Embodied Cognition - Monica Cowart
Embodied Cognition - Monica Cowart
Embodied Cognition - Monica Cowart
Table of Contents
1. Motivation for the Movement
2. General Characteristics of Embodied Cognition
1. Primacy of Goal-Directed Actions Occurring In Real-Time
1. Developmental Psychology
2. Robotics/Artificial Life
ii. Form of Embodiment Constrains Kinds of Cognitive Processes
iii. Cognition is Constructive
b. Embodied Cognition vs. Classicism/Cognitivism
c. Philosophical Implications of the Embodied Cognition Research Program
i. The Compatibilist Approach
ii. The Purist Approach
b. References and Further Reading
1. Motivation for the Movement
Although ideas applied in the embodied cognition research program can be traced
back to the seminal works of Heidegger, Piaget, Vygotsky, Merleau-Ponty, and
Dewey, the current thesis can be seen as a direct response and, in some cases, a
proposed alternative to the cognitivist/classicist view of the mind, which
conceptualizes cognitive functions in terms of a computer metaphor. The
cognitivist/classicist research program can be defined as a rule-based, information-
processing model of cognition that 1) characterizes problem-solving in terms of
inputs and outputs, 2) assumes the existence of symbolic, encoded representations
which enable the system to devise a solution by means of computation, and 3)
maintains that cognition can be understood by focusing primarily on an organism’s
internal cognitive processes (that is, specifically those involving computation and
representation). Although this research program is still prevalent, a number of
problems have been raised about its viability, including the symbol-grounding
problem (Searle 1980, Harnad 1990), the frame problem, the common-sense
problem (Horgan and Tienson 1989), and the rule-described/expertise problem
(Dreyfus 1992).
To say that cognition is embodied means that it arises from bodily interactions with the
world. From this point of view, cognition depends on the kinds of experiences that come
from having a body with particular perceptual and motor capacities that are inseparably
linked and that together form the matrix within which memory, emotion, language, and all
other aspects of life are meshed. The contemporary notion of embodied cognition stands in
contrast to the prevailing cognitivist stance which sees the mind as a device to manipulate
symbols and is thus concerned with the formal rules and processes by which the symbols
appropriately represent the world (xx).
Although embodied cognition accounts vary significantly across disciplines in terms
of the specific ways in which they attempt to apply the general theory, a few common
theoretical assumptions can be found in just about any embodied view one examines.
These further theoretical assumptions help to flesh out the central thesis, and include
1) the primacy of goal-directed actions occurring in real-time; 2) the belief that the
form of embodiment determines the type of cognition; and 3) the view that cognition
is constructive. Each theoretical assumption will be explained by considering the
work of a theorist whose research exemplifies the particular theoretical assumption
under investigation. The first theoretical assumption, the primacy of goal-directed
actions occurring in real time, is explained by considering research in
robotics/artificial life and developmental psychology.
i. Developmental Psychology
In order to understand how infants learn to reach, Thelen and Smith (1994)
examined four different infants from the time the babies were 3 weeks old until they
were 1 year old. What Thelen and Smith conclude is that each of the four infants
faced unique problems in learning to reach based on their individual energy level,
body mass and the different ways in which they initially tried to reach (that is, their
pre-reaching behaviors). Given these different pre-reaching movements, each of the
infants had to learn a different set of strategies for controlling their arms so that the
ultimate solution was specifically tailored to address the unique problem the
particular infant was encountering. Thus, each infant was eventually able to
overcome these developmental obstacles and learn to reach the toys, but the specific
ways in which they learned this behavior varied depending upon the specific problem
they were encountering. To understand how these different reaching problems
translated into unique reaching solutions, let’s consider two of the infants whose
reaching approaches varied considerably: Gabriel and Hannah.
Thelen and Smith describe Gabriel as an extremely active infant who was initially
unable to successfully reach the toy because he would excitedly flap his arms, in
seemingly random movements that were not focused enough to enable him to obtain
the toy. Consequently, he had to learn to control these energetic movements so that
this energy would become more focused. By learning to control these excited
movements, he would then be able to produce a more controlled reaching-action that
would propel his hand to the desired location. Gabriel eventually learned to reach
toys after multiple unsuccessful attempts; however, these unsuccessful reaching
attempts were instrumental in helping him realize how to adjust his muscle patterns
so that a successful reaching pattern finally emerged that enabled him to focus his
energy in the direction of the toy.
Hannah’s problem was different from Gabriel’s, but it was also the same. She, like Gabriel,
had to adjust the energy of forces moving her arm—in her case to make her arm sufficiently
stiff or forceful to lift it off her lap. What Gabriel and Hannah had in common, therefore, was
the ability to modulate the forces they delivered to the arms to change their ongoing, but non-
functional patterns to movements that brought their hands close enough to the toys for them
to make contact. Their solutions were discovered in relation to their own situations, carved
out of their individual landscapes, and not pre-figured by a synergy known ahead by the brain
or the genes (260).
The importance of Thelen and Smith’s research becomes clear when we contrast
their conclusions with the manner in which change is explained in other leading
developmental theories. Thelen notes that in other theories change is explained by
appealing to “some deus ex machina—’the genes,’ ‘maturation of the brain,’ ‘a shift
into a new stage,’ or ‘an increase in information-processing capacity’” (Thelen 1995:
91). Such moves are problematic, Thelen argues since they merely push the level of
explanation back a step so that in order to fully understand how change occurs this
new theoretical mechanism must also be explained. Moreover, Thelen notes that the
unique problems encountered and solved by individual infants make it extremely
unlikely that the solutions were innate, since no internal mechanism could know in
advance the specific “energy parameters of the system” (Thelen 1995: 90).
One advantage of a dynamic systems analysis is that it can account for how different
infants must learn unique pre-reaching strategies based on their specific energy level,
body mass and the different ways in which they initially tried to reach (that is, their
pre-reaching behaviors). Yet, despite these different techniques, Thelen and Smith’s
account still identifies the common factors that all of the infants had to learn to
control: the various forces surrounding arm control, such as gravitational resistance.
By developing a dynamical systems analysis of reaching behavior, Thelen and Smith
provide a theoretical mechanism that tries to explain the exact way in which these
different forces interact. The resulting analysis tracks how activity brings about
changes in the system, so that new types of behavior emerge from behaviors the
system already knows. This means of generating new patterns from those that
already exist results in ‘environmental scaffolding’, since a new behavior is generated
from the current resources of the system. Moreover, this dynamic systems analysis
enables the researcher to track how the different movements/actions change and
evolve over time. Consequently, behaviors, such as reaching, are explained in terms
of interactive forces, which are mathematically understood since they are grounded
in the physics of action.
These early movements often look to be entirely without form or meaning. But if what
neuroscientists tell us about the plasticity of the brain and how it changes is correct, infants
are also continually learning something about the perceptual-motor systems and their
relations to the world in their repeated spontaneous activity. That is, what infants sense and
what they feel in their ordinary looking and moving are teaching their brains about their
bodies and about their worlds. They are in fact exploring what range of forces delivered to
their muscles get their arms in particular places and then learning from their exploration,
remembering how certain categories of forces get their hands forward toward some-thing
interesting (90).
Consequently, infants must learn how to perform certain activity patterns, such as
reaching, and then remember when it is appropriate to generate those patterns again
to achieve a desired goal. In order to effectively perform these behaviors at the
appropriate times, the infant must learn to categorize particular situations and
correctly apply the action solution that corresponds with that situation. For example,
if a baby learns how to control its arm muscles so that it can reach a toy it desires,
then it will not take long for the infant to realize that the same type of reaching
behavior can also be used to grasp food. It is in this sense that the behaviors become
generalized as the infant learns to use its body to explore its environment. Moreover,
one might argue that the generalized categories formulated to perform these
reaching behaviors could be viewed as one instance of intentional categorization
emerging from action of a dynamical system.
Herbert was designed to wander around the MIT lab disposing of empty soda cans.
Although Herbert’s task might seem relatively simple, to accomplish it successfully
he had to perform a number sub-tasks; including identifying empty soda cans from
full ones, avoiding the stationary tables and chairs in his path, and maneuvering
around the seldom-stationary people who also inhabit the lab. In order to efficiently
accomplish his task of can removal, Herbert relied on what Brook’s called a
“subsumption architecture,” which consisted of a number of connected layers, each
responsible for performing a specific task; actions emerged from the suppression or
activation of various sub-systems. As Herbert moved through his environment, he
continuously encountered stimuli, which dictated which layer was activated at any
given time. For instance, once Herbert’s object-detection layer successfully detected a
wall obstructing its path, it activated the object-avoidance layer, which shut down the
layer responsible for forward motion. The various connected layers plus the
environmental stimuli ultimately determine the suppression or activation of a
particular layer. Brooks argued that the subsumption architecture enables Herbert to
“use the world as its own best representation” since Herbert does not need to refer to
a detailed map of his surroundings before determining how to react. Instead, in
systems such as Herbert, an effective interface is continually recreated between the
system and the world without relying on a central planning facility to dictate
commands, or encoding classicist representations.
In summary, Brooks’ research in artificial life, as well as the research of many other
roboticists (see also Mataric 1992, Agre and Chapman 1997, Tilden 1999, Mataric,
Clancey 1997), helps to clarify the first theoretical assumption of embodied
cognition: the primacy of goal-directed action occurring in real time. One reason that
Brooks’ research is an excellent example of this theoretical assumption is his
emphasis on developing robots that employ quick, cost-effective solutions to
“everyday” problems encountered in an environment. Although much more progress
needs to occur in Artificial Life before architectures are developed that are capable of
explaining behaviors associated with higher cognitive processes, these early
architectures are still able to do something the classicist/cognitivist systems have
not: provide a preliminary attempt at modeling some of the simple, low-level
behaviors that are necessary for survival.
In addition, the earlier examination of Thelen and Smith’s research provides us with
another example of why embodied cognition accounts maintain that action occurring
in real time is the essential to understanding cognitive development. Specifically, a
dynamic systems analysis is capable of tracking the way in which behaviors evolve
and unfold over time; this real-time analysis is completely missing from current
classicist/cognitivist accounts of developmental change.
To illustrate this point, consider how two very different organisms, a child and a
puppy, will try to play with a ball. If the child wishes to get the ball, she will most
likely use her hands, but she could also use her feet. Yet, she will not normally use
her mouth to get the ball, even if the size of the ball does not preclude this option.
This is because, aside from being culturally frowned upon, the other options enable
greater control, are easier to perform, and are culturally sanctioned. However, a
puppy has fewer options, and will most likely grab the ball with its mouth, since its
particular form of embodiment will not enable it to grasp the ball with its paws.
Although there are further differences related to how the child and puppy can
perceive and interact with the ball, including the fact that the child’s visual system
will include color cues, while the dog’s visual system will only enable it to see the ball
in black and white, the important point is that, in each case, the way the organism is
embodied constrains the options available to it.
A further point is that each of these different types of interactions (that is, grabbing
with one’s hands, clutching with one’s mouth, pouncing with one’s paws, etc.) has its
own set of corresponding sensorimotor experiences, which directly influence how the
organism interacts with the object. This is because the continuous feedback from
these sensorimotor experiences serves as the basis for how the organism understands
a specific interaction. Moreover, since activities always take place in a specific
environmental context, such as when a child plays soccer with a friend on a spring
day, the sensorimotor driven understanding of the situation that is gained from
performing the activity in these circumstances can further inform how the organism
might carry out future attempts at performing the same activity.
In general, environmental factors are very important because they can influence not
only what options are available to a particular organism, but also why an organism
might choose one option over another when performing a particular goal-directed
activity. For instance, weather conditions, the size of the ball, the rules of the game,
and whether or not an individual has any broken limbs will most likely factor into
their decision to throw the ball, or kick it. Yet, all of this person’s past experiences
with an object in these varied activity-based contexts will in some way contribute to
their current understanding of the activity. The individual’s understanding of these
past experiences is directly informed by the kinds of sensorimotor experiences their
form of embodiment allows.
In conclusion, the way in which we are embodied determines the type of action
patterns we can perform and these action patterns shape our cognitive functions
(that is, the way in which we can conceptualize and categorize). This is because most
embodied cognition theorists argue that category and concept formation is made
possible and constrained by the particular sensorimotor experiences of the organism.
It is in this sense that the form of embodiment partly determines the kind of
cognitive processes available to the organism. Psychologists, such as Barsalou (1983,
1997), Glenberg (1997,1999), and Thelen and Smith (1994), are but a few of the
cognitive scientists who adopt this theoretical assumption even though the specific
content of their individual views varies. For instance, Glenberg (1997) illustrates how
cognition results from embodiment due to’mesh,’ which refers to the particular way
in which affordances, knowledge, and goals combine. Yet, Barsalou (1997) develops a
theory of simulation, and as demonstrated earlier, Thelen and Smith (1994) explain
the emergence of this theoretical assumption according to a dynamical systems
framework. Thus, all of these individuals agree with the theoretical assumption that
the form of embodiment partly determines the cognitive processes available to the
organism, but they still debate precisely how this occurs.
c. Cognition is Constructive
If the way we conceptualize and categorize is based on the way we are embodied,
then according to embodied cognition theorists these concepts and categories are
actively constructed and not merely apprehended wholesale from an observer-
independent environment. The point here is that the way in which we are embodied
not only constrains the way we can interact in the world, but our particular form of
embodiment also partly determines the way the world appears to us. In effect, it does
not follow from the existence of an observer-independent world that this world is
seen in the same manner by all organisms. Instead, the claim is that certain
environmental features are re-constructed depending upon a number of relevant
factors, including the task at hand (that is, the goal-oriented action being
performed), the functioning sensorimotor modalities, the vantage point of the
organism, the form of embodiment, etc. The basic idea is that the organism actively
constructs a sensorimotor representation that is based on those environmental
features that are directly relevant to the goal-directed action it is currently
performing. Consequently, environmental space X could be viewed differently by the
same organism depending on the type of task the organism is performing in that
space, primarily because the goal-directed activity determines which environmental
features are relevant to the successful performance of the activity. For instance,
individuals attend to different features when they are preparing to mow a stretch of
grass with a lawn mover, than when they are playing soccer on it later the same day.
This is because the environmental features one must observe to successfully mow the
lawn are different from those that impact playing soccer well.
The embodied cognition theorist might respond that the classicist/cognitivist has
misinterpreted what it means to claim that cognition is a constructive process. By
constructive, Embodied theorists do not mean to imply that there is no objective,
external reality and that everything is subjective. Instead, the point is that a type of
mutual specification occurs between the organism and its environment, so that the
way the world looks and the way in which the organism can interact in the world is
primarily determined by the way the organism is embodied. So, an observer-
independent world can be granted, but embodied cognition theorists claim that an
organism will understand this world in terms of the unique sensorimotor relations it
experiences. These fundamental sensorimotor experiences achieved through acting
in the world are actively constructed to facilitate concept formation. For instance, we
view our bodies as having distinct fronts and backs. Due to the characteristics we
associate with each of these bodily spatial relations, linguist George Lakoff and
philosopher Mark Johnson (1999) argue that we also characterize objects in the
world according to these assignments (that is, go to the front of the house, that is the
back of her shirt, etc.). This process is considered to be constructive because we
project these characteristics onto the world because they reflect the foundational
understanding we have of our own bodies.
Consequently, if we were embodied differently then we would not see the world in
this particular way, but in terms of our new set of defining bodily characteristics.
However, by taking into account the bodies that we do have, our actual projected
spatial assignments can be traced back to sensorimotor experience, which enables
the formation of spatial schemas that are projected onto a scene to facilitate
reasoning without the use of deductive logic. These schemas are constructive because
they do not mirror what exists in the world. Instead, these schemas structure
elements within the world in such a way that the individuals can understand their
environment quickly. Given this, it should not be surprising that one way for an
organism to interpret its environment is in terms of something it already knows well:
its own bodily interactions.
In this case study, a painter (hereafter Mr. I) who completely lost his ability to see
colors after a car accident finds that this loss directly affected the way he experienced
other sensorimotor experiences, such as taste and sound. As a result of his accident,
he was only able to see the world in varying degrees of black, white and gray.
Moreover, Mr. I was not able to imagine colors, dream in colors, or remember what
colors looked like. Since he was no longer viewing the world as colored in any of
these ways, Mr. I reported that the nature of his experience of the world was also
affected dramatically. Reportedly, everything around him “had a distasteful, ‘dirty’
look, the whites glaring, yet discolored and off white, the black cavernous-everything
wrong, unnatural, stained, and impure.” Due to this abrupt change in the way he was
viewing his environment, he stated that he was no longer able to have sex or enjoy
food. Moreover, Mr. I was not able to enjoy music to the degree he had before the
accident since he was no longer able to visually transform musical notes into color
sequences.
After living with this condition for some time, Mr. I remarked that while he was
initially upset about his inability to perceive color, he now no longer misses it. In fact,
he reported that his actions, tastes and behaviors have naturally adjusted over time
to reflect that of a night person. He stated that “I love the night time….I often wonder
about people who work at night. They never see the sunlight. They prefer it….It’s a
different world: there’s a lot of space—you’re not hemmed in by streets, people….It’s
a whole new world. Gradually I am becoming a night person. At one time I felt kindly
toward color, very happy about it….Now I don’t even know it exists—it’s not even a
phantom” (164). Varela et al. concluded that:
This description provides rare insight into how our perceived world, which we usually take
for granted, is constituted through complex and delicate patterns of sensorimotor activity. Our
colored world is brought forth by complex processes of structural coupling. When these
processes are altered, some forms of behavior are no longer possible. One’s behavior changes
as one learns to cope with new conditions and situations. And, as one’s actions change, so too
does one’s sense of the world. If these changes are dramatic enough—as in Mr. I’s loss of
color—then a different perceived world will be enacted (164).
This case is meant to illustrate that if one’s ability to see color is completely removed,
then other sensorimotor experiences are also affected. Varela et al. argue that since
vision is not the only modality affected by Mr. I’s accident, his condition provides
some insight into the way in which “perception and action, sensorium and motorium,
are linked together as successively emergent and mutually selecting patterns” (163).
Although color is but one example of the way in which cognition is constructive, the
above case study might prompt one to ask what is the proper or correct way to view
the world? According to Embodied theorists, the answer is that there is no single
proper or correct way of viewing the world, since being able to correctly see the world
translates into using whatever sensorimotor modalities one has to act successfully in
one’s environment. Moreover, since an organism’s sensorimotor apparatus
determines the way it will experience the world, many embodied theorists argue that
instead of assuming that every organism shares the exact same view of the world
(that is, we all view an objective reality in the same way), it makes more sense to
acknowledge that an organism’s particular view of the world is the direct result of its
functioning sensorimotor experiences. The point is that an organism’s knowledge of
the world is primarily through its experiences within the world and these experiences
are constrained by the types of functioning sensorimotor modalities it has. When one
of these modalities is impaired, then its experience of the world will similarly be
affected on multiple levels, since these modalities influence one another. The case of
the colorblind painter illustrates the cross-modal natures of sensori-motor
experience by showing that the impairment of one modality (color) affected the way
the world was experienced in other modalities (taste, sound, etc.) to the point that
certain previously performed actions suddenly no longer make sense. Therefore, the
type of structural coupling that enables color perception to occur is a paradigm
example of constructive cognition.
The theoretical assumption that at least some forms of cognition are constructive is
supported by a growing number of theorists from a variety of disciplines. Varela et al.
argue that the coupling that occurs between organism and environment results in
constructive cognition. Lakoff and Johnson (1999) argue that cognition is
constructive since it involves projecting schemas (e.g., bodily) and combining these
schemas to create a metaphorical understanding of the world. Glenberg (1997, 1999),
Damasio (1994), and Fauconnier and Turner (2002) are but a few of the cognitive
scientists who maintain that cognition is in some way constructive. Thus, this
theoretical assumption is becoming more widely supported in the embodied
cognition literature.
2. Isolationist analysis – cognition can be understood by 2. Relational analysis-interplay among mind, body, and
focusing primarily on an organism’s internal processes. environment must be studied to understand cognition.