Child Development, November/December 2006, Volume 77, Number 6, Pages 1521 1538

Moving Toward a Grand Theory of Development: In Memory of Esther Thelen

Daniela Corbetta Beverly Ulrich

Patricia Buchanan Gregor Schoner

This paper is in memory of Esther Thelen, who passed away while President of the Society for Research in Child
Development. A survey of Esther Thelens career reveals a trajectory from early work on simple movements like
stepping, to the study of goal-directed reaching, to work on the embodiment of cognition, and, ultimately, to a
grand theory of developmentFdynamic systems theory. Four central concepts emerged during her career: (1) a
new emphasis on time; (2) the proposal that behavior is softly assembled from the interaction of multiple
subsystems; (3) the embodiment of perception, action, and cognition; and (4) a new respect for individuality.
Esther Thelen communicated these ideas to scientists and practitioners alike, so the ultimate benefactors of her
work were children.

Child Development traditionally publishes the Presi- Child Development. Esther was not only a mentor to
dential Address presented at the biennial meetings us all, but also a dear friend. Thus, we appreciate the
of the Society for Research in Child Development opportunity to share our collective vision of Esther
(SRCD). As President, Esther Thelen was slated to and of her work with you.
deliver the address in 2005, but her death on De- Although being a supportive and generous friend
cember 29, 2004, sadly precluded this event. In its was a core value Esther lived, our goal in this paper
place, a Memorial Symposium was held. One of Es- is to focus on her science. We use illustrations from
thers trademark characteristics was to be inclusive, her work to outline the themes emerging from her
supportive, and collaborativeFto bring people to- movement toward a grand theory of development.
gether. This is reflected in the large group of former We end by proposing several future directions that
postdocs, doctoral students, and research associates are inspired by her work.
who collaborated to prepare that symposium, and
then to prepare this archival paper for publication in
The First Steps: Setting the Stage
This research was supported by NIH HD 22830 awarded to We begin by examining Esthers early work in infant
Esther Thelen. Many people contributed to making the 2005 SRCD and child development. This early phase is perhaps
Presidential Address such a memorable event. We are deeply
grateful to the coauthors of the four talks: (Part 1) Donna Fisher- most vividly described as the kicking, stepping, and
Thompson, Jody Jensen, Beatrix Vereijken; (Part 2) Dexter Gorm- walking period. Many of the themes that character-
ley, J. Cole Galloway; (Part 3) Fred Diedrich, Linda B. Smith; (Part ize her work were foreshadowed in her earliest pa-
4) Jana Iverson. We would also like to thank Dexter Gormley, Scott pers. These themes were also foreshadowed in the
Robinson, and Vanessa Simmering for their invaluable technical context of her early career: Esther followed her own
assistance, and Brandi Dobbertin, Jeff Johnson, John Lipinski,
Larissa Samuelson, and Wendy Troob for their many contributions
unique path into the world of developmental theory.
both large and small. Finally, we would like to acknowledge Roger Her earliest graduate research focused on the
Bakeman, who hired Atlantas Harmony Chorus for the event. We grooming behaviors of wasps. Esther was drawn to
will never forget the moment when Esther finished speaking on this, in part, by the ethologists methods of observing
video and the voices of children echoed through the room singing
animal behavior in its natural environment and dis-
a South African freedom song. This gave everyone in the room a
concrete example of the power of self-organization. covering the underlying repeated patterns that
Correspondence concerning this article should be addressed to marked their function and social impact. Esther
John P. Spencer, Department of Psychology, 11 Seashore Hall E,
University of Iowa, Iowa City, IA 52242. Electronic mail may be r 2006 by the Society for Research in Child Development, Inc.
sent to john-spencer@uiowa.edu. All rights reserved. 0009-3920/2006/7706-0001
1522 Spencer et al.

discovered that actions that were repeated impacted mentFit was quite remarkable that someone trained
subsequent behaviorFeven if the link between past in biology drew such a strong and interested re-
and present behaviors was not transparent (Thelen & sponse with her first foray into the developmental
Farish, 1977). This pushed her to focus on process, on literature. Perhaps developmental psychologists
how and why these behaviors emerged and changed. could already glean from her cogent arguments the
seeds of a grand theory.

Rhythmical Stereotypies
Shifting Patterns of Bilateral Coordination: Newborn
Esther crossed over to the human side of animal
Stepping
studies with her dissertation. Here, she catalogued the
spontaneous behaviors of babies in their natural Of the many stereotypies infants generate, Esther
settingsFtheir homesFbiweekly throughout the first was drawn particularly to kicking because it was
year of life. Figure 1 illustrates the outcome of her repeated so frequently. Moreover, Esthers work re-
painstaking efforts. She grouped the actions babies vealed that the coordination patterns that underlie
repeated into stereotypy categories, such as waving the kicking were strikingly similar to the patterns that
arms, rocking in place, and kicking their legs. And she underlie newborn stepping. The puzzle was that
uncovered the developmental trajectory of each ste- newborn stepping disappeared within the first 3
reotypy, the timing of peak behaviors, and overlap months, whereas kicking continued and increased in
among these different behaviors as well as their rela- frequency. What were the secrets to this mystery?
tion to the emergence of functional motor milestones. Several researchers had proposed that maturing
For example, kicking movements had their greatest cortical centers inhibit the primitive stepping reflex
frequency just before the onset of locomotion and or that stepping was phylogenetically programmed
rocking on hands and knees appeared just before to disappear (Andre-Thomas & Autgaerden, 1966;
crawling (see Figure 1). Her first publications based on Oppenheim, 1981; Peiper, 1963; Touwen, 1976).
this work appeared in animal behavior journals (The- To probe the mystery of the disappearing steps,
len, 1979, 1981a). But in 1981, she published a seminal Esther conducted another longitudinal study, confi-
paper in Developmental Psychology entitled Rhythmical dent that the answer would be revealed by focusing
behavior in infancy: An ethological perspective. on the detailed development of individual infants and
She discovered that these simple, repetitive be- their individual differences. Thelen, Fisher, and Rid-
haviors not only provide a window for researchers ley-Johnson (1984) found a clue in the fact that
interested in studying motor control, but also pro- chubby babies and those who gained weight fastest
vide infants with opportunities to become active were the first to stop stepping. It appeared that for
participants in their own learning. This theme clearly young infants to flex and extend their legs when
resonated with researchers studying child develop- upright (stepping; see Figure 2a) demanded more

Rhythmical Behavior
Percent of Total Stereotypy Group

Legs
T-Hands & Knees
T-Sitting
T-Kneeling
T-Standing
Arms
Fingers
Hands
4

12

20

28

36

44

52

Age, in Weeks

Figure 1. Frequencies of groups of rhythmical stereotypies during the first year (T 5 torso) (adapted from Thelen, 1981a).
 Moving Toward a Grand Theory of Development 1523

Figure 2. Panel a shows newborn stepping. Panel b shows kicking. Panel c shows the reemergence of stepping when the infant is placed
waist-deep in water.

strength than to flex ones leg when supine (kicking; views of development to consider the many, often
see Figure 2b). To test this idea that strength was the nonobvious factors that influence developmental
key, Esther and her colleagues conducted two ingen- change. Critically, however, Esthers insights offered
ious studies. In one, they placed small leg weights on more than generalist statements that development
2-month-old babies, similar in amount to the weight was self-organizing and everything matters.
they would gain in the ensuing month. They ob- This was brought to the foreground as she began
served a significant drop in newborn stepping. In the to probe ties between infants behaviors and those of
other, they submerged older infants, whose stepping other complex thermodynamic systems. For in-
had begun to wane, in water up to chest levels (see stance, she noted that both showed evidence of
Figure 2c). Step frequency immediately increased; nonlinearity, both became increasingly less stable
more buoyant legs require less muscle force to lift. during periods of change, and both displayed a
These data demonstrated that traditional explan- tendency to shift toward increased organization
ations of neural maturation and innate capacities (Thelen, 1985; Thelen, Kelso, & Fogel, 1987).
were insufficient to explain the emergence of new These ideas emerged out of discussions with
patterns and the flexibility of behavior so evident Scott Kelso, who, along with Peter Kugler and Mi-
in this case. Esther ultimately proposed that step- chael Turvey, was beginning to use the principles of
pingFlike any other behavioral patternFis not complex systems theories developed in physics,
something one has. Rather, behavior emerges in the mathematics, and chemistry to explain adult human
moment from the self-organization of multiple compon- motor patterns and coordination (Kelso, Holt,
ents. This is elegantly illustrated in the stepping Kugler, & Turvey, 1980; Kelso & Tuller, 1984). For
studies which show how the posture of the infant, most of us in the field of development, this ap-
the strength of an infants muscles, and the pull from proach became known as the dynamic systems
the environment all cohere in a moment in time to approach.
create or hinder leg movements. And, further, this Using a clever (some might say bizarre) context,
example illustrates how changes in the components she began to support babies upright on the belt of a
of this system over the longer time scale of de- motorized treadmill (see Figure 3a) to show that
velopment interact with this real-time, self-organiz- behavior self-organizes and that multicausality is at
ing view. Quite literally, Esther and her colleagues the core of developmental change. She designed a
were able to create development in the moment by hin- new longitudinal experiment, studying infants from
dering the movements of young infants through the 1 month of age, monthly, for 7 10 months. Embed-
application of leg weights and recreating stepping in ded in the design were elements that allowed her to
older infants by cleverly changing the environment. incorporate concepts central to the study of complex
systems: (a) a collective variable captures the inte-
grated behavior of the system under study (in the
A Dynamic Theory Emerges as Infants Take a Walk on a
case of stepping, this was reflected in the phase re-
Treadmill
lation between the legs during each step cycle; see
The case of the disappearing steps led Esther and Figure 3b for data on one phase relationFalternat-
her colleagues to move away from single cause ing steps); (b) control parametersFsuch as changes in
1524 Spencer et al.

a b
25

20

Alternating Steps
15 jf
ch
10 bh

0
1 2 3 4 5 6 7 8 9
Age (mo)

Figure 3. Panel a shows an infant walking on a treadmill. Markers on the legs and torso allow the computer to track the motion of the joints
with a high degree of temporal and spatial precision. Panel B shows developmental changes in three infants alternating stepping patterns
(adapted from Thelen & Ulrich, 1991).

the strength of leg musclesFdrive the system Guiding Us Through the Next Steps
through periods of change; (c) behavioral change is
nonlinearFearly in the first year, infants produced One of the underlying goals of basic science is to
few steps on the treadmill, but there was a distinct understand how systems work sufficiently well to
increase in stepping, typically by 3 months of age, intervene, when necessary, and provide novel solu-
that occurred at different times for different infants tions to help people overcome behavioral problems.
(see Figure 3b); (d) behavior is self-organizedFthe Esthers early experiments utilized tasks that em-
motion of the treadmill helped assemble all the phasized kicking, stepping, and walking, behaviors
components involved in stepping such that infants of particular interest to therapists working with early
showed stepping patterns when placed on the belt motor disabilities. Researchers working on motor
even when they failed to show stepping in other problems continue to build on her work to study
contexts. The results of this study were published as neuromotor responses of infants born prematurely,
a monograph for the SRCD (Thelen & Ulrich, 1991). as well as infants born with intraventricular hemor-
This monograph served as a tutorial that explained rhages, Down syndrome, cerebral palsy, and spina
dynamic systems constructs, and it provided a tem- bifida (Angulo-Barroso, Ulrich, & Tiernan, 2004;
plate for how to apply these constructs to the study Heriza, 1988; Ulrich & Moerchen, 2005; Ulrich,
of behavioral development. Ulrich, Angulo-Barroso, & Yun, 2001). These trans-
A key finding from this monograph is that the lational efforts have proved quite promising, in part,
same context elicited different behavior patterns because the contexts Esther created readily elicit
over time and, further, that multiple factors affected motor activity. By carefully controlling and ma-
the patterns that emerged. Early stepping consisted nipulating the subsystems involved in, for instance,
of multiple unstable patterns, including alternating, stepping behaviors, one can push the system into
but also single, parallel, and double steps. As infants new forms of organization and uncover control pa-
acquired more muscle strength, an improved ability rametersFintrinsic or extrinsicFthat give us lever-
to control segments of the body, more experience in age to elicit behavioral change. Ultimately, these may
the upright posture, and so on, stepping reorgan- be used to help children assemble new, more stable
ized. Critically, although some form of stepping and more functional behavioral patterns.
pattern could be elicited over much of the first year Esther published nearly 30 empirical papers in
of postnatal life, walking only emerged after months which kicking and stepping were the primary focus.
of exploration. Esther and Bev Ulrich concluded that But these specific behaviors were merely tools for her
stepping and, ultimately, walking are not innate or focus on larger theoretical issues of learning and
prescribed. Rather, they are self-organized and development. By her painstaking efforts, she in-
emergent, reflecting the assembly of multiple sub- corporated profound empirical discoveries into her
systems within the infants history of activity in evolving theory. Her goal was to establish a grand
context (Thelen & Ulrich, 1991). theory of development with general principles that
 Moving Toward a Grand Theory of Development 1525

apply across varied phenomena and traditionally mensional (3-D) space. They need to understand
disparate domains. And in so doing, she wanted to whether the object is reachable, and they need to
reach out and reach in, so that the ultimate bene- transduce the perceived 3-D space into their body
factors of our work are children. Lofty goals, to be space. They need to be able to plan ahead and an-
sure, but what a wonderful roadmap she left us. ticipate how the trajectory will unfold. They need to
be able to correct their movements online as their
hand approaches the toy. They need to be able to lift
Learning to Reach: Mapping Intentions With and stabilize the arm as they reach while maintain-
Intrinsic Dynamics ing the stability of the head and the trunk. And, they
Kicking and stepping are repetitive movements that need to remember what works in context and dis-
are not always performed with an obvious goal in tinguish this from what does not work. Although
mind. What happens when infants movements vision is certainly involved in some of these chal-
become clearly goal-oriented? Can the principles of lenges, vision alone cannot account for how infants
dynamic systems theory (DST) be applied to the learn to reach given these many interacting factors.
development of goal-oriented behaviors? Do pat- The challenge, then, is to understand how infants
terns of, for instance, reaching movements self-or- manage to assemble all these factors to perform a
ganize as with kicking and stepping? What do successful reach.
infants need to learn to obtain a desired toy? During The second point that Esther challenged was the
the 1990s, Esther turned to these questions as she universality of the two-phase account of infant
sought to generalize her ideas about the nature of reaching. She thought the emphasis needed to be on
development. the individualFon the unique motor problems each
infant must overcome due to each infants unique
movement characteristics. In particular, infants need
to discover their own, individual, intrinsic move-
From a Classical to a Systems View of Infant Reaching
ment characteristics to develop proper control of the
Before Esthers work, the development of infant arm and improve movement coordination over time.
reaching was thought to occur in two phases that all
infants traversed in a similar fashion (see Bushnell,
Individual Development Matters: A Tale of Two Infants
1985, for review). In a first developmental phase
Learning to Reach
beginning at about 3 or 4 months of age, infants
reaches were characterized by very discontinuous, This idea is beautifully illustrated in the story of
zigzagging trajectories. In a second phase appearing Hannah and Gabriel learning to reach. Gabriel and
around 8 months of age, infants began to reach for Hannah were two of the four infants that Esther and
toys following a more direct path. The classic ac- her colleagues followed every week over the first
count of these two developmental phases focused year of life, from 3 to 52 weeks (e.g., see Thelen et al.,
almost entirely on visual control of reaching. During 1993; Thelen, Corbetta, & Spencer, 1996). Gabriel
the first phaseFcalled the visually guided reaching began reaching when he was 15 weeks old. Hannah
phaseFit was assumed that infants discontinuous performed her first reach attempts when she was 20
trajectories reflected their continuous effort to weeks old. Each child displayed very different be-
monitor and visually control the hand trajectory. havioral characteristics on the weeks before and at
In the second phaseFcalled the visually elicited reach onset. Gabriel was a very active, energetic in-
reaching phaseFinfants could look at the target, fant who was flapping both arms up and down along
anticipate an appropriate hand trajectory, and move the sides of his body. He was often flapping re-
their hand to the target using a fairly straight path gardless of whether the toy was visible or not. At
without visual monitoring. reach onset, these flapping patterns became an in-
Esthers work on infant reaching challenged this tegral part of his goal-oriented movements. Typical-
view of development in two key ways. First, Esther ly, when the toy was approaching his reaching space
objected to the heavy emphasis on visual control in from the side to the midline, Gabriel flapped and
earlier accounts. By her view, the development of reached for the toy by throwing his arm up and
reaching did not reflect changes in a single factor; forward. This resulted in swiping at the toy and,
rather, she saw infant reaching as emergent from the occasionally, a toy contact!
assembly of many components. To reach out and Hannah had a completely different style. She was
grab an object, infants need to be motivated. They a much more quiet, lower energy child. At reach
need to be able to localize the object in three-di- onset, her strategy was to stare at the toy intensely
1526 Spencer et al.

before moving her arm forward and making contact shed lines in Figure 4a) was embedded within the
with the toy. Because she was quiet before reaching ongoing stream of activity (solid lines), and the hand
and moved her arm forward slowly, her reaches path that resulted was very discontinuous.
seemed much more controlled and mature than When we contrast these data with Hannahs first
Gabriels, but that was only temporary. Indeed, a few reach (see Figure 4c and d), a very different picture
weeks after her first reaches, Hannah became more emerges. Hannahs movement trajectory (see Figure
active and she began to show the torturous reaching 4c) was much simpler and more direct than Gabriels
trajectories characteristic of Gabriels early reaches. trajectory. Why is this the case? Hannahs movement
By observing Hannah and Gabriel week by week velocity (see Figure 4d) was almost three times lower
(along with two other infants), Esther and her col- than Gabriels velocity on his first reaching attempt.
leagues discovered that the early discontinuities in Thus, while Gabriel was generating high motion-
reaching trajectories did not result from visual dependent forces that pushed his arm off the in-
guidance of the hand; rather, they emerged as the tended trajectory leading to many corrections, Han-
product of high movement speed and associated nah generated much smaller forces and was able to
large passive and active forces combined with a lack produce a fairly straight trajectory on her first reach.
of control. This is evident in the kinematic profiles of But Hannahs seemingly good reaching skill was
Gabriels and Hannahs first reaches. Figure 4a only temporary. During the weeks after these infants
shows the displacement of Gabriels hand during an began reaching, Gabriel and Hannah modulated
8-s segment of motion. This segment shows a con- their reaches in unique ways to overcome the chal-
tinuous movement performed before (solid line), lenges set up by their individual movement ten-
during (dashed line), and after (solid line) the reach. dencies. Specifically, within the few weeks after
Figure 4b shows the corresponding velocity profile reach onset, infants had to learn to either ramp up or
for this segment of movement, with the onset and clamp down on these movement tendencies to more
offset of the reaching segment marked by the two effectively bring their hands to the toy. Gabriel, who
vertical bars. Gabriel was initially producing fast was generating high and fast velocity peaks, had to
movementsFflapsFwith several high peaks of learn to slow down his movements. Within 3 weeks,
velocity before the toy came into view. Then, when he showed a dramatic decline in the number of vel-
the toy appeared, Gabriel turned a flap into a ocity peaks in his reaches with an associated increase
reaching movement, again generating a series of in path straightness. By contrast, Hannah, who pro-
high velocity peaks. This reaching movement (da- duced much slower movements, had to learn to

Figure 4. Kinematic profiles of Gabriels and Hannahs first reaches during an 8-s segment. (a) Endpoint trajectory of Gabriels right arm
movement before (solid line), during (dashed line), and after (solid line) reaching for a toy at midline. Gabriels reaches were embedded in
a continuous movement stream. (b) Corresponding resultant speed profile with demarcation of the reaching segment for Gabriel. (c)
Endpoint trajectory of Hannahs right arm movement when reaching for a toy at midline. (d) Corresponding resultant speed profile with
demarcation of the reaching segment for Hannah (adapted from Thelen et al., 1993).
 Moving Toward a Grand Theory of Development 1527

inject more energy into her reaches to lift her arm ation generated crucial sensory-motor experience
against gravity and extend her hand up and forward needed to learn to calibrate movements and feel the
to the toy. She became more active in the weeks boundaries of control within the reaching task. This
following reach onset, with an increase in movement resulted in selection during a second phase from 30
velocity and a decrease in path straightness. Thus, to 36 weeks through the end of the first year (see
Hannah, who looked pretty skilled on her first selection phase in Figure 5). In this phase, move-
reaching attempts, became much worse in the few ment parameters settled near particular values and
weeks following reach onset (Thelen et al., 1993). showed much more stability over time as infants
discovered, for instance, an optimal reaching speed
that led to more stable and efficient reaches (Thelen
Developmental Change Occurs Through Exploration and
& Corbetta, 1994; Thelen et al., 1996).
Selection
The third critical point illustrated in Figure 5 is
In addition to showing that infants follow their that Gabriel and Hannah (as well as the other two
own unique trajectories, this study of infant reaching infants in this study) converged on similar move-
revealed that exploration and selection is a key agent ment characteristics at the end of the first year des-
of developmental change. To improve over time, pite their very different starting points at reach onset.
each infant experienced and explored a wide range For instance, they converged to a similar number of
of movements, ultimately leading to the discovery movement units, comparable movement straight-
and selection of an optimal way to assemble the ness, and similar movement speed when contacting
many components involved in a reach. Such ex- the toy (see the small side arrows in Figure 5). This
ploration and selection is evident when we consider reveals that different developmental pathways can
the full set of changes in Gabriels and Hannahs lead to similar outcomes (Thelen et al., 1996).
reaching characteristics from reaching onset to the
end of their first year of life. Figure 5 shows changes
New Lessons About Learning to Reach
in several kinematic variables across the first year for
reaching (Figure 5a f and i n) and nonreaching In conclusion, Esther taught us a number of
movements (Figure 5g, h, o, and p) for Gabriel and wonderful lessons about the development of infant
Hannah, respectively (Thelen et al., 1996). reaching. She taught us that developmental change
We want to highlight three points about these happens at the level of the individualFall infants do
data. First, goal-oriented reaches were always em- not develop motor skills following the same mold.
bedded within a movement context that showed Instead, each infant experiences unique and different
very similar kinematic characteristics. For instance, movement challenges that call for different solutions
changes in the velocity of infants reaches over the and contribute to the formation of a distinct devel-
first year (Figure 5a c and i k) paralleled changes opmental pathway. In this way, she moved us away
in the velocity of infants nonreaching movements from earlier views about the visual guidance of the
(Figures 5g, h, o, and p). Thus, when infants were hand and universal phases of development. Rather,
more active as reflected in periods of higher average Esthers work demonstrated that reaches are carved
speeds in nonreaching movements, reaches were out from the intrinsic dynamics of infants self-gen-
also performed with higher movement speeds (see erated arm movements as they explore a range of
the dashed arrow line marking the active period for movement possibilities and select viable solutions to
Gabriel and Hannah). Likewise, when infants were meet the demands of the task. In this achievement,
less active as reflected in periods of lower average body and mind come together as infants assemble
speeds, their reaches were also slower. the many components that make a reach: the bio-
Second, these graphs illustrate the concepts of mechanics of the body, the details of the specific
exploration and selection. Exploration is reflected in environment including the perceived location of the
an early phase from reach onset to about 30 36 toy, the speed and force needed to extend the arms
weeks, where all the speed and trajectory parameters away from the body, the ongoing movement and
fluctuate up and down and show unstable and postural context, and so on. This integration of body
changing curves (see exploration phase in Figure and mind is a fundamental characteristic of all goal-
5). In this phase, infants explored a wide range of directed actions and creates a bridge to an embodied
movement parameters and movement solutions. view of cognition and behavior (for an additional
They tried fast and slow movements, learning the discussion of these lessons, see Corbetta & Thelen,
effect of these varied speeds on their ability to ac- 1994, 1996, 1999, 2002; Spencer & Thelen, 2000;
quire the toy (see Thelen et al., 1996). This explor- Spencer, Vereijken, Diedrich, & Thelen, 2000).
1528 Spencer et al.

Figure 5. Kinematic variables of Gabriels and Hannahs reaching and nonreaching movements from the onset of reaching to the end of the
first year. Panels a f: means and standard errors of movement speed at reach initiation, average reaching speed, maximum reach speed,
straightness of the path, number of movement units, and speed at hand toy contact for Gabriel. Panels g and h: means and standard
errors of the average speed of all nonreaching movements and the segment of motion before reach initiation for Gabriel. Panels i n: means
and standard errors of movement speed at reach initiation, average reaching speed, maximum reach speed, straightness of the path,
number of movement units, and speed at hand toy contact for Hannah. Panels o and p: means and standard errors of the average speed
of all nonreaching movements and the segment of motion before reach initiation for Hannah. Arrows at the top indicate phases in the
development of reaching skill. Arrows to the right indicate common values of kinematic variables for both infants at the end of the first
year (adapted from Thelen et al., 1996).

From Action to Embodied Cognition: Bridging the select their own unique solutions in the context of
Great Divide their intrinsic dynamics and movement history.
Finally, progress in motor development requires the
By this point, Esther had learned that infant motor integration of body and mind as infants bring to-
behaviorsFboth rhythmic movements with the legs gether their physical characteristics with the envir-
and goal-directed reachesFemerge as the product onmental and movement context to find optimal
and confluence of multiple factors. She had dis- and flexible behavioral solutions. These were key
covered that learning new movement patterns was insights in motor development, but did these in-
not the same for all infants; individual infants must sights extend beyond the realm of stepping, kicking,
explore a wide range of behaviors to discover and and reaching?
 Moving Toward a Grand Theory of Development 1529

To address this question, Esther started to dynamics of reaching first led her to a classic Pi-
think deeply about the connections between her agetian task: the A-not-B task. In this task, infants
work in motor development and work in other do- watch while a toy is repeatedly hidden in one loca-
mains of development. She turned naturally to tion. After a brief delay, infants reach to that location
cognition. Many of the dominant questions in cog- and uncover the toy. After several trials to this A
nitive development stem from Piaget, who asked location, infants watch while the toy is hidden in a
how children move from the sensorimotor origins of second B location a few inches away. Almost
thought to abstract cognition. Contemporary the- without fail, 8- to 10-month-old infants will reach
orists have built on this idea, emphasizing the back to the original A location after a short delay,
transformation from perceptual to conceptual pro- that is, they reach to A and not B (Piaget, 1954; Smith,
cessing (e.g., Mandler, 1988) or the construction of Thelen, Titzer, & McLin, 1999). According to Piaget
more sophisticated forms of cognition built upon (1954), this A-not-B error was indicative of infants
innate core knowledge modules (e.g., Spelke, incomplete object concept. More contemporary
1998). Critically, such approaches have postulated a theories have emphasized that problems with spatial
divide between the cognitive and the sensori- coding, search strategies, or fragile object represen-
motor, either through the pursuit of abstract forms tations underlie this error (see Marcovitch & Zelazo,
of cognition without an understanding of their 1999; Munakata, 1998; Wellman, Cross, & Bartsch,
sensorimotor origins or by creating a distinction 1987).
between the cognitive and the sensorimotor from Instead of thinking about what this task tells us
the beginning. about what infants know, Esther and her colleagues
This divide was evident when Esther contrasted began their analysis of the A-not-B error by focusing
the fields of motor development and infant cognition. on what infants do. And what they do in this task is
In motor development, researchers try to understand look and reach again and again to a poorly specified
the processes that result in a kick or a reach or a target at an A location, only to look and reach again
lookFthese behaviors are interesting in their own to this location after looking at an event at B. Thus,
right and reveal characteristics of how perception, Esther asked whether the processes that underlie
action, and cognition all come together to assemble a repeatedly looking at and reaching to a location
behavior in context. In infant cognition, by contrast, a could explain the complex pattern of behavior re-
reach or a look is just a way for infants to demonstrate vealed by decades of research on this odd error in
knowledge, that is a way to get at what infants know. infancy.
Given this, contemporary theories of infant cognition This question led Esther and her colleagues to
have little to say about the role of the body in mind. develop the dynamic field theory (DFT) of infant
Esther questioned the divide between pure perseverative reaching (Thelen, Schoner, Scheier, &
sensorimotor behavior and cognition. Indeed, in Smith, 2001). This theory captures the processes that
collaboration with Linda Smith, she denied the very underlie infants decisions to act based on the inte-
existence of this divide. She argued that mental ac- gration of the immediate environmental stimuli as
tivity is embodiedFthought is always grounded in well as the short-term and long-term history of
perception and action (e.g., Thelen, 2000; Thelen & reaching in the same and similar situations. The
Smith, 1994). This followed Piagets tradition in in- theory is captured schematically in Figure 6. The
voking the importance of the sensorimotor origins theory starts with the concept of an activation field
of thought. But rather than viewing development as that captures how infants plan and remember ac-
movement toward the abstract and away from per- tions. The activation field depicted in Figure 6 shows
ception action, Esther believed thatFfor infants two peaks of activation distributed across a
and adults alikeFcognition and action are not sep- movement parameter, reaching direction. There is
arate. Instead, cognition is inextricably linked to a large peak of activation over the A location and a
perception and movement. There is no cognition in smaller peak over the B location. This represents a
the absence of perception and action. stable decision to move to A: infants have a strong,
supra-threshold representation of A that can sub-
serve a reaching movement to this location.
What factors contributed to this activation peak at
Embodied Cognition and the Piagetian A-not-B Task
A? Figure 6 illustrates three inputs that have been
Esther began exploring this connection between shown to influence infants reaching decisions in the
cognition and action through several traditional A-not-B situation. The first input is the task input.
developmental tasks. Her detailed analysis of the This captures the pattern of activation generated by
1530 Spencer et al.

activation field

A location

B location

mo mete
par
task specific preshape

v em r
a
input input input

ent
Figure 6. Schematic of the A-not-B task. The infant (far left) sits in front of two targets (A and B locations). The activation field (in light gray)
captures the infants decision to reach as a pattern of activation distributed across the behavioral dimension (the direction of the reach). When
activation reaches threshold at one of the locations, the infant decides to reach to that location. Three inputs contribute to the activation level
(see labels at bottom right). The task input represents the perceptual layout of the space, including the two reaching targets. The specific input
is the transient event that draws attention to one location (i.e., the experimenter hides the toy at A). The preshape input represents the infants
perceptual-motor memory of previous reaches to A. In this example, the task input is equal for both the A and B locations, but there is a peak
at A from the specific input (the experimenter cued the A location) and from the preshape (the infant had previously reached to A). This
results in a suprathreshold peak at A (see light gray activation field), which results in a reach to A.

the perceptual layout of the task space, for instance, perceptual-motor memory which lingers after the
the two covers in Figure 6. It is through this level of reach and can preshape the activation field on
input that task variations like distinctive targets or subsequent trials. Thus, on the second trial to the A
multiple locations have an impact (e.g., Bjork & location, infants are already slightly biased by their
Cummings, 1984; Bremner, 1978; Butterworth, previous decision to reach in the same direction. With
Jarrett, & Hicks, 1982; Diedrich, Highlands, Thelen, more and more reaches to A, the perceptual-motor
& Smith, 2001). The second inputFthe specific memory builds up such that, by the first cue to B,
inputFcaptures the pattern of activation produced there is a relatively strong tendency to reach to A (i.e.,
by the attention-grabbing cue of waving the toy or a strong preshape input). This tendency, combined
the lid (or whatever the experimenter does to draw with the symmetric task input and a long delay on the
the infants attention to one location; see Clearfield, B trial, sets infants up to perseverate. In particular,
Dineva, Smith, Diedrich, & Thelen, 2006; Smith et al., infants perseverate because the specific input at B
1999). The final inputFthe preshape inputFcap- fades during the delay and is overtaken by the lin-
tures the pattern of activation produced by the just- gering memory of past reaches to A.
previous past. In particular, this input reflects in-
fants perceptual-motor memory of past reaches to
Bringing Together Theory and Experiment
the A or B locations. In Figure 6, this input has some
activation centered over the A location, reflecting an Through a combination of simulations and new
infants past reaches to A on a series of A trials. experimental evidence, Esther and her colleagues
How does the activation fieldFin concert with the showed that the decision to reach to A or B in the
three inputs depicted in Figure 6Fshed light on the A-not-B task could be entirely determined by infants
processes that underlie the A-not-B error? On the first short-term and long-term history in the task, their
A trial, infants face two identical reaching targets (the action planning abilities, and the inputs in the ex-
side-by-side covers or lids); thus, the task input is perimental context (e.g., Clearfield, Smith, Diedrich,
symmetricFit does not specify either location. As & Thelen, 2006; Diedrich, Thelen, Smith, & Corbetta,
infants watch, the experimenter cues them to reach to 2000; Diedrich et al., 2001). For instance, by thinking
the A location, by either hiding a toy or waving the about the dynamics of reaching and looking, these
lid. This produces a specific input centered at A. researchers demonstrated that hidden toys were un-
When the box is moved within reach after a short necessary to produce the error (Smith et al., 1999).
delay, the specific input to A is strong enough and Similarly, visual distractions had a profound influ-
the memory for that cue sufficiently long lasting that ence on infants pattern of reachingFa simple tap
they reach in this direction. This reach also creates a on the table near A or B could create stronger
 Moving Toward a Grand Theory of Development 1531

perseveration or create a tendency to reach correctly Embodied Cognition and the Dynamics of Infant
to B (Smith et al., 1999). They also showed that the Habituation
probability of perseverating on the B trials is a
Armed with this new understanding of embodi-
function of the number of reaches to A on the A
ment in infancy, Esther moved on to examine an-
trials: the more reaches to A, the stronger the pull
other task thought to provide a window into infants
to A on the B trials; the more spontaneous reaches to
mindsFinfant habituation. Infant habituation tasks
B on the A trials, the weaker the pull to A on the B
comprise the backbone of the field of infant cogni-
trials (Diedrich et al., 2001; Smith et al., 1999; see also
tion. Infants are shown a display or an event over
Marcovitch & Zelazo, 1999).
and over until looking time decreases. Then, they are
In several additional studies, Esther and her col-
shown a slightly different test display or event.
leagues demonstrated that infants decision to reach
If looking time increases, researchers state with cer-
to A or B was truly embodiedFthere was an ob-
tainty that infants discriminated between the two
ligatory coupling between body and mind. For in-
events and, perhaps, that this reflects infants
stance, perseveration is tightly linked to infants
knowledge or understanding of some concept. As
developing reaching abilities (Clearfield & Thelen,
with the A-not-B paradigm, this leap from what in-
2001; Clearfield, Smith, et al., 2006; Diedrich et al.,
fants do to what infants know troubled Esther. She
2001). Initially, when infants are very unskilled, they
began to think about what infants do in this task:
reach correctly in the A-not-B task. According to the
they attend to the display, they visually process it,
DFT, this occurs because infants perceptual-motor
and they look and look away from the display. Her
traces get smeared out as they reach for the A lo-
focus was on the multiple causes that produce these
cation in different ways from trial to trial (Clearfield,
behaviors, and how the history of the system might
Smith, et al., 2006). Thus, there is less of a bias to A
influence whether or not infants look at the display
on the B trial due to a less focused preshape input
at test.
(see Figure 6). As another example, Esther and her
To understand the complex looking dynamics at
colleagues showed that changes in the feel of the
work in infant habituation, Esther and Gregor
body between the A and B trials can disrupt per-
Schoner applied the concepts of the DFT to a classic
severation (Clearfield, Diedrich, Smith, & Thelen,
habituation task (Schoner & Thelen, 2006). In par-
2006). When infants arms were weighted during the
ticular, they focused on how the trial-to-trial history
A trials and the weights were removed in between
of perceiving, attending, and acting in context in-
the A and B trials, infants reached correctly to B on
fluences the decisions infants make to look or not
the subsequent B trials! Infants also reached correctly
look at stimuli in habituation tasks. Using a series of
on the B trials when weights were added in between
simulations, they showed how multiple factors in-
the A and B trials. By contrast, when infants wore
fluence infants decisions to look at a stimulus: the
weights through an entire session, they perseverated
history of looks across trials, the salience of the dis-
at the usual rate. Thus, disrupting the feel of the
plays, the number and order of habituation and
arms between the A and B trials was enough to
test trials, and the complexity of the displays.
disrupt perseveration; as long as the arms felt the
They concluded thatFas with the A-not-B para-
same throughout, infants perseverated at the normal
digmFhabituation tasks are not a window into in-
rate.
fants minds, divorced from attention, perception,
In her Presidential Address to the International
and action. Rather, infants decisions to look or not
Society on Infant Studies in 1998, Esther outlined the
to look are a result of the complex interactions
importance of these findings: what infants know is
among these diverse and fully embodied processes.
always assembled, in the moment, with contribu-
tions from memory, attention, and action (Thelen,
2000). Cognition is embodied. Infants decision to
From Knowing to Acting in the World
reach is based on much more than whether or not
they have an object concept. Instead, the decision to The importance of Esthers work on infant cog-
reach is based on what the infants have just done, nition is not just in the particular tasks or findings
their reaching skill, the feel of the body, the salience that she and her colleagues eloquently explained.
of the cue, and the perceptual layout of the task. Rather, what we take away from this work is that
Thus, this reaching task that was assumed to provide cognition is embodiedFand this has concrete con-
a direct window into infants abstract concepts is ceptual and methodological consequences. This view
actually a window into the complex interactions of cognition shifts the focus away from what babies
among perception, action, and cognition in infancy. know in the abstract to a shared emphasis on the
1532 Spencer et al.

perceptual and movement dynamics that produce lids and toys to looking at complex stimuli in infant
behavior. The traditional view that reaching or habituation.
looking gives us direct access to the contents of mind These commonalties led Esther to emphasize the
is no longer tenable. As long as our entry into infant role of general processes in shaping development. This
cognition is through reaching or looking, the dy- view ran counter to other modern proposals about
namics of these behaviors must be considered, that the state of developmental theory, which claim that
is, the developmental history and real-time dynam- Piaget was wrong: broad, general explanations
ics of the response cannot be separated from the seem increasingly implausible (Gopnik, 1996, p.
constructs the response is designed to measure. 221) and efforts are better spent working out the
Perception and action are not bystanders in cogni- details, domain-by-domain. The traditional big is-
tion. Rather, cognition always reflects the dynamic sues of developmental theory . . . should be cast aside
interplay of mental and bodily processes embedded in favor of specific theories about content (Thelen &
within a rich context. Bates, 2003, p. 378). In contrast to this view, Esther
and Elizabeth Bates wrote, we state forthrightly that
we do believe that there are general principles of
development: mechanisms and processes that hold
Reaching Out: Dynamic Systems in the World
true whatever the content domain (Thelen & Bates,
Movement played a central role in Esthers career, 2003, p. 378).
not simply in the topics she studied, but movement Esthers pursuit of these general developmental
in her own thinking as well. Two trajectories emerge processes highlights the second trajectory in her
from a survey of her career. The first is movement careerFmovement from metaphor to formal theory.
from studying simple responses to increasingly Her career began with an ethological and ecological
complex phenomena, including cognition. In Es- perspective that included intensive observations of
thers early career, she discovered how infants children in a number of detailed longitudinal studies
assemble simple movements like stepping and of kicking and reaching (e.g., Thelen, 1981b; Thelen
walking on a treadmill from multiple components et al., 1993). This large body of empirical work led
including the spring-like character of leg muscles. initially to a systems view and a rejection of nativism
Later, when studying goal-directed actions like as a viable approach to understanding development.
reaching, she discovered that the characteristics of Later, in a seminal paper coauthored with Beverly
infants bodies as well as their energy levels had a Ulrich, Esther introduced concepts of DST to devel-
profound influence on the emergence of goal-di- opmental researchersFcollective variables, control
rected actions and in generating the torturous paths parameters, nonlinearity, self-organization, attrac-
so typical of early reaching. But through a process of tors, stability, and multicausalityFlaying out a re-
exploration and selection, infants harnessed these cipe on how to study developmental process (Thelen
intrinsic dynamics as body and mind came together & Ulrich, 1991). These ideas were expanded in her
to make straighter, more efficient reaches. These 1994 book with Linda Smith (Thelen & Smith, 1994).
early hints at the embodiment of cognition were This book postulated that systems thinking was not
brought to the foreground as Esther moved into the solely applicable to motor development and motor
domain of infant cognition. Here, she and her col- control; rather, the concepts of DST could be applied
leagues specified the neural dynamics at work as across all domains of development including cogni-
infants made decisions about where to reach in the tion (for elegant demonstrations of this, see Fischer &
A-not-B task or where to look in a habituation task. Bidell, 1998; Lewis, 2000; van der Maas et al., in
Importantly, as Esther moved from simple re- press; Van Geert, 1998). Esther, Linda Smith, and
sponses to increasingly complex phenomena, she did their colleagues demonstrated this forcefully in a
not partition the child up as the focus expanded. series of elegant and counterintuitive experiments on
Rather, she viewed the child-in-context as a complex the Piagetian A-not-B error (Smith et al., 1999).
system of reciprocally coupled and reciprocally Next, Esther sought to formalize the concepts of
interactive components. Moreover, she showed an DST, creating a dialogue between formal theory and
impressive ability to integrate seemingly disparate empirical work. Her initial discussions with Scott
phenomena: from leg muscles and rhythmic move- Kelso led her to a relative phase model of rhythmic
ments to arm muscles and goal-directed actions; movements, but it was her collaboration with Gregor
from reaching movements and movement speed to Schoner and their work on the DFT that solidified
the A-not-B task and the role of motor skill in this the theory experiment link. This link produced two
classic cognitive task; from reaching and looking at formal models of classic findings in infant cognition
 Moving Toward a Grand Theory of Development 1533

(Schoner & Thelen, 2006; Thelen et al., 2001) and has according to DST, behavior is multiply determined and
led to a host of innovations in other domains (e.g., softly assembled from the nonlinear interactions of
Bastian, Schoner, & Riehle, 2003; Erlhagen & Schoner, multiple subsystems. The concept of soft assembly is
2002; Schutte, Spencer, & Schoner, 2003; Spencer & beautifully illustrated in some of Esthers earliest
Schoner, 2003; Spencer, Simmering, & Schutte, 2006). work. For example, in her work on the disappearing
Although Esther had immense enthusiasm for steps, Esther showed how stepping patterns come
formal theoretical work, she always emphasized the and go depending on the weight of the infants legs,
reciprocal interaction among different scientific ap- whether the infant is in water or not, whether the
proaches to the study of development. She refused to infant is upright, lying down, and so on (e.g., Thelen
invest too heavily in a particular mathematical et al., 1984). Note that this concept of soft assembly is
model, preferring to emphasize the importance of critical to allow the child to act in a changing and
conceptual thinking, telling a good story, and having variable world. Moreover, soft assembly provides a
a good metaphor: the role of formal models is to natural foundation for exploration and selection be-
make . . . underlying assumptions extremely precise. cause behavioral patterns are not fixed, but varying
The specific form of the model is, thus, less import- and flexible.
ant than the general principles of development on The third central concept of DST is embodi-
which it is based (Thelen & Bates, 2003, p. 378). mentFperception, action, and cognition form an in-
This emphasis came through in the clarity of her tegrated system that cannot be partitioned. Esther and
writing and the clarity of her thought. As an ex- Linda Smith emphasized this latter point in their 1994
ample, in a wonderful article targeted toward book: We, like the symbolic computational theorist,
undergraduates entitled The Improvising Infant: view cognition as all one kind; but in our view, it is all
Learning about Learning to Move, Esther compared embodied, all distributed, all activity, all a complex
development to improvisational jazz (Thelen, 1998). event in time (Thelen & Smith, 1994, p. 337). The
She discussed how infants create new solutions step embodiment of behavior is, perhaps, best illustrated in
by step over development as they carve out their the weighted limbs studies using the A-not-B task: by
own unique pathway. Thus, development is more simply changing the feel of the arms between the A
like improvisational jazz, with the infant as a musi- and B trials, Esther and her colleagues could create
cian and less like a mechanistic process driven by perseverative or accurate responding in the task
genes. And, as with jazz, the music infants create as (Diedrich, Clearfield, Smith, & Thelen, 2005). Fourth,
they learn to move and explore must be considered DST shows a new respect for individuality. Develop-
as a whole pattern rather than a sequence of indi- ment happens in individual children solving individ-
vidual notes. Note that Esther actually brought a ual problems in their own unique ways. This theme is
variant of this metaphor into reality by collaborating beautifully illustrated in the story of Gabriel and
with Helga Winold to study the dynamics of ex- Hannah learning to reach (e.g., Thelen et al., 1996).
pertise in cello performance (Winold, Thelen, & Putting these themes together, we can sketch the
Ulrich, 1994). view of development that Esther championed (see
Figure 7). We start with a child-in-context composed
of multiple components at different levels of an in-
tegrated system. These components include neural
A New Grand Theory of Development
dynamics captured by the DFT (Bastian, Riehle,
Given Esthers emphasis on general develop- Erlhagen, & Schoner, 1998; Erlhagen, Bastian, Jancke,
mental processes, we can ask where the two trajec- Riehle, & Schoner, 1999; Jancke et al., 1999; Thelen et
tories of her career led. The answer is that they led to al., 2001), neural oscillations critical to motor control
a new grand theory of development, DST. Four and rhythmic actions (Kelso, 1995), the spring-like
central concepts of the theory emerged across Es- characteristics of muscles (e.g., Thelen & Ulrich,
thers career and have been evident in the examples 1991), and so on. Critically, these components are
mentioned previously. First, DST creates a new em- fully embodied and reciprocally coupled (see the
phasis on timeFbehavior emerges in the moment, but bidirectional arrows in Figure 7). Moreover, they are
the effects of each behavioral decision accumulate coupled together in a softly assembled way that is
over longer time scales, as each change sets the stage grounded in the sensorimotor world.
for future changes. This theme is evident in Esthers What happens, then, as the child interacts with the
many detailed longitudinal studies showing a cas- world from second to second, minute to minute,
cade of influences over different time scales (e.g., situation to situation? Over short time scales, learn-
Thelen & Ulrich, 1991; Thelen et al., 1993). Second, ing occurs. What is learning from a dynamic systems
1534 Spencer et al.

Changing stability
within and across
levels of system

learning

Flexibly shifting
development among stable
solutions as
situation chanes
Figure 7. A depiction of learning and development from a dynamic systems perspective. Each image of the infant captures one time point
in learning/development with time moving from left to right. The infant is viewed as an integrated system consisting of multiple, re-
ciprocally coupled components (see the bidirectional arrows) embedded within a specific context. The components depicted include
neural dynamics (captured by the simulation of the DFT in the infants head), oscillatory dynamics (see oscillation in the brain and spinal
cord), and the springy character of muscles (see springs in the arm and leg). Learning is about changing stability within and across levels.
This is illustrated by the solid line highlighting simulation in the infants head, and the shift from dotted arrows in the first learning step to
solid arrows in the second. Development is about flexibly shifting among stable solutions, that is, being able to flexibly shift from the
dynamic organization needed to reach for the ball, to the dynamic organization needed to reach for the box, to the dynamic organization
needed to grab the blanket to slide the ball forward (see dashed highlighting and dashed arrows in the final time step).

view? At a general level, the infant is carving out (see Figure 7). Note that these solutions are particu-
individual solutions to the real-world problem she is lar to this child and this developmental trajectory is
facing, in this case, grabbing an attractive ball (see unique to this child. Nevertheless, the processes that
Figure 7). In dynamic systems concepts, she is work over these time scales are general.
forming stable attractors or patterns. Such stability Esther and Linda Smith conveyed their excite-
can emerge at one levelFfor instance, at the level of ment about this view of development in their 1994
neural dynamics captured by the DFTFor stability book: What can a dynamic approach do? A dy-
can be reflected in changes in the coupling of com- namic approach can change the way you think about
ponents (see the solid arrows in Figure 7). Import- development and it can change the way you conduct
antly, though, stability is not an end-stateFthere is research in development. Once we began to view
always a delicate balance between stability and in- development from a dynamic and selectionist ap-
stability. This allows for improvisation on a theme; proach, we found the ideas so powerful that we
for the infant to use stable solutions, and to also could never go back to other ways of thinking. Every
discover novel solutions that arise through explor- paper we read, every talk we heard, every new bit of
ation and often through accident. data from our labs took on new meaning . . . The final
Moving forward, these real-time and learning- test of dynamics in development, of course, is in its
time processes are integrated over longer time scales usefulness to a wide range of scholars. We hope
to form a unique developmental trajectory. Here, the readers will accept the challenge of the new way of
infant continues to carve out stable solutions, and thinking and working and we look forward to the
she learns to flexibly shift from one solution to an- report card (Thelen & Smith, 1994, pp. 341 342).
other as the situation or her motivations change.
Thus, she can flexibly shift from one stable pattern
Looking to the Future: Challenges for a Dynamic Systems
when reaching for the ball, to another stable pattern
Approach
when confronted with two identical hiding locations
in an A-not-B situation, to another stable pattern as As with all grand theories, there are many chal-
she has to reach for the blanket to retrieve the ball lenges that lie ahead. We highlight two that were
 Moving Toward a Grand Theory of Development 1535

emerging as critical next steps in Esthers work. The the influences of whole families and large social and
first challenge is to create dynamical systems that cultural groups (for steps in this direction, see Fogel,
change themselves and, in this way, begin to integrate Nwokah, Dedo, & Messinger, 1992; Lewis, 2000).
the multiple time scales of development. Here, Es- Second, Esther sought a stronger emphasis on dialog
ther thought dynamical systems theorists might with parents, practitioners, and policymakers to
borrow insights from connectionist approaches to translate theory into practice with infants, children,
learning (for related ideas, see Spencer & Thelen, and adults needing assistance. Thus, Esther sought
2003; Spencer, Thomas, & McClelland, in press). to take her knowledge of empirical findings, con-
Nevertheless, Esther thought that learning extended ceptual theory, and formal theory and bring it to the
beyond what could be captured by accumulating the level of individual children in the world.
statistics of experience. Rather, she thought a new We have already mentioned two of Esthers
view of learning might emerge by integrating translational projects: working with physical and
insights from connectionism with the rich under- occupational therapists to improve the lives of, for
standing of real-time dynamics she and her col- instance, children with Down syndrome (Ulrich
leagues explored (e.g., Schoner & Thelen, 2006; et al., 2001) and working with musical scholars to
Thelen et al., 2001). Esther also thought that devel- study the development of expertise in the training of
opmental scientists might gain insights about inte- cellists (Winold et al., 1994). A third translational
grating time scales by looking into the field of effort played a central role in the final phase of Es-
developmental robotics to discover the contingencies thers career: she became a Guild Certified Feldenkrais
that may structure the learning experience of the PractitionerCM. Here, she asked whether the prin-
perceiving and acting infant. ciples of DST that shed light on motor development
A second challenge facing DST is to develop dy- could foster a deeper understanding of the Fel-
s
namic systems that create themselves through pro- denkrais Method of somatic education. Importantly,
cesses of exploration and selection. The theme of she did not view this translational project as a one-
exploration and selection was present in Esthers way flow of information from researchers to the
work quite early and played a central role in her 1994 applied world. Rather, she had a vision of reciprocal
book with Linda Smith. There, they incorporated dialogFthat practitioners could gain theoretical
Gerald Edelmans ideas about Neural Darwinism grounding, but basic researchers could gain insights
(see Edelman, 1987) to understand how nervous into the generality of ideas as well as novel insights
systems can create new insights about the world via into developmental process.
the redundant, degenerate neural signals that Why was Esther drawn to the Feldenkrais Method?
underlie experience. To date, however, it is not clear The Feldenkrais Method uses subtle variation, innov-
how to integrate this view with, for instance, the ation, and explicit differentiation of the perception of
richly structured behavioral dimensions and neural movements to free people from habitual patterns
dynamics that underlie the DFT (Thelen et al., 2001). and allow new movement solutions to emerge. Thus,
Moreover, a critical challenge with selectionist ideas Feldenkrais practitioners introduce novel and/or dif-
is to understand generalizationFhow the novel as- ficult movement problems to help bring movement
sociations and insights created in-the-moment are perception into awareness. This also serves to foster
integrated together to extend across situations. Es- new modes of coordination and to destabilize old
ther thought that action might provide a common habits, allowing the discovery of new ways to move.
link across diverse experiences, given the embodi- To Esther and to many others, this view of move-
ment of cognition and the fact that the body is al- ment education has a natural affinity with concepts
ways present, shaping and structuring experience. of DST (Buchanan & Ulrich, 2001). Thus, toward the
end of her career, Esther began to conduct research
studies on the effects of the Feldenkrais Method on
Looking to the Future: Reaching Out
healthy adults; she participated in and organized
Another kind of challenge is to reach out and symposia to promote dialog among researchers and
bring DST into the world. There are two senses in Feldenkrais practitioners; and she looked forward to a
which this needs to be done. First, Esther thought second career giving Feldenkrais lessons to infants
that DST needed a richer sense of action in the real and children. Sadly, this second career was not
world to capture the emotional and motivational realized, but Esthers efforts to reach out and bring
side of development. Thus, there must be more than dynamic systems concepts into the world continue.
toys and treadmills in the childs world (see Figure We end with Esthers own wordsFwith a tran-
7); we must add mothers and fathers and consider script from a conversation between Roger Russell
1536 Spencer et al.

(a Certified Feldenkrais Trainer) and Esther that took been quite attractive and Im pleased, if, in fact,
place in Amsterdam in 1998. In this interview, Esther this will happen that these things will come to-
expresses the pleasure she received from translating gether.
dynamic systems ideas to others and her vision of Roger Russell: What pleases you about it?
the future. Central to this vision was the hope that Esther Thelen: Well, its nice to know that
her theoretical ideas would be useful, not just to re- something that youve done and that your col-
searchers in the field of child development, but to leagues and your students have done actually
people in general. That is, she hoped these ideas may be useful to someone, other than just, you
would have a profound effect on the way people know, giving your papers to other academics!
think about development, as well as how they think And especially . . . if it will inspire therapists to
of themselvesFas embodied, grounded, ever-chan- think differently and maybe try new things. I
ging, ever improvising people in the world. think that would be a lovely way of translating
basic laboratory and theoretical work. I started in
Roger Russell: Youve been involved in the aca- this field because Im interested in how things
demic field where most of the people who listen to work. I want to know, gee, why does it change
you are other academics asking academic ques- like this? What happens? without much thought
tions. of, well, is this useful to anyone. Basic, just basic
Esther Thelen: Ive actually talked to a lot of scientific curiosity. But, if at the same time it turns
practitioners of various kinds. out that it can be useful at least a little bit, then
Roger Russell: What interests them [the practi- thats wonderful!
tioners] in what youre saying?
Esther Thelen: I think that people dealing with
adults and children who have some sort of prob- References
lem . . . children with speech problems, with mo-
tor problems, adults with problems, emotional Andre-Thomas, & Autgaerden, S. (1966). Locomotion from
pre- to postnatal life. In S. Society (Ed.), Clinics in de-
and so on, are now really very interested in the
velopmental medicine (Vol. 24). London: William Heine-
parallels . . . between the processes of change as
mann.
they occur in development and the processes of Angulo-Barroso, R. M., Ulrich, D. A., & Tiernan, C. (2004,
change as they may happen in some sort of edu- March). Motor activity and adapted treadmill stepping
cational or therapeutic encounter. And a lot of patterns in infants born at risk for cerebral palsy. Paper
people have told me that theyre looking for some presented at the Gatlinburg Conference, San Diego, CA.
theoretical basis for what they do. Ive had this Bastian, A., Riehle, A., Erlhagen, W., & Schoner, G. (1998).
happen over and over where people will come up Prior information preshapes the population representa-
to me and say, well, this is very interesting and tion of movement direction in motor cortex. NeuroReport,
you are describing perfectly what I do in my 9, 315 319.
speech therapy! or what I do in my psychother- Bastian, A., Schoner, G., & Riehle, A. (2003). Preshaping
and continuous evolution of motor cortical representa-
apyFthis is how I approach therapy! But I have
tions during movement preparation. European Journal of
not had a theoretical kind of rationale for it before
Neuroscience, 18, 2047 2058.
[that is] so explicit about how people change. Bjork, E. L., & Cummings, E. M. (1984). Infant search er-
So, to have it spelled out as a developmental rors: Stage of concept development or stage of memory
processFI think its very honest to put change, development. Memory and Cognition, 12, 1 19.
dysfunctional outcomes and functional outcomes Bremner, J. G. (1978). Spatial errors made by infants: In-
all in the same vocabulary. In other words, people adequate spatial cues or evidence of egocentrism? British
are . . . looking for models that are not just medical Journal of Psychology, 69, 77 84.
models of how things go badly and how you Buchanan, P. A., & Ulrich, B. D. (2001). The Feldenkrais
s

change them. And the implications of the kinds of Method : A dynamic approach to changing motor
ways that were thinking are that you may have behavior. Research Quarterly for Exercise and Sport, 72,
315 323.
dysfunctional outcomes out of the same pro-
Bushnell, E. (1985). The decline of visually guided reaching
cessesFdevelopmental processesFthat led to
during infancy. Infant Behavior and Development, 8, 139
more adaptive outcomes . . . Therefore, you can 155.
look toward developmental processes to think Butterworth, G., Jarrett, N., & Hicks, L. (1982). Spatio-
about how to move people beyond their . . . mal- temporal identity in infancy: Perceptual competence
adaptive ways of interacting either in relation- or conceptual deficit? Developmental Psychology, 18,
ships or in their skills or so on. So I think thats 435 449.
 Moving Toward a Grand Theory of Development 1537

Clearfield, M. W., Diedrich, F. J., Smith, L. B., & Thelen, E. book of child psychology (5th ed.) Vol. 1: Theoretical models of
(2006). Piaget was right: Thoughts close to the sensori-motor human development (pp. 467 561). New York: Wiley.
surface. Manuscript in preparation. Fogel, A., Nwokah, E., Dedo, J. Y., & Messinger, D. (1992).
Clearfield, M. W., Dineva, E., Smith, L. B., Diedrich, F. J., & Social process theory of emotion: A dynamic systems
Thelen, E. (2006). Cue salience and memory in infant per- approach. Social Development, 1, 122 142.
severative reaching: A test of the Dynamic Field Model. Gopnik, A. (1996). The post-Piaget era. Psychological Sci-
Manuscript under review. ence, 7, 221 225.
Clearfield, M. W., Smith, L. B., Diedrich, F. J., & Thelen, E. Heriza, C. (1988). Organization of leg movements in
(2006). Young infants reach correctly on the A-not-B preterm infants. Physical Therapy, 68, 1340 1346.
task: On the development of stability and perseveration. Jancke, D., Erlhagen, W., Dinse, H. R., Akhavan, A. C.,
Infant Behavior and Development, 29, 435 444. Giese, M., Steinhage, A., et al. (1999). Parametric popu-
Clearfield, M. W., & Thelen, E. (2001). Stability and flexi- lation representation of retinal location: Neuronal
bility in the acquisition of skilled movement. In C. interaction dynamics in cat primary visual cortex. Jour-
Nelson (Ed.), Handbook of developmental cognitive neuro- nal of Neuroscience, 19, 9016 9028.
science (pp. 253 266). Cambridge, MA: MIT Press. Kelso, J. A. S. (1995). Dynamic patterns: The self-organization
Corbetta, D., & Thelen, E. (1994). Shifting patterns of of Brain and behavior. Cambridge, MA: The MIT Press.
interlimb coordination in infants reaching: A case study. Kelso, J. A. S., Holt, K. G., Kugler, P. N., & Turvey, M. T.
In S. P. Swinnen, H. Heuer, J. Massion, & P. Casaer (1980). On the concept of coordinative structures as
(Eds.), Interlimb coordination: Neural, dynamical, and cog- dissipative structures: II. Empirical lines of convergence.
nitive constraints (pp. 413 438). San Diego, CA: Aca- In G. E. Stelmach & J. Requin (Eds.), Tutorials in motor
demic Press. behavior (pp. 49 70). New York: North Holland.
Corbetta, D., & Thelen, E. (1996). The developmental ori- Kelso, J. A. S., & Tuller, B. (1984). A dynamical basis for
gins of two-handed coordination: A dynamic perspec- action systems. In M. S. Gazzaninga (Ed.), Handbook of
tive. Journal of Experimental Psychology: Human Perception cognitive neuroscience (pp. 321 356). New York: Plenum.
and Performance, 22, 502 522. Lewis, M. D. (2000). The promise of dynamic systems
Corbetta, D., & Thelen, E. (1999). Lateral biases and fluc- approaches for an integrated account of human devel-
tuations in infants spontaneous arm movements and opment. Child Development, 71, 36 43.
reaching. Developmental Psychobiology, 34, 237 255. Mandler, J. M. (1988). How to build a baby: On the de-
Corbetta, D., & Thelen, E. (2002). Behavioral fluctuations velopment of an accessible representational system.
and the development of manual asymmetries in infancy: Creative Development, 3, 113 136.
Contribution of the dynamic systems approach. In S. J. Marcovitch, S., & Zelazo, P. D. (1999). The A-not-B error:
Segalowitz & I. Rapin (Eds.), Handbook of neuropsychology Results from a logistic meta-analysis. Child Development,
(Vol. 8): Child neuropsychology, Part I (pp. 309 328). 70, 1297 1313.
Amsterdam: Elsevier Science Publishing Co. Munakata, Y. (1998). Infant perseveration and implications
Diedrich, F. J., Clearfield, M. W., Smith, L. B., & Thelen, E. for object permanence theories: A PDP model of the AB
(2005). Motor memory influences performance in the A-not-B task. Developmental Science, 1, 161 184.
task: Effects of arm-loading and posture. Unpublished Oppenheim, R. W. (1981). Ontogenetic adaptations and
manuscript. retrogressive processes in the development of the ner-
Diedrich, F. J., Highlands, T., Thelen, E., & Smith, L. B. vous system and behaviors: A neuroembryological
(2001). The role of target distinctiveness in infant per- perspective. In H. F. R. Prechtl & K. Connolly (Eds.),
severative reaching errors. Journal of Experimental Child Maturation and development: Biological and psychological
Psychology, 78, 263 290. perspectives (pp. 73 109). Philadelphia: Spastics Inter-
Diedrich, F. J., Thelen, E., Smith, L. B., & Corbetta, D. national Medical Publications by William Heinemann
(2000). Motor memory is a factor in infant perseverative Medical Books.
errors. Developmental Science, 3, 479 494. Peiper, A. (1963). Cerebral function in infancy and childhood.
Edelman, G. M. (1987). Neural Darwinism. New York: Basic New York: Consultants Bureau.
Books. Piaget, J. (1954). The construction of reality in the child. New
Erlhagen, W., Bastian, A., Jancke, D., Riehle, A., & Schoner, York: Basic Books.
G. (1999). The distribution of neuronal population acti- Schoner, G., & Thelen, E. (2006). Using dynamic field the-
vation (DPA) as a tool to study interaction and integra- ory to rethink infant habituation. Psychology Review, 113,
tion in cortical representations. Journal of Neuroscience 273 299.
Methods, 94, 53 66. Schutte, A. R., Spencer, J. P., & Schoner, G. (2003). Testing
Erlhagen, W., & Schoner, G. (2002). Dynamic field theory of the dynamic field theory: Working memory for locations
movement preparation. Psychological Review, 109, 545 becomes more spatially precise over development. Child
572. Development, 74, 1393 1417.
Fischer, K. W., & Bidell, T. R. (1998). Dynamic development Smith, L. B., Thelen, E., Titzer, R., & McLin, D. (1999).
of psychological structures in action and thought. In W. Knowing in the context of acting: The task dynamics of
Damon (Editor-in-Chief) & R. M. Lerner (Eds.), Hand- the A-not-B error. Psychological Review, 106, 235 260.
1538 Spencer et al.

Spelke, E. S. (1998). Nativism, empiricism, and the origins of Thelen, E., Corbetta, D., & Spencer, J. P. (1996). The
knowledge. Infant Behavior and Development, 21, 181 200. development of reaching during the first year: The
Spencer, J. P., & Schoner, G. (2003). Bridging the repre- role of movement speed. Journal of Experimental Psych-
sentational gap in the dynamical systems approach to ology: Human Perception and Performance, 22, 1059
development. Developmental Science, 6, 392 412. 1076.
Spencer, J. P., Simmering, V. R., & Schutte, A. R. (2006). To- Thelen, E., & Farish, D. J. (1977). An analysis of the
ward a formal theory of flexible spatial behavior: Geo- grooming behavior of wild and mutant strains of Bracon
metric category biases generalize across pointing and hebetor (Braconidae: Hymenoptera). Behaviour, 62, 699
verbal response types. Journal of Experimental Psychology: 715.
Human Perception and Performance, 32, 473 490. Thelen, E., Fisher, D. M., & Ridley-Johnson, R. (1984). The
Spencer, J. P., & Thelen, E. (2000). Spatially specific changes relationship between physical growth and a newborn
in infants muscle co-activity as they learn to reach. In- reflex. Infant Behavior and Development, 7, 479 493.
fancy, 1, 275 302. Thelen, E., Kelso, J. A. S., & Fogel, A. (1987). Self-organiz-
Spencer, J. P., & Thelen, E. (2003). Connectionism and ing systems and infant motor development. Develop-
dynamic systems theory: Are these really different mental Review, 7, 39 65.
approaches to development? Developmental Science, 6, Thelen, E., Schoner, G., Scheier, C., & Smith, L. B. (2001).
375 447 (Special Issue). The dynamics of embodiment: A field theory of infant
Spencer, J. P., Thomas, M. S., & McClelland, J. L. (Eds.). (in perseverative reaching. Behavioral and Brain Sciences, 24,
press). Toward a new grand theory of development? Con- 1 86.
nectionism and dynamic systems theory re-considered. New Thelen, E., & Smith, L. B. (1994). A dynamic systems approach
York: Oxford University Press. to the development of cognition and action. Cambridge, MA:
Spencer, J. P., Vereijken, B., Diedrich, F. J., & Thelen, E. MIT Press.
(2000). Posture and the emergence of manual skills. Thelen, E., & Ulrich, B. D. (1991). Hidden skills: A dynamic
Developmental Science, 3, 216 233. systems analysis of treadmill stepping during the first
Thelen, E. (1979). Rhythmical stereotypies in normal hu- year. Monographs of the Society for Research in Child De-
man infants. Animal Behaviour, 27, 699 715. velopment, 56 (1, Serial No. 223).
Thelen, E. (1981a). Kicking, rocking, and waving: Con- Touwen, B. (1976). Neurological development in infancy.
textual analysis of stereotyped behaviour in normal in- London: Spastic International Medical Publishers.
fants. Animal Behaviour, 29, 3 11. Ulrich, B. D., & Moerchen, V. (2005). Infants with spina bifida:
Thelen, E. (1981b). Rhythmical behavior in infancy: An Body weight supported treadmill stepping over the first
ethological perspective. Developmental Psychology, 17, postnatal year. Paper presented at the Motor develop-
237 257. ment and learning in infancy II: Fundamental concepts
Thelen, E. (1985). Developmental origins of motor coord- and applications, Murcia, Spain.
ination: Leg movements in human infants. Developmental Ulrich, D. A., Ulrich, B. D., Angulo-Barroso, R. M., & Yun,
Psychobiology, 18, 1 22. J.-K. (2001). Treadmill training of infants with Down
Thelen, E. (1998). The improvising infant: Learning about syndrome: Evidence-based developmental outcomes.
learning to move. In J. S. DeLoache, S. C. Mangelsdorf, & Pediatrics, 108, 1 7.
E. Pomerantz (Eds.), Current readings in child development van der Maas, H. L. J., Dolan, C. V., Grasman, R. P. P.,
(3rd ed., pp. 26 42). Boston: Allyn and Bacon. Wicherts, J. M., Huizenga, H. M., & Raijmakers, M. E. J.
Thelen, E. (2000). Motor development as foundation and (in press) A dynamical model of general intelligence:
future of developmental psychology. International Jour- The positive manifold of intelligence by mutualism.
nal of Behavioral Development, 24, 385 397. Psychological Review.
Thelen, E., & Bates, E. A. (2003). Connectionism and dy- Van Geert, P. (1998). A dynamic systems model of basic
namic systems: Are they really different? Developmental developmental mechanisms: Piaget, Vygotsky, and be-
Science, 6, 378 391. yond. Psychological Review, 105, 634 677.
Thelen, E., & Corbetta, D. (1994). Exploration and selection Wellman, H. M., Cross, D., & Bartsch, K. (1987). Infant
in the early acquisition of skills. International Review of search and object permanence: A meta-analysis of the A-
Neurobiology, 37, 75 102. not-B error. Monographs of the Society for Research in Child
Thelen, E., Corbetta, D., Kamm, K., Spencer, J. P., Schnei- Development, 51, 1 51.
der, K., & Zernicke, R. F. (1993). The transition to Winold, H., Thelen, E., & Ulrich, B. D. (1994). Coordination
reaching: Mapping intention and intrinsic dynamics. and control in the bow arm movements of highly skilled
Child Development, 64, 1058 1098. cellists. Ecological Psychology, 6, 1 31.