Copyright American Psychological Association

An Introduction to the Theory

of Embodied Cognition 1
A
rthur Glenberg, the preeminent cognitive psychologist who has exten-
sively tested and written about the theory of embodied cognition, starts
his course on the topic with questions such as these: “Do left-handers
perceive the world differently than right-handers, such that, for example,
right-handers like objects on the right more than objects on the left, but lefties
like the opposite?”; “When you say about a potential date, ‘He leaves me
cold,’ do you literally feel cold?”; “Does getting a Botox injection to remove
frown lines make it difficult to understand a sentence about sadness?”;
and, “When you are leaning backwards, are you more likely to think about
your past than your future?” These questions clearly implicate the body in
thought, and the surprising answer to all is “yes.” The empirical evidence
that provides this answer supports the theory of embodied cognition.

Philosophical Roots of the Theory

Before describing how a philosophical version of the theory of embodied
cognition evolved into a testable theory for cognitive psychologists, let me tell
you how I start my course on this topic. My approach is a bit different from
the above hook of Glenberg because it emphasizes the theory’s approach to
how we represent all the knowledge we have in our brain. The way I do this

http://dx.doi.org/10.1037/0000136-001
How the Body Shapes Knowledge: Empirical Support for Embodied Cognition, by R. Fincher-Kiefer
Copyright © 2019 by the American Psychological Association. All rights reserved.

3
 Copyright American Psychological Association
4 HOW THE BODY SHAPES KNOWLEDGE

is to have my students play a few rounds of “20 Q”—an electronic version of the game
“Twenty Questions”—with me on the first day of my advanced laboratory course
in cognition. My goal is to give them firsthand knowledge of “symbolic knowledge
representations,” or nonembodied knowledge, and they come to understand this when
playing this handheld game. Here’s why: The electronic version begins just as the
verbal version of the game does—you come up with your animal, vegetable, mineral,
or “other” thought, and then the handheld device begins the electronic questioning
by posing questions such as, “Is it bigger than a microwave?” You respond with the
“yes,” “no,” or “sometimes” button press on the device. This continues for 20 ques-
tions, or fewer if the device decides it has an early guess.
My students are always amused when they see that this toy can in fact “read
their mind” and, more often than not, guess their thought correctly. So I ask them,
“Does this game have a mind?” (They can’t be tricked into responding that it has a
physical brain, but to ask if it has a mind is a different issue.) Most tentatively say
no—the “mind” seems to be quintessential to being human, so it is hard to claim that
an inanimate object has one. But of course some students answer, “Well, sort of . . .”
I ask those pondering this, “Why do you think it has a mind?” They respond, “Because
it’s thinking and processing information and guessing.” The mind is the game’s intel-
lect, and it clearly has one because it is able to guess what someone is thinking. “So
how does it have this intellect?” I ask my students, and that makes them start thinking
about how the computer chip inside this game stores knowledge, how it is structured
and organized, and then I use the word represented.
Once it is acknowledged that the device has to have knowledge represented to
play the game, then we can discuss what the nature of that representation is and what
the processes are that eventually lead to it being able to guess what they are think-
ing (i.e., “20 Q” cognition). After playing the game, students intuit that once the first
answer is processed (e.g., “Yes” it is an animal), that narrows the possibilities for what
the correct guess is going to be and influences what questions should be asked (e.g.,
“Does it swim?”). Each new answer continues to constrain what the guess could be;
by the time 20 questions are asked, the pruning of possibilities in the game’s knowl-
edge base has left very few, if not one, best guess. And students understand that the
literal representation of that network of knowledge in “20 Q” is binary code, as in all
computers.
What my students have described as they work through how “20 Q” makes its
guess is that the game has a hierarchical semantic network of knowledge that uses
the process of spreading activation to get to its response. This is exactly what cogni-
tive psychologists theorized for human knowledge representation in the early days
of the subdiscipline. In fact, the representation of knowledge was perhaps the central
problem for cognitive psychologists when the subdiscipline first became recognized
as a branch of experimental psychology in the late 1950s and early 1960s. Cognitive
psychologists were questioning what philosophers had been debating for centuries—
what is knowledge? Where does it come from? How do we derive meaning from
our world? Some very early philosophy-of-mind theorists, starting with Descartes,
argued for mind–body dualism, such that mental phenomena were completely sepa-
rate from any body effects. However, in response to Cartesian dualism, Kant argued
 Copyright American Psychological Association
An Introduction to the Theory of Embodied Cognition 5

that knowledge results from the interaction of the mind and the external world and,
as an early precursor to the theory of embodied cognition, suggested that the body is
essential or central to human cognition.
More recently, some philosophers responsible for initially defining and devel-
oping the theory of embodied cognition have viewed cognition as serving the needs
of the body as it meets the demands of real-world situations (hence the name). From
this perspective, internal mental representations of the world are not essential, and
there is little need to discuss the actions of “the brain” (see dynamical systems theory
of Thelen & Smith, 1994, and Beer, 2003, as well as Brooks’s 1991a, 1991b, work
in robotics). This stance arose from a general philosophical belief that the concept of
internal mental representations had been overused in explaining cognition, as well
as a disillusionment with cognitive science’s lack of interest in how perception and
cognition were linked to action. This brought about the antirepresentational view
of “radical embodied cognitive science” (Chemero, 2009; Gallagher, 2005; Thelen &
Smith, 1994).
However, other philosophers flipped the relationship to suggest that the body
serves the mind (cognition), such that the external world leads our body to respond in
a way that will inform and guide the mental representations that constitute thought.
For example, Clark’s (1998; 1999; 2008) view is more moderate in terms of consid-
ering the role of internal representations in cognition and yet still suggests that the
interaction of the body and the external world can explain much of what has typically
been considered the work of the brain. Clark, as well as Varela, Thompson, and Rosch
(1991), argued that the body is essential in the production of cognition; in fact, the
body plays a constitutive role in cognitive processing. The position of these philoso-
phers and cognitive scientists has been referred to as “embodied cognitive science,”
and the central issue that distinguishes it from radical embodied cognitive science is
the presence of internal representations (Alsmith & de Vignemont, 2012).
There has been extensive discussion of how these philosophical approaches to
the theory of embodied cognition emerged and how they differ (e.g., Shapiro, 2011,
2014; and M. Wilson, 2002), but it is important to remember that in the philosophi-
cal tradition, empirical work is often cited to support a view, but it is not essential in
developing and testing the theory. Cognitive psychologists became interested in the
theory of embodied cognition because we, too, were interested in the body–mind
interplay. However, within the experimental psychological tradition, the theory’s
tenets required examination and testing. Cognitive psychologists adopted only one
perspective of the theory of embodied cognition, and from a philosophical and cogni-
tive science standpoint, it is considered a “narrow view” of the theory. This is because
the perspective is “brain centered”—it assumes internal representations and, as such,
shifts from the view that the mind operates to serve the body to a view that the body
serves the mind. As M. Wilson (2002) stated,

This takeover by the mind, and the concomitant ability to mentally represent
what is distant in time or space, may have been one of the driving forces
behind the runaway train of human intelligence that separated us from other
hominids. (p. 635)
 Copyright American Psychological Association
6 HOW THE BODY SHAPES KNOWLEDGE

Psychological Roots of the Theory

Getting back to the historical roots of cognitive psychology, the issue of internal rep-
resentations of knowledge was of central concern to the first cognitive psychologists.
We were emerging from the ashes of behaviorism, and the interest in the mind was
being driven by the emerging field of computer science. In 1956, John McCarthy, a
professor of mathematics at Dartmouth College, organized a conference titled Summer
Research Project on Artificial Intelligence, where the intent of the conference was to
“proceed on the basis of the conjecture that every aspect of learning or any other feature
of intelligence can in principle be so precisely described that a machine can be made to
simulate it” (see McCarthy, Minsky, Rochester, & Shannon, 2006, p. 12). Mathema-
ticians, computer scientists, linguists, and psychologists attended this conference, all
united by this curiosity in understanding human thought and cognitive processes to
the degree that a computer could mimic them.
Another important conference in the same year took place at the Massachusetts
Institute of Technology, where the lingering interest in behavior was fading quickly
and being replaced by an interest in the study of mental processes. At this confer-
ence, George Miller presented his idea that there are limits to human’s ability to hold
information in memory, that limit being somewhere around 7 “chunks” (G. A. Miller,
1956). Some psychologists were working on the “processing” aspect of cognition and
were trying to understand perception, attention, memory, and reasoning. Others
were working on the “structure” aspect of cognition—that is, how the knowledge
we have is organized and coherent so that it can be retrieved and used by cogni-
tive processes. Theories of knowledge were developed and tested, and they varied
from feature comparison models (cf., Smith, Shoben, & Rips, 1974) to prototype
models (cf., Rosch, 1973) to semantic network models (cf., Collins & Loftus, 1975;
and J. R. Anderson, 1983). Figure 1.1 shows a semantic network from Collins and
Loftus (1975), where the links between concept nodes are short or long depending
on personal experiences with these thoughts: The shorter the link, the stronger the
connection, and the faster the reaction time to respond to this connection in any
object identification or categorization task (think “20 Q” responses and early guesses
based on predictability in this network of knowledge). Semantic network theories
were widely tested and modified to address results that could not be explained, but
the newer versions became so general and flexible that they were criticized for being
difficult to falsify, thus losing explanatory power.
In the 1980s, a new approach called connectionism gained favor, which was an
approach to create computer models for representing knowledge that was based on
what was known at that time about neural networks. These models were called par-
allel distributed processing (PDP) models of knowledge representation, and in some
sense they were technological updates of more traditional semantic network models.
These PDP models argued that neuron-like nodes are excited or inhibited by the
action of other nodes, and thus knowledge was represented by patterns of excitation
or inhibition rather than particular nodes. Spreading activation still flows through
networks of linked concepts, but processes can occur in parallel, rather than the more
 Copyright American Psychological Association
An Introduction to the Theory of Embodied Cognition 7

FIGURE 1.1

STREET
VEHICLE

CAR
BUS
TRUCK

AMBULANCE FIRE
ENGINE HOUSE

FIRE
ORANGE
RED
YELLOW APPLES

GREEN
PEARS
CHERRIES

VIOLETS ROSES

FLOWERS SUNSETS

SUNRISES CLOUDS

A schematic representation of concept relatedness in a stereotypical

fragment of human memory (where a shorter line represents greater
relatedness). From “A Spreading-Activation Theory of Semantic
Processing,” by A. M. Collins and E. F. Loftus, 1975, Psychological Review,
82, p. 412. Copyright 1975 by the American Psychological Association.

hierarchical earlier models, and are distributed across brain sites (cf., McClelland &
Rogers, 2003).
The zeitgeist of the subdiscipline of cognitive psychology for its first 40 or so
years was to think about cognitive processes such as sensation, perception, atten-
tion, and memory as modular, that is, they existed in and of themselves with minimal
interaction with each other (Fodor & Pylyshyn, 1988; Newell & Simon, 1972). The
representation of knowledge was also functionally autonomous, and there was little
discussion of how processes would affect or determine what was being represented.
This could be likened to having a computer on one’s shoulders with input and output
modules that worked fairly independently from one another. This traditional view
 Copyright American Psychological Association
8 HOW THE BODY SHAPES KNOWLEDGE

was that mental representations were symbolic in nature, quasilinguistic (proposi-

tional), and abstract. Any symbol (or word), for example, chair, was abstract in the
sense that it referred to the general instance of the concept. Further, the symbol
(word) was an arbitrary referent to the concept because the way in which it looked
and sounded bore no relation to the physical or functional properties of the concept.
According to this view, these symbols that represented our knowledge of a
concept were amodal; they were not tied to any specific modality or bodily action.
Further, these symbols were organized propositionally, with their meaning emerg-
ing from relations to other symbols. Thus, this early theory holds that meaning is an
internal process, and cognition is not shaped by perception or action. Newell and
Simon’s (1976) physical symbol system hypothesis (PSSH) was a theoretical instan-
tiation of this view of knowledge: Abstract symbols can be found in both human
thought (propositions) and computer representation (binary code), and thinking is
the manipulation of these symbols, which does not involve perception or action.
As cognitive psychology grew as a discipline, there were many theoretical argu-
ments and empirical results that undermined a PSSH system of knowledge represen-
tation. One of the most damning arguments against a wholly symbol system view of
knowledge is the “symbol-grounding” problem (Harnad, 1990; Searle, 1980). This
problem basically states that abstract symbols (words) that are arbitrary and have
no connection to the external world can have no meaning. If thinking (or computa-
tion in a computer) is symbol manipulation that goes on internally, just within our
heads or in the computer, and there are no external referents, then we have what
Harnad (1990) referred to as the “symbol merry-go-round” problem. This is when
symbols cannot be assigned meaning according to what they refer to (which would
make them nonarbitrary); instead, their meaning is just in relation to other symbols.
Glenberg (2015) likened this to trying to find the meaning of a word in a foreign
language in a dictionary made up of only other foreign language words—the search
would never be successful if none of the symbols have any meaning to the person
searching. He described this problem in the following way:

Imagine that you land at an airport, perhaps in China, where you don’t speak
the local language. You see what appears to be a sign consisting of logograms
(or any other noniconic marks). Although you don’t speak the language, you
do have a dictionary written in that language. You look up the first logogram
and find its definition, but of course the definition consists of more logograms
whose meanings are obscure to you. Undaunted, you look up the definition of
the first logogram in the definition, and you find that its definition consists of
even more uninterpretable logograms. The point is that no matter how many
of the logograms you look up, this closed system of abstract symbols will never
produce any meaning for you. (p. 165)

What Glenberg (2015) pointed out about the standard cognitive models of knowl-
edge representation is that in those models, understanding the meaning of a symbol
could be likened to going round and round in an infinite circle. Looking for the mean-
ing of one symbol involves trying to find the meaning of another arbitrary symbol,
and the meaning of that symbol requires the same kind of internal search for another
 Copyright American Psychological Association
An Introduction to the Theory of Embodied Cognition 9

arbitrary symbol. If you look back at Figure 1.1, it seems as if the symbols (the ellipses
that have the labels underneath) are meaningful—but that is because we are provid-
ing them with meaning. The shape is not related to the meaning, and the label (the
word) is something we have provided. The words are there for the convenience of the
reader; if we opened up someone’s head and looked inside, we would not see labels
on ellipses. Figure 1.2 more accurately represents the semantic network theory. Here
the abstract symbols (ellipses) are not labeled, and thus they are authentically amodal,
abstract symbols. This portrayal of the theory makes clear this symbol merry-go-round
argument. One can start at any node and trace relations to any and all other nodes, but
no matter how many relations are traced, no meaning will ever be available.

FIGURE 1.2

A schematic representation of the semantic network theory with amodal

abstract symbols. Adapted from “A Spreading-Activation Theory
of Semantic Processing,” by A. M. Collins and E. F. Loftus, 1975,
Psychological Review, 82, p. 412. Copyright 1975 by the American
Psychological Association.
 Copyright American Psychological Association
10 HOW THE BODY SHAPES KNOWLEDGE

Never leaving “the head” (or the computer) to determine what a symbol is refer-
ring to in the external world would mean an infinite regress of search for mean-
ing. We can only break from this symbol-grounding problem if the symbols refer to
something in our external world, that is, have referents that ground those symbols in
perception, action, and/or emotion. This grounding represents the meaning of the
symbol. The embodied approach to cognition argues that all knowledge is grounded
in sensory, perceptual, and motoric processes and that these processes are a function
of one’s own morphology (shape and size) and physiology (internal processes).
Evidence from cognitive neuroscience supports embodiment in revealing that
thinking involves the reactivation and reuse of processes and representations involved
in perception and action. For example, Hauk, Johnsrude, and Pulvermüller (2004)
recorded neural activity while readers listened to action verbs such as kick and pick.
They found that as readers read those action verbs, there was greater activation in the
specific part of the motor cortex responsible for producing that action than other motor
cortex areas. This suggests that understanding an action concept may be grounded in
action-specific motor neurons.
The discovery of mirror neurons in the motor cortex of monkeys may help
us understand a similar neural system in humans that would allow us to compre-
hend others’ actions, goals, and intentions (e.g., Gallese, Keysers, & Rizzolatti, 2004;
Iacoboni, 2009; Rizzolatti & Craighero, 2004). In monkeys, single-cell recordings from
mirror neurons show that a neuron in this tract is similarly active when a monkey is
performing a task as when the monkey is observing another monkey engage in the
same task. Researchers have suggested that the purpose of having mirror neurons that
are active both during action and during observation is that this provides the animal
with the ability to infer goals and emotions to another. If an animal acts in a purpose-
ful way, the neurons that fire during that action, and then are later “reused” while
observing another’s action, must reflect the meaning of that action. This is evidence
for that neuron serving both action and cognition, supporting the view that the body
plays a constitutive role in cognition.
This evidence seriously challenges the view that conceptual representations are
abstract and amodal. Of course there are quite disparate views of the extent to which
mental representations are modal or both amodal and modal. And even those that
argue that meaning is found in modality-specific representations have quite different
approaches to the theory of embodiment (see Gentsch, Weber, Synofzik, Vosgerau, &
Schütz-Bosbach, 2016; A. D. Wilson & Golonka, 2013).

Tripartite Framework of the Theory

In this book, I will take a three-pronged approach to organizing and synthesizing the
extensive experimental results concerning this theory. This is not entirely original, as
others have noticed this tripartite distinction in the embodiment literature. Although
the experimental work can be loosely organized within this tripartite framework,
I will continue to remind the reader that these approaches are not contradictory but
instead complementary.
 Copyright American Psychological Association
An Introduction to the Theory of Embodied Cognition 11

These three approaches are represented by

❚ Arthur Glenberg’s (2010, 2015) view that the body is essential to knowledge
(Chapters 2, 3, 4, and 5);
❚ Lawrence Barsalou’s (1999) perceptual symbol system hypothesis (PSS), which
relies on representations that come about from perceptual experiences and sen-
sorimotor simulation as the “core computational” process of thought (Chapters 6
and 7); and
❚ Lakoff and Johnson’s (1980) Metaphors We Live By proposal that language reflects
our representation of knowledge (Chapter 8).
Although this book will allow the reader to compare and contrast these three
approaches to embodiment, there is no attempt to choose one as the prominent or more
persuasive perspective. Because it is at the core of my presentation of this topic, I will
say it again: These approaches are not contradictory, but instead are complementary.
This book organizes results from different laboratories of experimental psychol-
ogy examining multiple types of cognitive processes to give weight to the theory of
embodiment. Why would I use such a metaphor concerning the weight of a topic?
Very intentionally—it provides a way to illustrate what I am trying to do in this book.
For instance, if I want you to understand the role of the body in this theory of embodi-
ment, I would emphasize that when we hold something heavy, the perceived value
of that which we are holding is increased compared with when it is light (Jostmann,
Lakens, & Schubert, 2009). Further, our judgments concerning the importance of
an object affect how heavy we believe that object to be (Schneider, Parzuchowski,
Wojciszke, Schwarz, & Koole, 2015). These empirical examples suggest that our body
is essential to our conceptualization of importance.
However, if I want you to understand that thinking involves Barsalou’s (1999)
PSS hypothesis of sensorimotor simulation, I would argue that it is the simulation of
a prior experience of weight that underlies a judgment of value. In our life’s physical
experiences, more likely than not, heavier objects have been more important than
lighter objects (e.g., full bottles of milk compared with empty ones), and these patterns
of experience are reenacted in thinking about value (Barsalou, 2008b).
Or, if I wanted you to understand how we give meaning to something abstract,
like importance, I would emphasize that we do this by linking abstract concepts
to superficially dissimilar but understood concrete concepts. This is reflected in our
language. For example, our metaphors concerning “weighty decisions” and “heavy
topics” reflect that our mental representation of value or importance is linked to the
concrete property of weight. Abstract concepts can be understood in terms of mul-
tiple linkages with concrete concepts, yielding expressions that could not be easily
explained by embodied simulations (Landau, Meier, & Keefer, 2010).
In Chapters 2 through 5 of this book, I present material that supports the role
of the body in different cognitive processes—perception of distance, perception of
size—but also how our bodies affect our sense of control, power, even judgments
about free will and include a discussion of how culture may mediate some of these
judgments. Chapters 6 and 7 address how empirical results can demonstrate that
 Copyright American Psychological Association
12 HOW THE BODY SHAPES KNOWLEDGE

sensorimotor simulation is the mechanism that provides meaning. Evidence will

include such research as that demonstrating that emotional facial responses are sim-
ulations of others’ emotional reactions even in the absence of motoric mimicry. Or
how simulation, in the form of how we imagine our bodies moving in space, may
determine our conceptual representation of time more so than the actual movement
of our bodies.
Finally, in Chapter 8, I explore how language, specifically metaphor, reflects the
grounding of meaning. In this chapter, I present results that demonstrate the bidirec-
tionality of metaphor-consistent psychological effects. For example, in exploring the
common metaphor that someone we care for is “warm,” it has been shown that if we
are physically warm (e.g., holding a hot cup of coffee), we feel closer to people in our
lives (even strangers in our physical space) than when holding something cold, and
when we are thinking about a time in which we were included in a social experience,
we judge the temperature of a room to be physically warmer than when we remember
a time of social exclusion. It appears that the conceptualization of affiliation may be
grounded in the physical experience of temperature (Williams & Bargh, 2008).
I conclude with Chapter 9 that addresses the current status of embodiment and
attempts to simplify the complex reactions to the theory. In the end, it will be left to
experimental cognitive psychologists, cognitive neuroscientists, and cognitive scien-
tists to continue with the empirical investigations and modeling of cognitive processes
to determine if the theory of embodiment will in fact be the paradigm shift that uni-
fies psychology as a discipline.
One final note about the book: At the end of each of these content Chapters 2–8,
I have left a “takeaway” for the reader. This is simply a statement or two that represents
the “big picture” of the chapter; the intent was to capture the essence of the empiri-
cal work discussed in that chapter. You will find some redundancy in these takeaways
because they all are describing how embodiment theory explains cognition—and yet
each explanation is within a specific area (e.g., perception, emotion, language) or spe-
cific to a concept (e.g., simulation). The takeaway pays homage to the primary theo-
rists in each of the areas by simply stating their names, but of course the reader will
understand that many others are cited in the chapter to support these views. I hope
these takeaways serve as reminders of the overarching principles guiding the empirical
investigations of the theory of embodied cognition.