DOCUMENT RESUME

ED 466 734 SE 066 311

AUTHOR Nunez, Rafael E.

TITLE Mathematical Idea Analysis: What Embodied Cognitive Science
Can Say about the Human Nature of Mathematics.
PUB DATE 2000-07-00
NOTE 21p.; In: Proceedings of the Conference of the International
Group for the Psychology of Mathematics Education (PME)
(24th, Hiroshima, Japan, July 23-27, 2000), Volume 1; see ED
452 031.
PUB TYPE Reports Evaluative (142) Speeches/Meeting Papers (150)
EDRS PRICE EDRS Price MF01/PC01 Plus Postage.
DESCRIPTORS *Cognitive Psychology; Cognitive Structures; Elementary
Secondary Education; Knowledge Representation; Learning
Theories; Mathematics Education; Metaphors; *Theory Practice
Relationship

ABSTRACT
This paper gives a brief introduction to a discipline called
the cognitive science of mathematics. The theoretical background of the
arguments is based on embodied cognition and findings in cognitive
linguistics. It discusses Mathematical Idea Analysis, a set of techniques for
studying implicit structures in mathematics. Particular attention is paid to
everyday cognitive mechanisms such as image schemas and conceptual metaphors.
Some implications for mathematics education are discussed. (Contains 36
references.) (DDR)

Reproductions supplied by EDRS are the best that can be made

from the original document.
 MATHEMATICAL IDEA ANALYSIS:
WHAT EMBODIED COGNITIVE SCIENCE CAN SAY ABOUT
THE HUMAN NATURE OF MATHEMATICS

Rafael E. Ntifiez
. University of Freiburg
University of California at Berkeley

PERMISSION TO REPRODUCE AND ABSTRACT:

DISSEMINATE THIS MATERIAL HAS
BEEN GRANTED BY This article gives a brief introduction to a new discipline called the cognitive science
of mathematics (Lakoff & Nunez, 2000), that is, the empirical and multidisciplinary
study of mathematics (itself) as a scientific subject matter. The theoretical background
vo4N,UtFINAP of the arguments is based on embodied cognition, and on relatively recent findings in
cognitive linguistics. The article discusses Mathematical Idea Analysisthe set of
TO THE EDUCATIONAL RESOURCES
INFORMATION CENTER (ERIC) techniques for studying implicit (largely unconscious) conceptual structures in
mathematics. Particular attention is paid to everyday cognitive mechanisms such as
role
image schemas and conceptual metaphors, showing how they play a fundamental
in constituting the very fabric of mathematics. The analyses, illustrated with a
discussion of some issues of set and hyperset theory, show that it is (human) meaning
what makes mathematics what it is: Mathematics is not transcendentally objective,
U.S. DEPARTMENT OF EDUCATION
Office of Educational Research and improvement but it is not arbitrary either (not the result of pure social conventions). Some
EDUCATIONAL RESOURCES INFORMATION
CENTER (ERIC)
implications for mathematics education are suggested.
This document has been reproduced as
received from the person or organization
originating it.
Minor changes have been made to
improve reproduction quality.
Have you ever thought why (I mean, really why) the multiplication of two
negative numbers yields a positive one? Or why the empty class is a subclass of all
Points of view or opinions stated in this classes? And why is it a class at all, if it cannot be a class of anything? And why is it
document do not necessarily represent
official OERI position or policy. unique? For most people, including mathematicians, physicists, engineers, and
computer scientists, the answers to these questions have a strong dogmatic component
(try these questions with your own colleagues!). It is common to encounter answers
such as "well, that's the way it is", or "I don't know exactly why, but 1 know it works
that way", and so on.
Within the culture of those who practice mathematics professionally, the
dogmatic answers to these questions usually follow from definitions, axioms, and
rules, they don't necessarily follow from genuine understanding. In those cases, the
validation of the answer is provided by proof, not necessarily by meaning. This
profound difference between determining that something is true and explaining why it
is true, can be seen in the following historical anecdote.
Benjamin Peirce, one of Harvard's leading mathematicians in the 19th century
(and the father of Charles Sanders Peirce), was once lecturing at Harvard on Euler's
proof that e°' +1= 0. In teaching this famous equation and its proof, he remarked,
"Gentlemen, that is surely true, it is absolutely paradoxical; we cannot
understand it, and we don't know what it means. But we have proved
1-3

2
BEST COPY AVAILABLE
 it, and therefore we know it must be truth." (cited in Maor, 1994, p.
160)
Of course Peirce was not the only mathematician (or mathematics teacher) to
fail to understand what eit1 + 1 = 0 means. Even today, relatively few mathematics
teachers and students understand what the equation actually means. Yet generation
after generation of mathematics teachers and students continue to go
uncomprehendingly through one version or another of Euler's proof, understanding
only the regularity in the manipulations of the symbols, but not the ideas that make it
true. This is hardly an isolated example. Meaningless truth and meaningful sense-
making are fundamental components of many debates involving the nature of
mathematics.
In this plenary address, I want to show that it is meaning (i.e., human
meaningful ideas), what makes mathematics what it is, and that this meaning is not
arbitrary, not the result of pure social conventions. My arguments will be based on
contemporary embodied cognitive science. More specifically, I intend to show the
following:
1. That the nature of mathematics is about human ideas, not just, formal proofs,
axioms, and definitions (proofs, axioms, and definitions constitute only a part of
mathematics, which are also realized through precise sets of ideas).
2. That these ideas are grounded in species-specific everyday cognitive and bodily
mechanisms, therefore making mathematics a human enterprise, not a platonic and
transcendental entity.
3. That because of this grounding, mathematical ideas are not arbitrary, that is, they
are not the product of purely social and cultural conventions (although socio-
historical dimensions play key roles in the formation and development of ideas).
4. That the conceptual (and idea) structure that constitutes mathematics can be
studied empirically, through scientific methods.
5 . That a particular methodology based on embodied cognitive science
Mathematical Idea Analysiscan serve this purpose.
Most of the material I will present here is based on the work I have been developing
for several years in close collaboration with the cognitive linguist George Lakoff in
Berkeley (Lakoff & M.-Inez, 1997, 1998, 2000; Niinez & Lakoff, 1998).

THE CONTEMPORARY STUDY OF IDEAS:

FROM ARMCHAIR PHILOSOPHY TO SCIENTIFIC UNDERSTANDING
Throughout history, many mathematicians have tried to answer the question of
the nature of meaning, truth, and ideas in mathematics. In the last century or so,
various influential mathematicians, such as Dedekind, Cantor, Hilbert, Poincare, and
Weyl, to mention only a few, suggested some answers which share important
elements. They all considered, in one way or another, human intuition as a
fundamental starting point for their philosophical investigations: Intuitions of small
integers, intuitions of collections, intuitions of movement in space, and so on (see

3 1-4
BEST COPY AVAILABLE
 Dedekind, 1888/1976; Dauben on Cantor (1979); Kitcher on Hilbert (1976); Poincare,
1913/1963; Weyl, 1918/1994). They saw these fundamental intuitions of the human
mind as being stable and profound to serve as basis for mathematics.'
These philosophical insights tell us something important. They implicitly say
that the edifice of mathematics is based on aspects of the human mind that lie outside
of mathematics proper (i.e., these intuitions themselves are not theorems, axioms or
definitions). However, beyond the philosophical and historical interest these insights
may have, when seen from the perspective of nowadays' scientific standards, they
present important limitations:
First, those mathematicians were professionally trained to do mathematics, not
necessarily to study ideas and intuitions. And their discipline, mathematics (as
such), does not study ideas or intuitions. Today, the study of ideas (concepts and
intuitions) itself is a scientific subject matter, and it is not anymore just a vague
and elusive philosophical object.
Second, the methodology they used was mainly introspectionthe subjective
investigation of one's own impressions, feelings, and thoughts. Now we know,
form substantial evidence in the scientific study of intuition and cognition, that
there are fundamental aspects of mental activity that are unconscious in nature and
therefore inaccessible to introspection.
The moral here is that pure philosophical inquiry and introspectionalthough very
importantgive, at best, a very limited picture of the conceptual structure that makes
mathematics possible. What is needed, in order to understand the nature and origin of
mathematics and of mathematical meaning, is to study mathematics itself (with its
intuitive grounding, its inferential structure, its symbol systems, etc.) as a scientific
subject matter. What is needed is a cognitive science of mathematics, a science of
mind-based mathematics (Lakoff & Ntifiez, 1997, 2000). From this perspective, the
answers to these issues should be in terms of those mechanisms underlying our
intuitions and ideas. That is, in terms of human cognitive, biological, and cultural
mechanisms, and not in terms of axioms, definitions, formal proofs, and theorems. Let
us see what important findings are helpful in providing those answers.

Embodied Cognitive Science and Recent Empirical Findings

about the Nature of Mind
In recent years, there have been revolutionary advances in cognitive science
the multidisciplinary scientific study of the mind. These advances have an important

But, they didn't think of these intuitions and basic ideas as being "rigorous" enough. This was a
major reason why, later, formalism would explicitly eliminate ideas, and go on to dominate the
foundational debates. Unfortunately, at that time philosophers and mathematicians didn't have the
scientific and theoretical tools we have today to see that human intuitions and ideas are indeed very
precise and rigorous, and that therefore the problems they were facing didn't have to do with lack of
rigor of ideas and intuitions. For details, see Nunez & Lakoff, 1998, and Lakoff & Nunez, 2000).

1-5
BEST COPY AVAILABLE
4
bearing on our understanding of mathematics. Among the most profound of these new
insights are the following:
1. The embodiment of mind. The detailed nature and dynamics of our bodies, our
brains, and our everyday functioning in the world structures human concepts and
human reason. This includes mathematical concepts and mathematical reason.
2. The cognitive unconscious. Most cognitive processes is unconsciousnot
repressed in the Freudian sense, but simply inaccessible to direct conscious
introspection. We cannot through introspection look directly at our conceptual
systems and at our low-level cognitive processes. This includes most mathematical
thought.
3. Metaphorical thought. For the most part, human beings conceptualize abstract
concepts in concrete terms, using precise inferential structure and modes of
reasoning grounded in the sensory motor system. The cognitive mechanism by
which the abstract is comprehended in terms of the concrete is called conceptual
metaphor'. Mathematical thought also makes use of conceptual metaphor, as when
we conceptualize numbers as points on a line, or space as sets of points.
In what follows I intend to give a general overview of how to apply these empirical
findings to the realm of mathematical ideas. That is, while taking mathematics as a
subject matter for cognitive science I will ask how certain domains in mathematics are
created and conceptualized. In doing so, I will show that it is with these recent
advances in cognitive science that a deep and grounded Mathematical Idea Analysis
becomes possible (for details, see Lakoff & Nunez, 2000). Keep in mind that the
major concern then is not just with what is true in mathematics, but with what
mathematical ideas mean, and why mathematical truths are true by virtue of what they
mean.
At this point it is important to mention that when I refer to cognitive science, I
refer to contemporary embodied oriented approaches (see, for instance, Johnson,
1987; Lakoff, 1987; Varela, Thompson, & Rosch, 1991; Nthiez, 1995, 1999),
which
are radically different from orthodox cognitive science. The latter builds on dualist,
functionalist, and objectivist assumptions, while the former has explicitly denied
them, especially, the mind-body split (dualism). For embodied oriented approaches
any theory of mind must take into account the peculiarities of brains, bodies, and the
environment in which they exist. Because of these reasons analyses of the sort I will
be giving below were not even imaginable in the days of orthodox cognitive science
of the disembodied mind, developed in the 1960s and early 1970s. In general, within
the traditional perspective, which under the form of neo-cognitivism (Freeman &
111iiiez, 1999) is still very active today, thought is addressed in terms of the
manipulation of purely abstract symbols and concepts are seen as literal free of all
biological constraints and of discoveries about the brain.
I mention this, because, unfortunately, within the mathematics education
community, for many, cognitive science is synonymous with the orthodox view.

2
As we will see later, this is a technical term.

1-6
Because of the various limitations that this traditional view has manifested over the
years, many researchers in mathematics education concerned with developmental,
social, and cultural factors have rejected cognitive science as a whole, assuming that it
had little to offer (Ntifiez, Edwards, & Matos, 1998). I want to make clear then, that
Mathematical Idea Analysis comes out of embodied oriented approaches to cognitive
science. For a deeper discussion of the differences between orthodox cognitive
science and recent embodied oriented cognitive science, see Nthiez (1997), Lakoff &
Johnson (1999), and Ntitiez & Freeman (1999).

Ordinary Cognition and Mathematical Cognition

Substantial research in neuropsychology, child development, and animal
cognition suggests that all individuals of the species Homo Sapiens are born with a
capacity to distinguish among very small numbers of objects and events (e.g.,
subitizing) and to do the simplest arithmeticthe arithmetic of very small numbers
(for recent reviews on these and related issues, see Dehaene, 1997, and Butterworth,
1999). These findings are important for the understanding of the biological rudiments
of basic arithmetic. However, they tell us very little about the full complexity and
abstraction of mathematics. There is a lot more to mathematics than the arithmetic of
very small numbers. Trigonometry and calculus are very far from "3 minus 1 equals
2". Even realizing that zero is a number and that negative numbers are numbers took
centuries of sophisticated development. Extending numbers to the rationals, the reals,
the imaginaries, and the hyperreals requires an enormous cognitive apparatus that
goes well beyond what babies and animals and a normal adult without instruction can
do. So the question of the nature, origin, and meaning of mathematical ideas remains
open: What are the embodied cognitive capacities that allow one to go from such
innate basic numerical abilities to a deep and rich understanding of, say, college-level
mathematics?
George Lakoff and I have addressed this question, using methodologies from
the growing field of cognitive linguistics and psycholinguistics (more about this
below). According to what we have found to date, it appears that such advanced
mathematical abilities are not independent of the cognitive apparatus used outside of
mathematics. Rather, it appears that the cognitive structure of advanced mathematics
makes use of the kind of conceptual apparatus that is the stuff of ordinary everyday
thought such as image schemas, aspectual schemas, conceptual blends, and conceptual
metaphor3. Indeed, the last one is one of the most important ones, constituting the
very fabric of mathematics. It is present in all subfields of mathematics, as when we
conceptualize functions as sets of points, infinite sums as having a final unique
resultant state, or dynamic continuity as being static preservation of closeness
(Weierstrass' s 6-5 criteria).

3
Because of the scope of this presentation, here I will refer only to image schemas and conceptual
metaphor. I will describe them in the next section.

1-7 BEST COPY AVAILABLE

 Let us now have a look at the theoretical background of Mathematical Idea
Analysis.

MATHEMATICAL IDEA ANALYSIS

Extending the study of the cognitive unconscious to mathematical cognition,

implies analyzing the way in which we implicitly understand mathematics as we do it
or talk about it. A large part of unconscious thought involves implicit rather than
explicit, automatic, immediate understandingmaking sense of things without having
conscious access to the cognitive mechanisms by which we make sense of things.
Ordinary everyday mathematical sense-making is not in the form of conscious proofs
from axioms nor is it always the result of explicit, conscious, goal-oriented
instruction. Most of our everyday mathematical understanding takes place without our
being able to explain exactly what we understood and how we understood it. What
Lakoff and I have done is to study everyday mathematical understanding of this
automatic unconscious sort and to ask the following crucial questions:
How much of mathematical understanding makes use of the same kinds of
conceptual mechanisms that are used in the understanding of ordinary,
nonmathematical domains?
Are the same cognitive mechanisms used to characterize ordinary ideas also used
to characterize mathematical ideas?
If yes, what is the biological or bodily grounding of such mechanisms?
We have found that a great many cognitive mechanisms that are not specifically
mathematical are used to characterize mathematical ideas. These include such
ordinary cognitive mechanisms as those used for basic spatial relations, groupings,
small quantities, motion, distributions of things in space, changes, bodily orientations;
basic manipulations of objects (e.g., rotating and stretching), iterated actions, and so
on.
Thus, for example:
Conceptualizing the technical mathematical concept of a class makes use of the
everyday concept of a collection of objects in a bounded region of space.
Conceptualizing the technical mathematical concept of recursion makes use of the
everyday concept of a repeated action.
Conceptualizing the technical mathematical concept of complex arithmetic makes
use of the everyday concept of rotation.
Conceptualizing derivatives in calculus requires making use of such everyday
concepts as motion, approaching a boundary, and so on.
From a nontechnical perspective, this should be completely obvious. But from the
technical perspective of cognitive science, there is a challenging question one must
ask:
Exactly what everyday concepts and cognitive mechanisms are used in exactly
what ways in the unconscious conceptualization of technical ideas, such that they
provide the precise inferential structure observed in mathematics?

1-8
BEST COPY AVAILABLE
 Mathematical Idea Analysis, depends crucially on the answers to this question. We
have found that mathematical ideas, are grounded in bodily-based mechanisms and
everyday experience. Many mathematical ideas are ways of mathematicizing ordinary
ideas, as when the idea of subtraction mathematizes the ordinary idea of distance, or
as when the idea of a derivative mathematicizes the ordinary idea of instantaneous
change. I will illustrate these findings in more detail with some examples taken from
set theory and hyperset theory. But because of the technicalities involved we must
first go over some basic notions of cognitive linguistics, necessary to understand those
examples.

Some Basic Notions of Cognitive Linguistics and the Embodied Mind

Recent developments in cognitive linguistics have been very fruitful in studying
high-level cognition from an embodiment perspective (e.g., natural language
understanding and conceptual systems). In particular, cognitive semantics (Sweetser,
1990, Talmy, 1999), conceptual integration (Fauconnier, 1997; Fauconnier & Turner,
1998) and conceptual metaphor theory (Lakoff, 1993; Lakoff & Johnson, 1980, 1999;
Gibbs, 1994) have proven to be very powerful. These approaches offer the possibility
of empirically studying the conceptual structure of vast systems of abstract concepts
through the largely unconscious, effortless, everyday linguistic manifestations. They
provide an excellent background for the development of Mathematical Idea Analysis.
Conceptual metaphor
An important finding in cognitive linguistics is that concepts are systematically
organized through vast networks of conceptual mappings, occurring in highly-
coordinated systems and combining in complex ways. For the most part these
conceptual mappings are used unconsciously and effortlessly in everyday
communication. An important kind of mapping is the one mentioned earlier,
conceptual metaphor.
It is important to keep in mind that conceptual metaphors are not mere figures
of speech, and that they are not just pedagogical tools used to illustrate some
educational material. Conceptual metaphors are in fact fundamental cognitive
mechanisms (technically, they are inference-preserving cross-domain mappings)
which project the inferential structure of a source domain onto a target domain,
allowing the use of effortless species-specific body-based inference to structure
abstract inference. For example, humans naturally conceptualize Time (target domain)
primarily in terms of Uni-dimensional Motion (source domain), either the motions of
future times toward an observer (as in "Christmas is approaching") or the motion of
an observer over a time landscape (as in "We're approaching Christmas"). That is,
our everyday concept of Time is inextricably related to the experience of uni-
dimensional motion. There are, of course, many more important details and variations
of this general Time As Motion mapping but their analyses would go beyond the
scope of this presentation. The point here, is that these conceptual metaphors (and
conceptual mappings in general) are irreducible, they are extremely precise (e.g., in
the Time As Motion example, their inferential structure preserves transitive relations),

1-9
they are used extensively, effortlessly, unconsciously, and they are ultimately bodily
grounded (for details, see Lakoff & Johnson, 1999, Chapter 10; Ntliiez, 1999).
Contrary to what some people think, conceptual metaphors (and conceptual
mappings in general) are not mere arbitrary social conventions. They are not arbitrary,
because they are structured by species-specific constrains underlying our everyday
experienceespecially bodily experience. For example, in most cultures Affection is
conceptualized in terms of thermic experience: Warmth (as in "He greeted me
warmly", or as in "send her my warm helloes"). The grounding of this mapping
doesn't depend (only) on social conventions. It emerges from the correlation all
individuals of the species experience, from early ontogenetic development, between
affection and the bodily experience of warmth. It is also important to mention that a
huge amount of the conceptual metaphors we use in everyday communication, such as
Affection Is Warmth, is not learned through explicit goal-oriented educational
intervention.
Research in contemporary conceptual metaphor theory indicates that there is an
extensive conventional system of conceptual metaphors in every human conceptual
system. As I said earlier, unlike traditional studies of metaphor, contemporary
embodied views don't see conceptual metaphors as residing in words, but in thought.
Metaphorical linguistic expressions thus are only surface manifestations of
metaphorical thought. These theoretical claims are based on substantial empirical
evidence from a variety of sources, including among others, psycholinguistic
experiments (Gibbs, 1994), cross-linguistic studies (Yu, 1998), generalizations over
inference patterns ( Lakoff, 1987), generalizations over conventional and novel
language (Lakoff and Turner, 1989), the study of historical semantic change
(Sweetser, 1990), of language acquisition (C. Johnson, 1997), of spontaneous gestures
(McNeill, 1992), and of American sign language (Taub, 1997). Conceptual mappings
thus can be studied empirically, and stated precisely.
In what concerns mathematical concepts, Lakoff & Ntiiiez (2000) distinguish,
three important types of conceptual metaphors:
Grounding metaphors, which ground our understanding of mathematical ideas in
terms of everyday experience. In these cases, the target domain of the metaphor is
mathematical, but the source domain lies outside of mathematics. Examples
include the metaphor Classes Are Container Schemas (see below) and other
conceptual metaphors for arithmetic.
Redefinitional metaphors, which are metaphors that impose a technical
understanding replacing ordinary concepts (such as the conceptual metaphor used
by Georg Cantor to reconceptualize the notions of "more than" and "as many as"
for infinite sets).
Linking metaphors, which are metaphors within mathematics itself that allow us to
conceptualize one mathematical domain in terms of another mathematical domain.
In these cases, both domains of the mapping are mathematical. Examples include
Von Neumann's Numbers Are Sets metaphor, Functions Are Sets of Points, and as
we will see later, the Sets Are Graphs metaphor.

1-10 ;JEST COPY AVAILABLE

The linking metaphors are in many ways the most interesting of these, since they are
part of the fabric of mathematics itself. They occur whenever one branch of
mathematics is used to model another, as happens frequently. Moreover, linking
metaphors are central to the creation, not only of new mathematical concepts, but
often to the creation of new branches of mathematics. Such classical branches of
mathematics as analytic geometry, trigonometry, and complex analysis owe their
existence to linking metaphors.
Spatial relation concepts and image schemas
Another important finding in cognitive linguistics is that conceptual systems
can be ultimately decomposed into primitive spatial relations concepts called image
schemas. Image schemas are basic dynamic topological and orientation structures that
characterize spatial inferences and link language to visual-motor experience (Johnson,
1987; Lakoff and Johnson, 1999). As we will see, an extremely important feature of
image schemas is that their inferential structure is preserved under metaphorical
mappings. Image schemas can be studied empirically through language (and
spontaneous gestures), in particular through the linguistic manifestation of spatial
relations.
Every language has a system of spatial relations, though they differ radically
from language to language. In English there are prepositions like in, on, through,
above, and so on. Other languages have systems that often differ radically from the
English system. However, the spatial relations in a given language decompose into
conceptual primitives (image schemas) that appear to be universal.
For example, the English word "on," in the sense used in "The book is on the
desk" is a composite of three primitive image schemas:
The Above Schema (the book is above the desk)
The Contact Schema (the book is in contact with the desk)
The Support Schema (the book is supported by the desk)
The Above Schema is orientational; it specifies an orientation in space relative to the
gravitational pull one feels on one's body. The Contact Schema is one of a number of
topological schemas; it indicates an absence of a gap. The Support Schema is force-
dynamic in nature; it indicates the direction and nature of a force. In general, static
image schemas fall into one of these categories: orientational, topological, and force-
dynamic. In other languages, the primitives may combine in very different ways. Not
all languages have a single concept like English on. For instance, even in a language
as close as German, the on in on the table is rendered as auf, while the on in on the
wall (which does not contain the Above Schema) is translated as an.
A common image schema that is of great importance in mathematics is the
Container Schema, which in everyday cognition occurs as the central part of the
meaning of words like in and out. The Container Schema has three parts: an Interior, a
Boundary, and an Exterior. This structure forms a gestalt, in the sense that the parts
make no sense without the whole. There is no Interior without a Boundary and an
Exterior, no Exterior without a Boundary and an Interior, and no Boundary without
sides, in this case an Inside and an Outside. This structure is topological in the sense

BEST COPY AVAILABLE

 that the boundary can be made larger, smaller, or distorted and still remain the
boundary of a Container Schema.
The schemas for the concepts In and Out, have a bit more structure than the
plain Container Schema. The concept In requires that the Interior of the Container
Schema be profiled, that is, that it must be highlighted over the Exterior and
Boundary. In addition, there is also a figure-ground distinction. For example, in a
sentence like "The car is in the garage," the garage is the ground, that is, it is the
landmark relative to which the car, the figure, is located. In cognitive linguistics, the
ground in an image schema is called the Landmark, and the figure is called the
Trajector. Thus, the In-Schema has the following structure:
Container Schema, with Interior, Boundary, and Exterior
Profiled: The Interior
Landmark: The Interior
Image schemas have a special cognitive function: they are both perceptual and
conceptual in nature. As such, they provide a bridge between language and reasoning
on the one hand and vision on the other. Image schemas can fit visual perception, as
when we see the milk as being in the glass. They can also be imposed on visual
scenes, as when we see the bees swarming in the garden, where there is no physical
container that the bees are in. Because spatial relations terms in a given language
name complex image schemas, image schemas are the link between language and
spatial perception.
We can now analyze how the inferential structure of image schemas (for
example, the Container Schema) is preserved under metaphorical mappings to
generate more abstract concepts (such as the concept of Boolean class). We shall see
exactly how image schemas provide the inferential structure to the source domain of
the conceptual metaphor, which via the mapping is projected onto the target domain
of the metaphor to generate Boolean-class inferences.
Image schema structure and metaphorical projections
When we draw illustrations of Container Schemas, we find that they look rather
like Venn Diagrams for Boolean classes. This is by no means an accident. The reason
is that classes are normally conceptualized in terms of Container Schemas. For
instance, we think (and speak) of elements as being in or out of a class. Venn
Diagrams are visual instantiations of Container Schemas. The reason that Venn
diagrams work as symbolizations of classes is that classes are usually metaphorically
conceptualized as containers that is, as bounded regions in space.
Container Schemas have a logic that appears to arise from the structure of our
visual and imaging system, adapted for more general use. More specifically,
Container Schemas appear to be realized neurally using such brain mechanisms as
topographic maps of the visual field, center-surround receptive fields, gating circuitry,
and so on (Regier, 1996). The inferential structure of these schemas can be used both
for structuring space and for more abstract reason, and is projected onto our everyday
conceptual system by a particular conceptual metaphor, the Categories (or 'Classes')
Are Containers metaphor. This accounts for part (by no means all!) of our reasoning

BEST COPY AVAILABLE

about conceptual categories. Boolean logic also arises from our capacity to perceive
the world in terms of container schemas and to form mental images using them.
So, how do we normally conceptualize the intuitive premathematical notion of
classes? The answer is in terms of Container Schemas. In other words, we normally
conceptualize a class of entities in terms of a bounded region of space, with members
of the class all inside the bounded region and nonmembers outside of the bounded
region. From a cognitive perspective, intuitive classes are thus metaphorical
conceptual containers, characterized cognitively by a metaphorical mapping a
grounding metaphor the Classes Are Containers Schemas metaphor. The following
is the mapping of such conceptual metaphor.
Classes Are Containers
Source Domain Target Domain
Container Schemas Classes
Interiors Of Container Schemas + Classes
Objects in Interiors --> Class members
Being an Object in an Interior ---> The Membership Relation
An Interior of one Container Schema ---* A subclass in a Larger Class
within a Larger One
The Overlap of the Interiors of Two --> The Intersection of Two Classes
Container Schemas
The Totality of the Interiors of Two + The Union of Two Classes
Container Schemas
The Exterior of a Container Schemas The Complement of a Class
This is our natural, everyday unconscious conceptual metaphor for what a class is. It
is a grounding metaphor. It grounds our concept of a class in our concept of a
bounded region in space, via the conceptual apparatus of the image schema for
containment. This is the way we conceptualize classes in everyday life.
We can now analyze, how conceptual image schemas (in this case, Container
Schemas) are the source of four fundamental inferential laws of logic. The structural
constraints on Container Schemas mentioned earlier (i.e., brain mechanisms such as
topographic maps of the visual field, center-surround receptive fields, gating circuitry,
etc.) give them an inferential structure, which Lakoff and I called "Laws of Container
Schemas" (Lakoff & 1111iiez, 2000). These so-called "laws" are conceptual in nature
and are reflections at the cognitive level of brain structures at the neural level (see
Figure 1). The four inferential laws are Container Schema versions of classical logical
laws: Excluded Middle, Modus Ponens, Hypothetical Syllogism, and Modus Tollens.
Let's see the details.
Inferential Laws of Embodied Container Schemas:
Excluded Middle. Every object X is either in Container Schema A or out of
Container Schema A.
 Modus Ponens: Given two Container Schemas A and B and an object X, if A is in B
and X is in A, then X is in B.
Hypothetical Syllogism: Given three Container Schemas A, B and C, if A is in B
and B is in C, then A is in C.
Modus Tollens: Given two Container Schemas A and B and an object Y, if A is in
B and Y is outside of B, then Y is outside of A.

Figure 1. The "laws" of cognitive Container Schemas. The figure shows one cognitive
Container Schema, A, occurring inside another, B. By inspection, one can see that if X is
in A, then Xis in B, and that if Y is outside of B, then Y is outside of A. We conceptualize
physical containers in terms of cognitive containers. Cognitive Container Schemas are
used not only in perception and imagination but also in conceptualization, as when we
conceptualize bees as swarming in the garden. Container Schemas are the cognitive
structures that allow us to make sense of familiar Venn diagrams.

Now, recall that conceptual metaphors allow the inferential structure of the
source domain to be used to structure the target domain. So, the Classes Are
Containers Metaphor maps the inferential laws given above for embodied Container
Schemas (source domain) onto conceptual classes (target domain). These include both
everyday classes and Boolean classes, which are metaphorical extensions of everyday
classes. The entailment of such conceptual mapping is the following:
Inferential Laws for Classes Mapped from Embodied Container Schemas
Excluded Middle. Every element X is either a member of class A or not a member
of class A.
Modus Ponens: Given two classes A and B and an element X, if A is a subclass B
and X is a member of A, then X is a member of B.
Hypothetical Syllogism: Given three classes A, B, and C, if A is a subclass of B and
B is a subclass of C, then A is a subclass of C.

1-14

BEST COPY AVM BLE

 Modus Tollens: Given two classes A and B and an element Y, if A is a subclass of B
and Y is not a member of B, then Y is not a member of A.
The moral then is that these traditional laws of logic are in fact cognitive entities
and, as such, are grounded in the neural structures that characterize Container
Schemas. In other words, these laws are part of our bodies. Since they do not
transcend our bodies, they are not laws of any transcendent reason. The truths of these
traditional laws of logic are thus not dogmatic. They are true by virtue of what they
mean.
This completes our brief and general overview of some crucial concepts of
cognitive linguistics. Let us now see how this background can be used to apply a
Mathematical Idea Analysis to some specific mathematical domains, Set theory and
Hyperset theory.

ARE HYPERSETS, SETS?

A VIEW FROM MATHEMATICAL IDEA ANALYSIS
Consider the following question in modern mathematics: Are hypersets, sets? If
not, what are they? We will now see, what embodied cognitive science can say about
this. Since hypersets and sets are human (technical, mathematical) ideas we can
provide an answer through Mathematical Idea Analysis. This is what we can say.
Sets
On the formalist view of the axiomatic method, a "set" is any mathematical
structure that "satisfies" the axioms of set theory as written in symbols. The
traditional axioms for set theory (the Zermelo-Fraenkel axioms) are often taught as
being about sets conceptualized as containers. Many writers speak of sets as
"containing" their members, and most students think of them that way. Even the
choice of the word "member" suggests such a reading, as do the Venn diagrams used
to introduce the subject. But if you look carefully through those axioms, you will find
nothing in them that characterizes a container. The terms "set" and "member of are
both taken as undefined primitives. In formal mathematics, that means that they can
be anything that fits the axioms. Here are the classic Zermelo-Fraenkel axioms,
including the axiom of choice, what are commonly called the ZFC axioms.
The axiom of extension: Two sets are equal if and only if they have the same
members. In other words, a set is uniquely determined by its members.
The axiom of specification: Given a set A and a one-place predicate, P(x) that is
either true or false of each member of A, there exists a subset of A whose members
are exactly those members of A for which P(x) is true.
The axiom of pairing: For any two sets, there exists a set that they are both
members of.
The axiom of union: For every collection of sets, there is a set whose members are
exactly the members of the sets of that collection.
The axiom of powers: For each set A, there is a set P(A) whose members are
exactly the subsets of set A.
 The axiom of infinity: There exists a set A such that (1) the empty set is a member
of A, and (ii) if x is a member of A,.then the successor of x is a member of A.
The axiom of choice: Given a disjointed set S whose members are nonempty sets,
there exists a set C which has as its members one and only one element from each
member of S.
You can see that there is absolutely nothing in these axioms that explicitly requires
sets to be containers. What these axioms do, collectively, is to create entities called
"sets," first from elements and then from previously created sets. The axioms do not
say explicitly how sets are to be conceptualized.
The point here is that, within formal mathematics, where all mathematical
concepts are mapped onto set-theoretical structures, the "sets" used in these structures
are not technically conceptualized as the Container Schemas we described above.
They do not have container-schema structure with an interior, boundary, and exterior
at all. Indeed, within formal mathematics, there are no concepts at all, and hence sets
are not conceptualized as anything in particular. They are undefined entities whose
only constraints are that they must "fit" the axioms. For formal logicians and model
theorists, sets are those entities that fit the axioms and are used in the modeling of
other branches of mathematics.
Of course, most of us do conceptualize sets in terms of Container Schemas, and
that is perfectly consistent with the axioms given above. However, when we
conceptualize sets as Container Schemas, a particular entailment follows
automatically: Sets cannot be members of themselves, since containers cannot be
inside themselves. But strictly speaking, this entailment does not follow from the
axioms themselves, but rather from our metaphorical understanding of sets in terms of
containers. The above axioms do not rule out sets that contain themselves. Indeed, an
extra axiom was proposed by Von Neumann to rule out this possibility:
The Axiom of Foundation: There are no infinite descending sequences of sets
under the membership relation. That is, . . . E e . . . e S is ruled out.
Since allowing sets to be members of themselves would result in such a sequence, this
axiom has the indirect effect of ruling out self-membership.
Hypersets
Technically within formal mathematics, model theory has nothing to do with
everyday understanding. Model-theorists do not depend upon our ordinary container-
based concept of a set. Indeed, certain model-theorists have found that our ordinary
grounding metaphor that Classes Are Container Schemas gets in the way of modeling
kinds of phenomena they want to model, especially recursive phenomena. For
example, take expressions like
1
x=1+
1

1+
1+ ...
If we observe carefully, we can see that the denominator of the main fraction has in
fact the value defined for x itself. In other words the above expression is equivalent to

x=t+x
1-16 5
Such recursive expressions are common in mathematics and computer science. The
possibilities for modeling such expressions using "sets" are ruled out if the only kind
of "sets" used in the modeling must be ones that cannot have themselves as members.
Set-theorists have realized that a new non-container metaphor is needed for thinking
about sets, and have explicitly constructed one (see Barwise and Moss, 1991).
The idea is to use graphs, not containers, for characterizing sets. The kinds of
graphs used are Accessible Pointed Graphs, or APGs. "Pointed" indicates an
asymmetric relation between nodes in the graph, indicated visually by an arrow
pointing from one node to anotheror from one node back to that node itself (see
Figure 2). "Accessible" indicates that there is a single node which is linked to all other
nodes in the graph, and can therefore be "accessed" from any other node.

(a-)

Figure 2. Hypersets: Sets conceptualized as graphs, with the empty set as the graph with
no arrows leading from it. The set containing the empty set is a graph whose root has
one arrow leading to the empty set (a). Illustration (b) depicts a graph of a set that is a
"member" of itself, under the Sets Are Graphs Metaphor. Illustration (c) depicts an
infinitely long chain of nodes in an infinite graph, which is equivalent to (b).
From the axiomatic perspective, they have replaced the Axiom of Foundation
with another axiom that implies its negation, the "Anti-Foundation Axiom." From the
perspective of Mathematical Idea Analysis they have implicitly used a conceptual
metaphor, a linking metaphor whose mapping is the following:
The Sets Are Graphs Metaphor
Source Domain Target Domain
Accessible Pointed Graphs Sets
An APG The Membership Structure of a Set
An Arrow The Membership Relation
Nodes That Are Tails of Arrows * Sets
Decorations on Nodes that are Heads of ----> Members
Arrows
APG's With No Loops ---) Classical Sets With The Foundation
Axiom
APG's With or Without Loops --> Hypersets With the Anti-Foundation
Axiom

1
The effect of this metaphor is to eliminate the notion of containment from the concept
of a "set." The graphs have no notion of containment built into them at all. And
containment is not modeled by the graphs.
Graphs that have no loops satisfy the ZFC axioms and the Axiom of
Foundation. They thus work just like sets conceptualized as containers. But graphs
that do have loops model sets that can "have themselves as members." They do not
work like sets that are conceptualized as containers, and they do not satisfy the Axiom
of Foundation.
A "hyperset" is an APG that may or may not contain loops. Hypersets thus do
not fit the Axiom of Foundation, but rather another axiom with the opposite intent:
The Anti-Foundation Axiom: Every APG pictures a unique set.
The fact that hypersets satisfy the Zermelo-Fraenkel axioms confirms what we
said above: The Zennelo-Fraenkel axioms for set theorythe ones generally accepted
in mathematicsdo not define our ordinary concept of a set as a container at all!
That is, the axioms of "set theory" are not, and were never meant to be, about what we
ordinarily call "sets", which we conceptualize in terms of Container Schemas.
So. What are sets, really?
Here we see the power of conceptual metaphor in mathematics. Sets,
conceptualized in everyday terms as containers, do not have the right properties to
model everything needed. So we can now metaphorically reconceptualize "sets" to
exclude containment by using certain kinds of graphs. The only confusing thing is that
this special case of graph theory is still called "set theory" for historical reasons.
Because of this misleading terminology, it is sometimes said that the theory of
hypersets is "a set theory in which sets can contain themselves." From a cognitive
point of view this is completely misleading because it is not a theory of "sets" as we
ordinarily understand them in terms of containment. The reason that these graph
theoretical objects are called "sets" is a functional one: they play the role in modeling
axioms that classical sets with the Axiom of Foundation used to play.
The moral is that mathematics has (at least) two quite inconsistent metaphorical
conceptions of sets, one in terms of Container Schemas (a grounding metaphor) and
one in terms of graphs (a linking metaphor). Is one of these conceptions right and the
other wrong? There is a perspective from which one might think so, a perspective that
says that there must be only one literal correct notion of a "set". But from the
perspective of Mathematical Idea Analysis these two distinct notions of "set" define
different and mutually inconsistent subject matters, conceptualized via radically
different conceptual metaphors. This situation is much more common in mathematics
than the general public normally recognizes. It is Mathematical Idea Analysis that
helps us to see and analyze these situations, by making explicit what is cognitively
implicit.

1-18 17
 EPILOGUE: SOME SPECULATIONS ABOUT THE IMPLICATIONS OF MATHEMATICAL
IDEA ANALYSIS FOR MATHEMATICS EDUCATION
I would like to close my presentation, making some general remarks about
possible implications of Mathematical Idea Analysis for mathematics education in
general, and for the psychology of mathematics education in particular. This is by no
means an exhaustive list. It is simply an open list to be taken as a proposal for
discussion during the various sessions of the PME-2000 meeting.
In a nutshell, I could say that the deepest implication that Mathematical Idea
Analysis provides, is the kind of philosophy of mathematics and of mathematics
education that it brings forth. The approach presented here gives a portrait of
mathematics that is fundamentally human. Concepts and ideas are human, and the
truths that come out of them are relative to human conceptual systems. This includes
mathematics. It follows from this perspective that teaching mathematics implies
teaching human meaning, and teaching why theorems are true by virtue of what the
elements involved actually mean. From this perspective at least the following
implications can be mentioned:
Mathematics education should demystify truth, proof, definitions, and formalisms.
Although they are relevant, they should be taught in the context of the underlying
human ideas. Therefore questions like those in the first paragraph of this article
should be taken very seriously in the educational process.
Mathematics Education should also demystify the belief that meaning, intuition,
and ideas are vague and (purely) subjective. Human ideas and meaning have an
impressive amount of bodily grounded constrains that make them non-arbitrary.
Mathematics should be taught as a human enterprise, with its cultural and
historical dimensions (which shouldn't just be a presentation of dates and a
chronological list of events). These human dimensions should include those
moments of doubts, hesitations, triumphs, and insights that shaped the historical
process of sense-making.
New generations of mathematics teachers, not only should have a good
background in education, history, and philosophy, but they should also have some
knowledge of cognitive science, in particular of the empirical study of conceptual
structures and of everyday unconscious inferential mechanisms. They should know
what is the implicit conceptual structure of the ideas they have to teach.
The so-called "misconceptions" are not really misconceptions. This term as it is
implies that there is a "wrong" conception, wrong relative to some "truth". But
Mathematical Idea Analysis shows that there are no wrong conceptions as such,
but rather variations of ideas and conceptual systems with different inferential
structures (sometimes even inconsistent with each other, as we saw for the case of
sets and hypersets).
From a pedagogical point of view then, it would be very important to identify what
are exactly the variations of inferential structure that generate the so-called
misconceptions.. By making this explicit, a pedagogical intervention should follow

1-19 8 BEST COPY AVAILABLE

 for induCing students to operate with the appropriate conceptual mappings that
bring forth the inferential structure required by the mathematical idea in question.
When applied properly, Mathematical Idea Analysis can serve as a tool for helping
people (especially adolescents and adults) to become aware of the organization,
limitations, and potentials of their own conceptual systems, making explicit (and
conscious) what in everyday life is implicit (and unconscious).
"Being good at mathematics" doesn't necessarily mean being good at doing
calculations and running algorithms. It means knowing how to keep one's
metaphors straight, when to operate with the appropriate metaphors, when to shift
from one to another one, when to combine them, and so on. Teaching how to
master this conceptual gymnastics should be a goal for mathematics education.
Beyond the mathematical content as such, the empirical study of conceptual
systems can also give important insights into the attitudes and beliefs, students
have about mathematics. The detailed study of students' conceptual structures
underlying their linguistic expressions can reveal the origin of difficulties, lack of
motivation, anxieties, and so on, that may be interfering with the learning of
mathematics.
As you can see, this is far from being an exhaustive list. I believe that the cognitive
science of mathematics, and Mathematical Idea Analysis in particular, provide a rich
and deep tool, with a solid theoretical background, for bringing back human meaning
into mathematics. The invitation is then extended for exploring how this can be
accomplished in the process of teaching this astonishing conceptual structure called
mathematics.

REFERENCES
Barwise, J., & L. Moss (1991). Hypersets. The Mathematical Intelligencer, 13 (4): 31-41.
Butterworth, B. (1999). What Counts : How Every Brain is Hardwired for Math. New York: Free
Press.
Dauben, J. W. (1979). Georg Cantor: His Mathematics and Philosophy of the Infinite. Princeton:
Princeton University Press.
Dedekind, R. (1888/1976). Was sind and was sollen die Zahlen. In P. Dugac, Richard Dedekind et
les fondements des mathematiques. Paris: J. Vrin.
Dehaene, S. (1997). The Number Sense: How the Mind Creates Mathematics. New York: Oxford
University Press.
Fauconnier, G. (1997). Mappings in Thought and Language. New York: Cambridge University
Press.
Fauconnier, G., & M. Turner (1998). Conceptual integration networks. Cognitive Science, 22 (2):
133-187.
Freeman, W. J., & R. Nunez (1999). Restoring to cognition the forgotten primacy of action,
intention, and emotion. In R. Nuilez & W. J. Freeman (eds.), Reclaiming Cognition: The
Primacy of Action, Intention, and Emotion. Thorverton, UK: Imprint Academic.
Gibbs, R. (1994). The Poetics of Mind: Figurative Thought, Language, and Understanding. New
York: Cambridge University Press.

1-20
9
 Johnson, C. (1997). Metaphor vs. conflation in the acquisition of polysemy: The case of SEE. In M.
K. Hiraga, C. Sinha, & S. Wilcox, (eds.), Cultural, Typological and Psychological Issues in
Cognitive Linguistics. Current Issues in Linguistic Theory. Amsterdam: John Benjamins.
Johnson, M. (1987). The Body in the Mind: The Bodily Basis of Meaning, Imagination, and Reason.
Chicago: University of Chicago Press.
Kitcher, P. (1976). Hilbert's epistemology. Philosophy of Science, 43: 99-115.
Lakoff, G. (1987). Women, Fire, and Dangerous Things: What Categories Reveal about the Mind.
Chicago: University of Chicago Press.
Lakoff, G. (1993). The contemporary theory of metaphor. In A. Ortony (ed.), Metaphor and
Thought, 2nd ed. New York: Cambridge University Press.
Lakoff, G., & M. Johnson (1980). Metaphors We Live By. Chicago: University of Chicago Press.
Lakoff, G., & M. Johnson (1999). Philosophy in the Flesh. New York: Basic Books.
Lakoff, G., & R. Ntinez (1997). The metaphorical structure of mathematics: Sketching out cognitive
foundations for a mind-based mathematics. In L. English (ed.) Mathematical Reasoning:
Analogies, Metaphors, and Images. Mahwah, NJ: Erlbaum.
Lakoff, G., & R. Ntifiez (1998). Conceptual metaphor in mathematics. In J. P. Koenig (ed.),
Discourse and Cognition: Bridging the Gap. Stanford, CA: CSLI/Cambridge.
Lakoff, G. & R. litinez (2000). Where Mathematics Comes From: How the Embodied Mind Brings
Mathematics into Being. New York: Basic Books.
Lakoff, G., & M. Turner (1989). More than Cool Reason: A Field Guide to Poetic Metaphor.
Chicago: University of Chicago Press.
Maor, E. (1994). e: The Story of a Number. Princeton: Princeton University Press.
McNeill, D. (1992). Hand and Mind: What Gestures Reveal about Thought. Chicago: Chicago
University Press.
Nthiez, R. (1995). What brain for God's-eye? Biological naturalism, ontological objectivism, and
Searle. Journal of Consciousness Studies, 2 (2): 149-166.
Niiez, R. (1997). Eating soup with chopsticks: Dogmas, difficulties, and alternatives in the study of
conscious experience. Journal of Consciousness Studies, 4 (2): 143-166.
1\4-Inez, R. (1999). Could the future taste purple? Reclaiming mind, body, and cognition. In R. Nunez
& W. J. Freeman (eds.), Reclaiming Cognition: The Primacy of Action, Intention, and
Emotion. Thorverton, UK: Imprint Academic.
Naiez, R., L. Edwards, & J. F. Matos (1999). Embodied cognition as grounding for situatedness and
context in mathematics education. Educational Studies in Mathematics, 39 (1-3): 45-65.
Nunez, R., & W. J. Freeman (eds.) (1999). Reclaiming Cognition: The Primacy of Action, Intention,
and Emotion. Thorverton, UK: Imprint Academic.
Nifriez, R., & G. Lakoff (1998). What did Weierstrass really define? The cognitive structure of
natural and 6-8 continuity. Mathematical Cognition, 4 (2): 85-101.
Po i ncare, H. (1913/1963). Mathematics and Science: Last Essays [Dern ieres pen sees]. New York:
Dover.
Regier, T. (1996). The Human Semantic Potential. Cambridge: MIT Press.
Sweetser, E. (1990). From Etymology to Pragmatics: Metaphorical and Cultural Aspects of
Semantic Structure. New York: Cambridge University Press.
Talmy, L. (1999). Toward a Cognitive Linguistics. Cambridge: MIT Press.
Taub, S. (1997). Language in the Body: Iconicity and Metaphor in American Sign Language. PhD

1-21
20
 dissertation, Linguistics Department, University of California at Berkeley.
Varela, F., E. Thompson, & E. Rosch (1991). The Embodied Mind: Cognitive Science and Human
Experience. Cambridge: MIT Press.
Weyl, H. (1918/1994). The Continuum: A Critical Examination of the Foundation of Analysis. New
York: Dover.
Yu, N. (1998). The Contemporary Theory of Metaphor: A Perspective from Chinese. Amsterdam:
John Benjamins.

1-22 21 BEST COPY AVAILABLE

 U.S. Department of Education
Office of Educational Research and Improvement (OEM)
National Library of Education (NLE)
Educational Resources Information Center (ERIC)
ERIC

NOTICE

Reproduction Basis

This document is covered by a signed "Reproduction Release

(Blanket)" form (on file within the ERIC system), encompassing all
or classes of documents from its source organization and, therefore,
does not require a "Specific Document" Release form.

This document is Federally-funded, or carries its own permission to

reproduce, or is otherwise in the public domain and, therefore, may
be reproduced by ERIC without a signed Reproduction Release form
(either "Specific Document" or "Blanket").

EFL' -089 (3/2000)