for injury
prevention training.
Training Antifragility for Injury Prevention
We have previously developed neuromuscular training (NMT) as a
mechanism to mediate the confluence of biomechanical events that lead to lower
limb musculoskeletal injuries and, chiefly, non-contact ACL injuries (Myer,
Ford, Palumbo, & Hewett, 2005). The logic behind the formulation of NMT is
that through the modification of mechanics known to predispose athletes to ACL
injury the system can become more adaptable to external challenges (Hewett et
al., 2005). From the outset, NMT has consisted of core strength, resistance and
balance training in combination with more dynamic exercises such as
plyometrics, speed, and movement training in order to regulate the full body
mechanics for proper technique and performance (Hewett, Lindenfeld,
Riccobene, & Noyes, 1999; Hewett, Stroupe, Nance, & Noyes, 1996). In its
early form, NMT built robustness into the system. That is, the athlete became
more resistant to external noise given learned biomechanical patterns that were
consistently replicated throughout competition. These general principles of
NMT prescribe exercises so that the athlete can first gain control of fundamental
movement patterns as part of a progression toward more complex movements.
This, in turn, helps influence transfer to, and potentially injury risk reduction on,
the field of play. However, resistance does not necessarily make the movement
system more adaptable to external noise. By 2008, we learned from the evidence
494 NDPLS, 19(4), Kiefer & Myer
to further develop the training systems (Myer, Ford, McLean, & Hewett, 2006;
Myer et al., 2005; Myer & Ford, 2006) to target trunk and hip stabilizers—two
biomechanical mechanisms of ACL injury—in order to counterbalance trunk
motion via unilateral loading, e.g., using a medicine ball to weigh down the
trunk on the right side during a single leg lunge with the left leg (Myer, Chu,
Brent, & Hewett, 2008). Importantly, the integration of late phase progressions
with unanticipated movements (i.e., artificially introducing external noise) chal-
lenges and perturbs the neuromuscular system throughout the NMT protocol.
Moreover, the exercises are designed to train the system to be better prepared to
respond to, and gain from, volatility while at the same time honing the athlete's
overall neuromuscular control towards a successful performance outcome. Thus,
NMT in its current form is intended to build robustness in movement patterns,
and then introduce external fluctuations to the system in an effort to increase the
metaflexibility, and ultimately the antifragility, of the movement system.
An important and open question still remains as to whether NMT, as
well as the external challenges introduced as part of the NMT progression,
fundamentally alter the local dynamics of a movement system to exhibit a more
flexible dynamic signature as a result of motor learning. This is an important
first step to eventually understand the nature of flexibility and adaptability as it
pertains to injury prevention. In the present experiment we elected to examine
the dynamics of the muscle tonus—the low-level activity of the muscle—prior
to a land and jump movement state. Given that the environment and associated
challenges shape the measureable behavioral dynamics of an athlete, and may
overwhelm any measurement of the system's tonus activity, the muscle tonus
dynamics as measured prior to a given action state may identify the region
where the system lives on the spectrum of metaflexibility. From a nonlinear
dynamics perspective, intermittency specifically, might be used to pinpoint the
metaflexibile signature of muscle tonus.
Adolescent female athletes took part in either an NMT hip and trunk
focused training intervention in conjunction with their normal offseason strength
and conditioning activities or, alternatively, their normal offseason strength and
conditioning activities only (Myer, Brent, Ford, & Hewett, 2008; Myer, Chu et
al., 2008). The purpose of the current investigation was to evaluate the
adaptability of gluteal muscle tonus following targeted neuromuscular training.
It was hypothesized that during the performance of a drop vertical jump as
tested prior to and after NMT (relative to normal offseason activities), the
gluteal tonus dynamics of the athletes that participated in NMT would exhibit
increased intermittency (as measured by Recurrence Quantification Analysis;
see method section) in the pre-contact phase of landing after NMT compared to
their pre-test, and also compared to the control athletes’ post-test following 10
weeks of normal offseason training. We also hypothesized that increases in
intermittency would be accompanied by concomitant increases in random
processes (an additional potential indicator of increased flexibility) that make up
the tonus dynamics for the NMT athletes during their post-test when compared
to their pre-test and the post-test of the control athletes.
 NDPLS, 19(4), Antifragile Athlete 495
METHOD
Participants
Twenty-six high school volleyball players were recruited and, prior to
an initial testing session, were divided into two groups: an NMT group (N = 19)
and a no training control group (N = 7). The NMT athletes participated in an
NMT intervention twice a week over a 10-week period in addition to standard
off-season strength training that took place once a week. Pre-testing took place
one week prior to the first day of training, and post-testing took place approxi-
mately 11 weeks after the pretest on all participants. Due to participant drop-out
or failure to meet minimum compliance requirements (i.e., participation in at
least two-thirds, or 14 of 20, training sessions), in combination with missing
post-test data as a result of testing session dropout, data from 10 NMT athletes
(Age M = 15.50 ± 1.64 years; Height M = 170.40 ± 4.93 cm; Weight M = 63.20
± 9.98 kg) and 6 control athletes (Age M = 15.66 ± 1.58 years; Height M =
173.33 ± 10.97 cm; Weight M = 63.65 ± 5.02 kg) were included in the present
analysis.
Procedure
All participants provided informed consent prior to participation in the
first testing session, and the Institutional Review Board approved all procedures.
Surface EMG Measurements
For the current study, we elected to evaluate the gluteus medius given
that it is a key modulator of hip neuromuscular control. Surface EMG activity
was continuously recorded at the right gluteus medius muscle on all participants
through a surface telemetry EMG system (Telemyo 2400 R G2, Noraxon, USA).
Disposable, self-adhesive Ag/AgCl snap electrodes (2 cm inter-electrode
distance) were connected to a transmitter attached to the waistband of all
athletes. The surface EMG leads had 1st order high-pass filters set to 10 Hz ±
10% cutoff. The signals were bandpass filtered (7-500 Hz), differentially
amplified with a common mode rejection ratio > 130 dB, a gain of 1000, a noise
level equal to 1μV, a signal-to-noise ratio > 300, and were sampled at 1200 Hz.
Drop Vertical Jump
At week one and week 11 of the experiment, the EMG activity of all
athletes was assessed during the performance of a drop vertical jump task, as
part of a larger testing battery. The drop vertical jump is a well-defined and
well-understood test of neuromuscular control and has been demonstrated to
identify athletes at risk for future primary and secondary ACL injury (Ford,
Myer, & Hewett, 2003; Hewett et al., 2005; Paterno et al., 2010). In the present
experiment, the athletes stood on a 30 cm tall box and were instructed to drop
off the box, leave both feet at the same time, land, and then immediately per-
form a maximum vertical jump. Three trials were performed during each testing
session. Prior to performing the drop vertical jump, each participant was instru-
496 NDPLS, 19(4), Kiefer & Myer
mentedd with retrorefflective markers, as previouusly described (Ford, Myer, &
Hewettt, 2003). Two force
f platformss (AMTI, Wateertown, MA) w were positionedd 8
cm apaart to ensure th
hat each foot coontacted a sepaarate force plaatform. The forrce
data waas collected in sync with the EMG data at 11200 Hz. The first contact onnto
the plattforms ≥ 10 N was used to iddentify the startt of the landingg phase.

Fig. 2. Three example e techniques foor how externa al noise was introduced into tthe
perceptual-motor sys stem througho out the NMT intervention. The athlete: (1)
balancees on a single leg on a Bosu uTM Balance Trrainer while sh e simultaneously
must reeact to and ca atch a ball tha
at is thrown fro om different locations (left), (2)
kneels on a SwissballTM and mainta ains balance w while the instrucctor kicks the b ball
at unprredictable interrvals and locaations (middle) , and (3) perfo orms a series of
tuck jum
mps during which, at an unprredictable time while the athle ete is in flight, tthe
trainer points right orr left and as soon
s as the atthlete lands sh he is required to
perform
m a 45° cut in th
he direction tha
at the instructo r pointed (rightt).
Training Interventio
on
The trunk and
a hip focuseed neuromuscuular training iintervention w was
used too target trunk control deficits and improvve hip strengthh and power ((cf.
Myer ete al., 2008). Thhe intervention n was compriseed of five phasses for each exer-
cise to promote progrression in levell of difficulty, and the exerciises were desiggn-
ed to perturb
p the tru
unk through an n increase in eexternal noise during standaard
training
g activities to facilitate the athletes’
a contrrol of the trunnk and concom mit-
antly immprove “core stability” during dynamic acctivities. Withhin the individuual
phases,, the intensityy of each exerrcise progressiively increaseed to continuaally
challennge the athletess. End-stage prrogressions inccorporated impposed challengges
to the lateral
l trunk an
nd forced the athletes
a to deceelerate and conntrol the trunkk in
order to
t successfully y execute the trained
t techniqques. For exam mple, during laate
stage progressions
p off the single legg balance task on a BosuTM Balance Trainner,
athletess were required to react to anda catch a balll that came tooward them froom
unpreddictable location ns (Fig. 2, leftt). Athletes alsso balanced onn their knees onn a
SwissbballTM while a trainer kicked the ball at unnpredictable inttervals and locca-
tions to
o force the athllete’s core muscles to remainn poised for suuch perturbatioons
(Fig. 2, middle). Thee most demand ding of tasks w was a reactionaary cut followiing
the landing of a tuck jump. Specificcally, as athlettes repeatedly pperformed a tuuck
jump, the
t instructor would
w point eitther right or leeft in an unpreedictable mannner,
while the
t athlete waas in the flightt phase and noot in contact w with the grounnd.
 NDPLS, 19((4), Antifragilee Athlete 4497
Based on the directiion the instrucctor pointed, tthe athlete hadd to prepare hher
posturee and muscularr activations ass she came dow wn into a landding position aand
immed diately executed d a 45° cut in the direction tthat the instrucctor pointed (F
Fig.
2, righht). Training progressions
p such
s as these served to inntroduce externnal
fluctuaations, or noisee, into the perceptual-motorr system in orrder to facilitaate
changees in the dynam mics of muscle activation andd, ultimately, inn the tonus dynna-
mics off crucial musclle groups (i.e., gluteus mediuus) related to trrunk and hip coon-
trol. Alll of the initial exercises weree adapted from
m previous studdies that reportted
ACL in njury risk redu uction (Hewett et al., 1996; M Mandelbaum, 22005; Myklebuust
et al., 2003;
2 Petersenn et al., 2005),, with progresssions developeed from previoous
researcch that demonstrated post-trraining reductiion in knee abbduction load in
female athletes (Myeer, Ford, Brentt, & Hewett, 22007; Myer et al., 2006, 20005;
Myer & Ford, 2006).
Data Redu
uction and An
nalysis
The data of interest was th
he raw EMG ssignal, or tonuus, of the gluteeus
mediuss 100ms prior to landing. Th hus, each EMG G time series wwas not subjectted
to addiitional filtering
g or smoothin
ng (cf. Liu, Kaankaanpää, Zbbilut, & Webbber,
2004), and all time seeries were 120 data points in length (Fig. 3)).

Fig. 3. A sample EM MG time series s for the gluteu

us medius durring a single trrial
drop veertical jump tria
al (bottom pane
el), and the 100
0 ms epoch fro
om the same tim me
series that correspon nds to the 1000 ms prior to initial landing during the drrop
verticall jump (gray highlighted portion in bottom ppanel, and entire time series of
top pannel).
Recurrrence Quantiffication Analysis (RQA)
RQA is a univariate
u non
nlinear analysiss that is usedd to quantify tthe
presencce or absence and nature of recurrent paatterns, or repeeating segmennts,
498 NDPLS, 19(4), Kiefer & Myer
within a single time series. The computational details of RQA have been
discussed in detail elsewhere (Webber & Zbilut, 2005; Marwan, Romano, Thiel,
& Kurths, 2007). RQA is a multistep process in which the time series is first
embedded in a multi-dimensional phase space. Next, the data are analyzed to
identify when the time series visits similar states, or recurrences, within the
reconstructed phase space, and these recurrences are quantified by consideration
of how numerous they are and based on their temporal organization.
RQA requires the selection of several parameters in order to optimize
the analysis technique. The first is the embedding dimension (m) that is specified
to unfold the EMG time series (x) into a higher-dimensional space of dimension
m. This step serves to remove distortions in the data that result from the
projection of the data from a high-dimensional system onto a one-dimensional
axis of measurement. In combination with the second parameter, time delay (t),
a set of m vectors is constructed where the first vector is a copy of the time
series and each subsequent vector is a copy of the previous one offset by t
(truncated to maintain equal lengths). A 6-dimensional space was used in the
present analysis, based on a false nearest neighbors analysis on the original
signal and time-delayed copies, x(t), x(t + delay), x(t+{2 x delay]), and so on. A
delay of 10 data points was used to guarantee the minimal average mutual
information between points separated by that distance so as to obtain nearly
orthogonal dimensions in embedding space. This process allows the dynamics of
the signal to unfold in the appropriate number of dimensions so that the under-
lying dynamic processes are better revealed. Next, the trajectory of data points
of the time series in the m-dimensional phase space is compared to determine
when they intersect within some tolerance as determined by the radius (r). A
hyper-sphere with radius r is positioned at each value of the time series, and for
each of those values and other values that fall within the hyper-sphere are deter-
mined. Each instance of a data point that falls within the radius is a single recur-
rence. Thus the radius essentially acts as a sensitivity parameter for the identifi-
cation of recurring points in the reconstructed data. The ordering of points in
phase space indicates time points of common configurations within the given
time series, and for example, represents when the EMG signal revisits the same
activity pattern(s) over time. The results of the recurrence calculations are visu-
ally represented on a recurrence plot, or RP (Fig. 4)—a two dimensional plot of
time points of xi vs. time points of xj in which a darkened pixel at each (xi, xj)
coordinate where a recurrence was identified. The radius of inclusion for points
to be considered recurrent in the reconstructed phase space was 1.25% of the
actual Euclidean distance separating points in reconstructed phase space.
Once the RP is generated, the recurrences can be quantified based on
the structure present in the RP, and specifically the number and temporal
distribution of the recurrences. While RQA typically outputs several variables of
interest, two quantitative measures were of primary importance in the present
experiment: %LAM (percent laminarity) and %DET (percent determinism).
%LAM is the percentage of darkened pixels that fall along vertical lines in the
RP. Vertical lines mark time intervals in which a state does not change or
 NDPLS, 19((4), Antifragilee Athlete 4499
changees very slowly (Marwan
( et al.., 2007)—in otther words, thee state of the tim
me
series is
i trapped for some period of o time; a typee of behavior tthat is typical of
laminarr states (deriv ved from the stacking of m multiple paralleel trajectories as
lamina; Zbilut & Weebber, 2008) and a related to intermittency.. Hence, %LA AM
will deecrease if the RPR consists off more single recurrences annd fewer verticcal
line strructures, indicative of fewerr laminar phasses. %DET is computationaally
analogo ous to %LAM M, but based on n the recurrent points that foorm diagonal liine
structurres. Thus, it iss the percentag
ge of darkenedd pixels that falll along diagonnal
lines in
n the RP and is a measure off the degree too which the futture states of tthe
time seeries are govern ned by previouus and present states (Riley & Turvey, 2002).
Similarr to %LAM, %DET will decrease d if thee RP consists of more singgle
recurreences and feweer diagonal linee structures, inndicative of a ggreater degree of
randommness in a tim me series. All RQAR results w
were computedd using the crooss
recurreence plot toolbo ox (Marwan ett al., 2007).

Fig. 4. A sample EM MG time series s and correspo

onding recurre
ence plot for o
one
particip
pant during a siingle experimental trial.
Statisstical Analysiss
The dependeent measures %LAM
% and %%DET were subbmitted to a onne-
tailed 2 × 2 mixed-mmodel ANOVA A with group ((No training vvs. Training) ass a
betweeen-subjects facttor and time of test (pre- vs.. post-test) as a within subjects
factor (α = .05). Paiired-samples t--tests were useed for follow--up comparisonns,
when appropriate.
a Peercent change scores were allso computed for both %LA AM
and %D DET, and subm mitted to separrate one-tailed independent ssamples t-tests (α
= .05).
500 NDPLS, 19(4), Kiefer & Myer
RESULTS
R
See Fig. 5 for
f sample timee series and reecurrence plotss during both tthe
nd post-test forr an athlete th
pre- an hat participatedd in NMT (botttom) and for an
athlete that did not (to
op).

Fig. 5. Sample EMG G time series and

a associated
d recurrence plots for a conttrol
particip
pant (top) and an a NMT participant (bottom) during a pre-te est trial (left) a
and
a post-test trial (right)). Note the qua
alitative change
e in the tonus EEMG times seriies
from prre- to post-test,, and the resulttant recurrence
e plot for the NM
MT participant..
A significan
nt time of test × group interacction F(1,14) = 3.24, p = .0447,
η2 = .1 19 was observ ved for %LAM M (Fig. 6, topp). Follow-up tests revealedd a
significcant decrease for
f %LAM in the training grroup during thhe post-test (M M=
0.58 ± 0.03), comparred to their pree-test (M = 0.880 ± 0.01), andd the post-test of
the con ntrols (M = 0..81 ± 0.01). Similarly,
S the aanalysis of %DDET indicatedd a
significcant time of teest × group intteraction F(1,114) = 3.87, p = .035, η2 = ..22
(Fig. 6,, bottom), and follow-up testts indicated a cconcomitant deecrease of %DE ET
in the training group p during the po ost-test M = 00.51 ± 0.03) coompared to thheir
 NDPLS, 19((4), Antifragilee Athlete 5501
own prre-test M = 0.733 ± 0.01), and the post-test off the controls M = 0.75 ± 0.01).
There were
w no other significant
s effeects.
Significant decreases
d in th he percent ch ange of both %LAM, t(14)) =
1.99, p = .034, and %DET, t(14) = 2.15, p = .025, were alsso observed; tthe
g group exhibitted a negative percent changee (%LAM M = -26.06 ± 3.833%
training
and %D DET M = -29.9 96 ± 4.21%), whereas
w the coontrol group didd not (%LAM M
= 18.633 ± 8.59% and %DET M = 23 3.18 ± 9.50%)..

Fig. 6. A time of test × group interac

ction for %LAM
M (top) and %DET (bottom).
DIISCUSSION
NMT is dessigned to increease both the robustness andd adaptability of
athletess’ movement behaviors
b to ex
xternal challennges that they fface during com-
petition
n. Early progreessions of NMT T focused on llearning stereootyped movemeent
pattern
ns to make the athlete more robust to exteernal noise. Addvanced progrees-
sions have
h built on thhis foundation and emphasizee controlled, bbut unpredictabble,
externaal fluctuations as the athlete trains such thaat she must leaarn to coordinaate
and con ntrol her behav vior to encouraage movement patterns that ppersist in the faace
of volaatility and, ultiimately, gain from
f it (i.e., aantifragility). W
We hypothesizzed
that su
uch an adaptab ble system wou uld be identifiied by its tonuus dynamics aand
charactterized by impeermanency of local states (i.ee., increased inntermittency) aand
a blendd of random an nd deterministtic processes (ii.e., piecewise determinism) in
the EMMG signal. To test whether NMT N gave ris e to such channges in the loccal
musclee physiologies,, we compared d the tonus dy dynamics of addolescent femaale
athletess prior to the laand and jump phase
p of a dropp vertical jumpp before and affter
10 weeeks of NMT, an nd also with control particippants who did nnot participate in
502 NDPLS, 19(4), Kiefer & Myer
NMT. The athletes that participated in NMT exhibited more random, intermit-
tent tonus dynamics after NMT compared to both their own pre-training perfor-
mance and the post-test performance of athletes who participated only in normal
offseason activities. Thus, the results supported our hypotheses and provide evi-
dence of a fundamental change to the tonus dynamics of muscles following tar-
geted neuromuscular training as measured prior to a specific action state.
A prominent view of human movement is that motor behavior emerges
from the combined deterministic and random processes present internal and
external to the behavioral system (Diniz et al., 2011; Kelso & Ding, 1993; Riley
& Turvey, 2002). Stable movement patterns (represented as trajectories of
dynamic state variables in state space) are governed by attractor states that arise
from solutions to motion equations (i.e., a function of the current state of the
system)—a dynamical system (Warren, 2006). These stable solutions, and
therefore the behavior of the system as a whole, are contingent on the optimal
blend of deterministic and random processes (i.e., piecewise determinism),
which give rise to stable behavior. Changes in or transitions between patterns of
coordination, or movement patterns more generally, arise due to the influx of
random noise and lead to instability. Specifically, the increased random
fluctuations push the system out of a stable arrangement (Kelso & Ding, 1993),
the attractor destabilizes, and the system falls into an impermanency of states as
it searches for a solution to the next movement pattern. These transitions are
accompanied by increased movement variability exhibited by the motor system
due to the influx of random noise. As the system converges on the next stable
solution the variability decreases and stability is achieved once again, and as the
system settles into a new stable solution a different blend of random and deter-
ministic processes is present and optimized for the new, stable coordination
regime. Thus, a system that is inherently more flexible, up to a point, will likely
transition more easily between different movements. More specifically, a system
with tonus dynamics that exhibit more intermittent behavior can remain poised
to quickly transition into new firing patterns, and thus a variety of movement
states, and for that reason is likely more adaptable to external noise.
In light of this view, the present findings have the potential to optimize
current approaches in the development of training interventions for the preven-
tion of musculoskeletal injury in sport. Given that the timing of injuries are not
inherently predictable, prevention strategies are extremely difficult to develop
and implement with high success (Sugimoto et al., 2012). While biomechanical
injury-risk profiles can help identify high-risk athletes, the specific injury event
is unpredictable—it is, in effect, a black swan event (Taleb, 2007). For example,
when an athlete incurs an ACL injury such an outcome is an extreme outlier
from prior typical movement outcomes, and it is an event that carries with it an
extreme impact on the athlete’s current and future health.
We are not the first to discuss ACL injuries in this way (cf. Hewett,
2009). Hewett’s main thesis was that neuromuscular imbalances should be used
to screen for at-risk athletes and then channel them into NMT programs that tar-
get the internal variability inherent to each athlete. He argues for this in the con-
 NDPLS, 19(4), Antifragile Athlete 503
text of a gray swan—an event that can be anticipated to a certain degree—and,
over the years, variations of this approach have contributed to a nearly 50%
reduction in non-contact ACL injury for female athletes (Myer et al., 2013).
Likewise, successful biomechanical approaches (Hewett et al., 2005, 1999;
Myer, Brent, et al., 2008; Myer, Chu, et al., 2008; Myer et al., 2012a; Myer,
Ford, Brent, & Hewett, 2012b) for risk reduction utilize targeted training to
reduce movement variability and create a movement system that is robust to
predictable contexts (i.e., planned runs, cuts, jumps and landings) in the absence
of extreme external noise. In a sense it serves to “narrow the window” in which
a non-contact ACL injury can occur under most conditions. However, this
approach is only the foundation, as the results of these efforts still leave a large
number of athletes who will be affected by an ACL injury with no apparent
remedy. Thus, the idea of an ACL injury as a black swan event must be pushed
even further to reach its ultimate potential as a theoretical basis for how we
approach injury prevention.
Fundamentally, the categorization of an ACL injury as a black swan
event allows us to rethink the meaning of the question, "How do we prevent an
ACL injury?" Given the unpredictable nature of an ACL injury event, linear
causal approaches that attempt to predict the event itself are fruitless.
Interventions designed to address (bio)mechanistic links made to ACL injury in
the context of extreme external fluctuations have the potential to lead us into the
same trap that Taleb (2007) adamantly argues we should avoid (building robust
systems based on linear additivity). Instead, training interventions must be
expanded to prepare the athlete’s internal movement system to avoid injury
when faced with unpredictable, external fluctuations and noise during athletic
actions. Such a strategy would overcome the limitations of current approaches
that focus on building interventions based on linearly causal models of injury
risk (cf. Myer et al., 2013 for a review). It would also expand on current NMT
approaches for prevention by taking a movement system beyond robustness and
to the edge of chaos—to a system that exhibits disorder.
The ultimate answer to prevention likely lies in Taleb’s concept of
antifragility (Taleb, 2012), which allows for the system to gain from volatility.
Specifically, to prevent the undesirable consequences of unknown demands
during competition, prevention strategies require the training of a dynamic and
functional system in such a way that it is able to not only respond, but more
importantly gain and adapt from the new challenges or the never before
experienced variability it will potentially face. A system such as this would
benefit from intermittency and piecewise determinism, as these behaviors would
provide the flexibility necessary for the system to efficiently flow through and
between a variety of behavioral states.
Maybe the most easily understood example of the “gain from
volatility” concept in the context of behavior is that of post-traumatic growth,
the converse of post-traumatic disorder. In the former, patients actually grow
stronger (viz. healthier) after stressful past events (Barskova & Oesterreich,
2009; Calhoun & Tedeschi, 2006; Collicutt McGrath, & Linley, 2006; Pat-
504 NDPLS, 19(4), Kiefer & Myer
Horenczyk & Brom, 2007). Robustness does not allow for such a thing, and
while resilience incorporates adaptability, there is no mention in the literature of
a resilient system gaining from the external challenges with which it is faced.
Thus, there is much work that needs to be done to uncover the subtle, yet
important differences in behavior between robust, resilient, and antifragile
systems and how they relate to biology and behavior. It is possible that these
three concepts are independent from one another; however, it is more likely that
they are overlapping concepts that build on one another much in the way that the
NMT training protocol builds from robust movement patterns up through
resilience and toward antifragility.
It is likely that biological systems are not only robust and resilient, but
also inherently antifragile: They gain or evolve from external volatility at a
variety of time scales while the system can remain relatively unchanged at the
global level. For example, consider the remodeling process of bones that cause
positive adaptation of a bone’s histological structure to increases in load (Cowin
& Hegedus, 1976). This process can be induced naturally, as through normal
weight-bearing activities following recovery from a bone fracture, or artificially
from weight lifting as load is systematically increased over time. Physiological
(e.g., neuromuscular) adaptations occur in much the same way. Specifically,
strength gains during exercise are greatest when the system is stressed due to a
change in load (e.g., increased weight of dumbbells), and an entire training
principle called specific adaptations to imposed demands (SAID)—the body will
adapt to demands placed upon it—has been developed around this phenomenon
(Mathews & Fox, 1976). However, traditional applications of the SAID
principle rely on linear increases in demand, or load, and thus simply make the
system more robust. They do not necessarily induce antifragility for injury
prevention (Fig. 7, top) and likely would not lead to the type of changes in
%DET and %LAM that we found in the current study. Instead, interventions that
build on the SAID principle and push the system even farther toward its
biological limits are required. Specifically, novel training progressions are
needed that not only add noise to the system, but do so in a way that the system
is pseudo-randomly brought to the brink of critical failure while it continues to
maintain global integrity. This allows the local level dynamics (i.e., muscle
physiologies) to be pushed to the limit as they “learn” to live in a state of
intermittency and piecewise determinism, while global coordination between
limbs and body segments is more or less preserved. These dynamics allow the
system to remain poised for and even thrive from volatility at the local level,
without putting the global system at increased risk of injury (Fig. 7, bottom).
The result of such training would be a system that, when faced with external
fluctuations that would normally lead to increased load and potentially injury,
would respond more rapidly and correct itself before the load hits a critical
threshold, and even allow the system to react faster or more efficiently the next
time it faces any type of external noise.
Figure 7 illustrates schema of linear increases in training load to induce
specific adaptations to imposed demands (top) and a schema of a training
 NDPLS, 19((4), Antifragilee Athlete 5505
progresssion of NMT T designed aroound the conccept of antifraagility (bottom m).
Phase I fortifies techn nique and imp
proves strengthh to a baseline functional levvel.
Phase III introduces progressive noise to expand thhe window of resilience so thhat
a system can with hstand larger amounts of variability w while preserviing
form/teechnique. Phasse III is a stark
k departure froom previous trraining protocools.
In this phase, extrem me random fluuctuations in vvariability aree introduced thhat
continuuously push th he system to the brink of collapse (thee robust windoow
continuues to expan nd) while sim multaneously fluctuating nooise at randoom
magnittudes to force the system to co ontinuously addapt to unprediictable variabillity
(antifraagility increasees).

Fig. 7. A schema of linear incrreases in traiining load to induce specific

adaptations to impose
ed demands (toop) and a sche ema of a trainin
ng progression
n of
NMT deesigned aroundd the concept of
o antifragility ( bottom).
There are sttill several outsstanding questtions that needd to be addresssed
before a conclusive link between intermittencyy, piecewise ddeterminism, aand
antifrag
gility can be made.
m An open question still remains as to the link betweeen
resilien
nce (as defineed in some ciircles) and anntifragility. W When consideriing
intermiittency in the context
c of resillience and antiffragility, Woodds and Wreathhall
(2008) seem to comee the closest to o the idea of ggaining from vvolatility in thheir
definitiion of resiliencce; however, th hey do not expplicitly state ssuch an outcomme.
Interesttingly, their nootion of energgy and resourcees as it relatess to efficiencyy is
perhaps a critical question
q for future
f researchh, and may bbe an importaant
consideeration in link king their con ncept of resilieence with Talleb’s concept of
antifrag
gility (Taleb, 2012).
2 Anotheer issue is thatt, based on thhe distribution of
responsses that can beb measured when w a systemm is exposed too randomness, a
systemm can be characcterized as frag gile (concavity)) or antifragilee (convexity): tthe
506 NDPLS, 19(4), Kiefer & Myer
convex gains more than it loses when faced with variability (Taleb, 2012).
However, this is at the global level of a system. At a local level the nature of the
distribution of the system, and the dynamics of its behavior more specifically,
remains unclear. The observed decreases in %LAM and %DET as measured in
the gluteal tonus dynamics lend support for the idea of intermittency and
piecewise determinism as foundational for antifragility, but this is based only on
literature that has examined the characteristics of systems as performance relates
to optimal variability (Diniz et al., 2011; Riley & Turvey, 2002; Van Orden et
al., 2011; West, 2006). For more conclusive support of this argument, the
adaptability of the system needs to be examined in relation to the level or
amount of intermittency (and randomness) present in a given system. Further,
dynamical systems models of specific behaviors need to be developed and
tested, and parameter spaces need to be probed, in order to understand the role
of external fluctuations on behavior. While this has been done quite well in the
area of rhythmic motor coordination (e.g., Kelso & Ding, 1993; Kelso, Scholz,
& Schoner, 1986; Kelso, 1984; Kugler, Kelso, & Turvey, 1980; Kugler et al.,
1982; Scholz, Kelso, & Schoner, 1987), aside from locomotor path trajectories
(Fajen & Warren, 2003, 2007; Warren, 2006), it is a less developed approach for
more aperiodic, trajectory-driven behaviors. These aperiodic behaviors tend to
make up more of the overall strings of action that humans perform in sport, and
throughout activities of daily living. Thus, the measurement and modeling of
these behaviors require a greater amount of attention to support further
initiatives for injury prevention and improved health outcomes
CONCLUSION
NMT, at its core, is an antifragile training protocol that capitalizes on
the inherently antifragile nature of biological systems, and thus physiological
adaptations lend themselves to such an approach and should be the focus for
injury prevention. The identification and assessment of antifragile physiologies
and specifically, the muscle tonus dynamics therefore become paramount for
prevention. Nonlinear analyses such as RQA are well suited to this endeavor and
the current data provide evidence that piecewise determinism and intermittency
are two modifiable characteristics that provide evidence of a physiological sys-
tem that lives at an optimal point of metaflexibility. More research is needed to
understand whether these dynamic signatures are truly representative of antifra-
gility, and what specific role piecewise determinism and intermittency play in
the adaptability of the system to external noise. Regardless, the current results
provide a novel approach of study of injury prevention in a new and exciting
framework, and open the door to extremely innovative questions related to
motor learning and behavioral dynamics across the spectrum of human
movement.
REFERENCES
Barskova, T., & Oesterreich, R. (2009). Post-traumatic growth in people living with a
serious medical condition and its relations to physical and mental health: A
systematic review. Disability and Rehabilitation, 31, 1709–1733.
 NDPLS, 19(4), Antifragile Athlete 507
Calhoun, L., & Tedeschi, R., (Eds.) (2006). Handbook of posttraumatic growth:
Research & practice. Mahwah, NJ. Lawrence Erlbaum Associates Publishers.
Collicutt McGrath, J., & Linley, P. A. (2006). Post-traumatic growth in acquired brain
injury: a preliminary small scale study. Brain Injury, 20, 767–773.
Cowin, S., & Hegedus, D. (1976). Bone remodeling I: Theory of adaptive elasticity.
Journal of Elasticity, 6, 313–326.
Diniz, A., Wijnants, M. L., Torre, K., Barreiros, J., Crato, N., Bosman, A. M. T., …
Delignières, D. (2011). Contemporary theories of 1/f noise in motor control.
Human Movement Science, 30, 889–905.
Fajen, B. R., & Warren, W. H. (2003). Behavioral dynamics of steering, obstable
avoidance, and route selection. Journal of Experimental Psychology: Human
Perception and Performance, 29, 343–362.
Fajen, B. R., & Warren, W. H. (2007). Behavioral dynamics of intercepting a moving
target. Experimental Brain Research. Experimentelle Hirnforschung.
Expérimentation Cérébrale, 180, 303–319.
Ford, K., Myer, G., & Hewett, T. (2003). Valgus knee motion during landing in high
school female and male basketball players. Medicine and Science in Sports and
Exercise, 35, 1745-1750.
Glass, L., & Mackey, M. C. (1979). Pathological conditions resulting from instabilities in
physiological control systems. Annals of the New York Academy of Sciences,
316, 214–35.
Guastello, S. J., & Liebovitch, L. S. (2009). Introduction to nonlinear dynamics and
complexity. In S. J. Guastello, M. Koopmans, & D. Pincus (Eds.), Chaos and
complexity in psychology: The theory of nonlinear dynamical systems (pp. 1-
40). New York: Cambridge University Press.
Hausdorff, J., & Peng, C. (1996). Multiscaled randomness: A possible source of 1/f noise
in biology. Physical Review. E, Statistical Physics, Plasmas, Fluids, and
Related Interdisciplinary Topics, 54, 2154–2157.
Hewett, T. (2009). Prevention of non-contact ACL injuries in Women: Use of the Core of
Evidence to Clip the Wings of a “Black Swan.” Current Sports Medicine
Reports, 8, 219–221.
Hewett, T., Lindenfeld, T. N., Riccobene, J. V, & Noyes, F. R. (1999). The effect of
neuromuscular training on the incidence of knee injury in female athletes. A
prospective study. The American Journal of Sports Medicine, 27, 699–706.
Hewett, T., Myer, G., Ford, K., Heidt, R., Colosimo, A., McLean, S., … Succop, P.
(2005). Biomechanical measures of neuromuscular control and valgus loading
of the knee predict anterior cruciate ligament injury risk in female athletes: A
prospective study. The American Journal of Sports Medicine, 33, 492–501.
Hewett, T., Stroupe, A., Nance, T., & Noyes, F. (1996). Plyometric training in female
athletes decreased impact forces and increased hamstring torques. The
American Journal of Sports Medicine, 24, 765-773.
Kelso, J. (1984). Phase transitions and critical behavior in human bimanual coordination.
American Journal of Physiology, 246, R1000-1004.
Kelso, J., & Ding, M. (1993). Fluctuations, intermittency, and controllable chaos in
biological coordination. In K. M. Newell & D. M. Corcos (Eds.), Variability
and motor control (pp. 291–316). Champaign, IL: Human Kinetics Publishers.
Kelso, J., Scholz, J., & Schoner, G. (1986). Nonequilibrium phase transitions in
coordinated biological motion: critical fluctuations. Physics Letters A, 118,
279–284.
Krosshaug, T., Nakamae, A., Boden, B. P., Engebretsen, L., Smith, G., Slauterbeck, J. R.,
… Bahr, R. (2007). Mechanisms of anterior cruciate ligament injury in
basketball: video analysis of 39 cases. The American Journal of Sports
508 NDPLS, 19(4), Kiefer & Myer
Medicine, 35, 359–367.
Kugler, P. N., Kelso, J. A. S, & Turvey, M. T. (1980). On the concept of coordinative
structures as dissipative structures: I. Theoretical lines of convergence.
Advances in Psychology, 1, 3-47.
Kugler, P. N., Kelso, J. A. S., & Turvey, M. T. (1982). On the control and co-ordination
of naturally developing systems. In J. Kelso & J. Clark (Eds.), The development
of movement control and co-ordination (pp. 5–78). Chichester: John Wiley &
Sons, Ltd.
Liu, Y., Kankaanpää, M., Zbilut, J. P., & Webber, C. L. (2004). EMG recurrence
quantifications in dynamic exercise. Biological Cybernetics, 90, 337–348.
Mackey, M., & Glass, L. (1977). Oscillation and Chaos in Physiological Control
Systems. Science, 197, 287–289.
Mandelbaum, B. (2005). Effectiveness of a neuromuscular and proprioceptive training
program in preventing the incidence of ACL injuries in female athletes: two-
year follow up. American Journal of Sports Medicine, 33, 1003–1010.
Marwan, N., Carmenromano, M., Thiel, M., & Kurths, J. (2007). Recurrence plots for the
analysis of complex systems. Physics Reports, 438, 237–329.
Mathews, D., & Fox, E. (1976). The physiological basis of physical education and
athletics. Philadelphia: Saunders.
Myer, G., Brent, J. L., Ford, K., & Hewett, T. (2008). A pilot study to determine the
effect of trunk and hip focused neuromuscular training on hip and knee
isokinetic strength. British Journal of Sports Medicine, 42, 614–619.
Myer, G., Chu, D. A, Brent, J. L., & Hewett, T. (2008). Trunk and hip control
neuromuscular training for the prevention of knee joint injury. Clinics in Sports
Medicine, 27, 425–448.
Myer, G. D., Ford, K. R., Brent, J. L., & Hewett, T. E. (2006). The effects of plyometric
vs. dynamic stabilization and balance training on power, balance, and landing
force in female athletes. The Journal of Strength & Conditioning Research, 20,
345-353.
Myer, G., Ford, K., Brent, J., & Hewett, T. (2012a). An integrated approach to change the
outcome part I: neuromuscular screening methods to identify high ACL injury
risk athletes. Journal of Strength and Conditioning Research / National
Strength & Conditioning Association, 26, 2265–2271.
Myer, G., Ford, K., Brent, J. L., & Hewett, T. (2007). Differential neuromuscular training
effects on ACL injury risk factors in“high-risk” versus “low-risk” athletes.
BMC Musculoskeletal Disorders, 8:39.
Myer, G., Ford, K., Brent, J. L., & Hewett, T. (2012b). An integrated approach to change
the outcome part II: targeted neuromuscular training techniques to reduce
identified ACL injury risk factors. Journal of Strength and Conditioning
Research / National Strength & Conditioning Association, 26, 2272–2292.
Myer, G., Ford, K., Khoury, J., Succop, P., & Hewett, T. (2011). Biomechanics
laboratory-based prediction algorithm to identify female athletes with high
knee loads that increase risk of ACL injury. British Journal of Sports Medicine,
45, 245–252.
Myer, G., Ford, K., McLean, S., & Hewett, T. (2006). The effects of plyometric versus
dynamic stabilization and balance training on lower extremity biomechanics.
The American Journal of Sports Medicine, 34, 445–455.
Myer, G., Ford, K., Palumbo, J., & Hewett, T. (2005). Neuromuscular training improves
performance and lower-extremity biomechanics in female athletes. The Journal
of Strength & Conditioning, 19, 51–60.
Myer, G., Sugimoto, D., Thomas, S., & Hewett, T. E. (2013). The influence of age on the
effectiveness of neuromuscular training to reduce anterior cruciate ligament
 NDPLS, 19(4), Antifragile Athlete 509
injury in female athletes: a meta-analysis. The American Journal of Sports
Medicine, 41, 203–215.
Myklebust, G., Engebretsen, L., Braekken, I., Skjolberg, A., Olsen, O.-E., & Bahr, R.
(2003). Prevention of anterior cruciate ligament injuries in female team
handball players: A prospective intervention study over three seasons. Clinical
Journal of Sport Medicine, 13, 71-78.
Nemeth, C. P. (2008). Resilience engineering: The birth of a notion. In E. Hollnagel, C.
P. Nemeth, & S. Dekker (Eds.), Resilience engineering perspectives, vol. 1 (pp.
5-6). Burlington, VT: Ashgate Publishing Ltd.
Pat-Horenczyk, R., & Brom, D. (2007). The multiple faces of post-traumatic growth.
Applied Psychology, 56, 379–385.
Paterno, M. V, Schmitt, L. C., Ford, K., Rauh, M. J., Myer, G., Huang, B., & Hewett, T.
(2010). Biomechanical measures during landing and postural stability predict
second anterior cruciate ligament injury after anterior cruciate ligament
reconstruction and return to sport. The American Journal of Sports Medicine,
38, 1968–1978.
Petersen, W., Braun, C., Bock, W., Schmidt, K., Weimann, A., Drescher, W., … Zantop,
T. (2005). A controlled prospective case control study of a prevention training
program in female team handball players: the German experience. Archives of
Orthopaedic and Trauma Surgery, 125, 614–621.
Pincus, D., & Metten, A. (2010). Nonlinear dynamics in biopsychosocial resilience.
Nonlinear Dynamics, Psychology, and Life Sciences, 14, 353–380.
Riley, M. A, & Turvey, M. T. (2002). Variability of determinism in motor behavior.
Journal of Motor Behavior, 34, 99–125.
Scholz, J. P., Kelso, J. A. S., & Schoner, G. (1987). Nonequilibrium phase transitions in
coordinated biological motion: critical slowing down and switching time.
Physics Letters A, 123, 8–12.
Sugimoto, D., Myer, G., McKeon, J. M., & Hewett, T. (2012). Evaluation of the
effectiveness of neuromuscular training to reduce anterior cruciate ligament
injury in female athletes: A critical review of relative risk reduction and
numbers-needed-to-treat analyses. British Journal of Sports Medicine, 46, 979–
88.
Taleb, N. (2007). The black swan: The impact of the highly improbable (2nd ed.). New
York: Random House.
Taleb, N. (2012). Antifragile: Things that gain from disorder (Incerto). New York:
Random House.
Turvey, M. T. (1990). Coordination. American Psychologist, 45, 938–953.
Van Orden, G. C., Kloos, H., & Wallot, S. (2011). Living in the pink: Intentionality,
wellbeing, and complexity. In Hooker, C. (Ed.), Handbook of the philosophy of
science, vol. 10: Philosophy of complex systems (pp. 639-683). Oxford, UK:
Elsevier.
Warren, W. H. (2006). The dynamics of perception and action. Psychological Review,
113, 358–389.
Webber, C., & Zbilut, J. (2005). Recurrence quantification analysis of nonlinear
dynamical systems. In M. Riley and G. Van Orden (Eds.), Tutorials in
contemporary nonlinear methods for the behavioral sciences web book (pp.
26–94). Washington, DC: National Science Foundation.
West, B. J. (2006). Where medicine went wrong: Rediscovering the path to complexity.
Hackensak, NJ: World Scientific.
Woods, D. D. & Wreathall, J. (2008). Stress-strain plots as a basis for assessing system
resilience. In E. Hollnagel, C. P. Nemeth, and S. Dekker (Eds.), Resilience
Engineering Perspectives, vol 1 (pp. 5-6). Burlington, VT: Ashgate Publishing
510 NDPLS, 19(4), Kiefer & Myer
Ltd.
Zbilut, J., Dixon, D., & Zak, M. (2002). Detecting singularities of piecewise deterministic
(terminal) dynamics in experimental data. Physics Letters A, 304, 95–101.
Zbilut, J. P., & Webber, C. L. (2008). Laminar recurrences, maxline, unstable
singularities and biological dynamics. The European Physical Journal Special
Topics, 164, 55–65.