EMBODIED, SYMBOLIC AND FORMAL ASPECTS OF BASIC

LINEAR ALGEBRA CONCEPTS

Sepideh Stewart & Michael O. J. Thomas
The University of Auckland
Many students find their first experience with linear algebra at university very
challenging. They may cope with the procedural aspects of the subject, solving
linear systems and manipulating matrices, but struggle to understand the crucial
conceptual ideas underpinning them. This makes it very difficult to make progress in
more advanced courses. In this research we have sought to apply APOS theory, in
the context of Tall’s three worlds of mathematics, to the learning of the linear
algebra concepts of linear combination, span, and subspace by a group of second
year university students. The results suggest that the students struggled to
understand the concepts through mainly process conceptions, but embodied, visual
ideas proved valuable for them.
BACKGROUND
The motivation for considering student understanding of linear algebra is well
summed up by Carlson (1993, p. 39),
My students first learn how to solve systems of linear equations, and how to calculate
products of matrices. These are easy for them. But when we get to subspaces, spanning,
and linear independence, my students become confused and disoriented.
Many university teachers will have had a similar experience. Students start well and
cope with the procedural aspects of first courses, solving linear systems and
manipulating matrices, but struggle to understand some of the crucial conceptual
ideas underpinning the material, such as subspace, span, and linear independence,
mentioned by Carlson. The action-process-object-schema (APOS) development in
learning proposed by Dubinsky and others (Dubinsky & McDonald, 2001) suggests
an approach different from the definition-theorem-proof that often characterises
university courses. Instead mathematical concepts are described in terms of a genetic
decomposition into their constituent actions, process and objects in the order these
should be experienced by the learner. For example, there is little point presenting
students with the concept of span if they do not understand linear combination, since
span is an object constructed from the objects of scalar multiple and linear
combination, each of which must be encapsulated from mathematical processes.
Tall and others (Gray & Tall, 1994) have extended these ideas to talk about procepts,
the symbolisation of both a process and an object, so that symbols such as
3v, a1u1 + a2 u2 + ...+ an un etc. may be viewed from either perspective. In more recent
developments of the theory Tall has introduced the idea of three worlds of
mathematics, the embodied, symbolic and formal (Tall, 2004). The worlds describe a
hierarchy of qualitatively different ways of thinking that individuals develop as new

2007. In Woo, J. H., Lew, H. C., Park, K. S. & Seo, D. Y. (Eds.). Proceedings of the 31st Conference of
the International Group for the Psychology of Mathematics Education, Vol. 4, pp. 201-208. Seoul: PME. 4-201
Stewart & Thomas
conceptions are compressed into more thinkable concepts (Tall & Mejia-Ramos,
2006). The embodied world, containing embodied objects (Gray & Tall, 2001), is
where we think about the things around us in the physical world, and it “includes not
only our mental perceptions of real-world objects, but also our internal conceptions
that involve visuo-spatial imagery.” (Tall, 2004, p. 30). The symbolic world is the
world of procepts, where actions, processes and their corresponding objects are
realized and symbolized. The formal world of thinking comprises defined objects
(Tall, Thomas, Davis, Gray, & Simpson, 2000), presented in terms of their properties,
with new properties deduced from objects by formal proof. This theoretical stance
implies that students can benefit from constructing embodied notions underpinning
concepts by performing actions that have physical manifestations, condensing these
to processes and encapsulating these as objects in the embodied world, alongside
working in the symbolic world and, finally, the formal world. Many linear algebra
concepts have embodied and symbolic representations; in fact several
representations (Hillel, 2000). Thus a linear combination of two vectors may be
experienced as a triangle of vector lines, symbolized as au + bv , a(u1,u2 ,u3 ) + b(v1,v 2 ,v 3 ),
or otherwise. In linear algebra few students are given time and opportunities to
develop embodied notions of basic ideas that may be considered trivial by the
teacher. The research presented here used a framework (Figure 1) based on genetic
decompositions of linear combination, span and subspace to investigate student
understanding of these concepts and whether embodied constructs are useful.
Symbolic World
APOS Embodied World Algebra Matrix Formal World
Can calculate with linear
Action Can create a new vector
combinations, e.g.
w by, say addition, e.g.
w = 3u + 5u = 8u

Can determine whether a

Can add multiples of two vector w is a linear
given vectors combination of u and v
using row reduction
Can consider operations on
Process Can think of linear vectors without performing
combinations of vectors them e.g.
e.g. w = 3u + 5u
without having to
Can generalise addition of perform operations
multiples of vectors
Can operate on a linear
Object Can operate on a linear
combination e.g. Sees a general linear
combination e.g.
T(3u + 5v) combination as an element
Sees resultant as new vector M.
object and can operate on it of a vector space V

Figure 1. Part of a framework for linear algebra concepts.

METHOD
This research comprised a case study of a small group of 2nd year 2006
undergraduates at the University of Auckland studying their second general
mathematics course: 40% linear algebra and 60% calculus. The students were offered

4-202 PME31―2007
 Stewart & Thomas
two supplementary linear algebra tutorials at the end of the course, 4 days prior to
the examination, taught by the first-named researcher, and attended by ten students.
Prior to the first they were given linear algebra questions to assess their existing
conceptual thinking (see Figure 2 for questions). The tutorials covered the concepts
of linear combination, span, linear independence, subspace, and basis. The aim was
to give students an explanation of these topics including elements of embodied,
symbolic, and formal worlds. For example, linear combinations were presented by
showing embodied, visual aspects of the addition of scalar multiples of directed line
segments, along with algebraic and matrix symbolisations. This was generalised to
describe the notion of span and the two concepts were linked using a variety of
diagrams. In each case the formal definition was given after the symbolic and visual
aspects were addressed. Following the course examination three of the ten students
returned and did a second, parallel test, although a controlled experiment was not
intended, and two of them, students J and Y, were also interviewed. A post-doctoral
mathematics student did the final test for comparison purposes.
RESULTS
Linear algebra is a large subject and we identified the sequence of concepts: vector,
scalar multiple, linear combination, span, subspace, as the initial focus of attention
for the research. Only the last three of these received attention in the course.

Some formatting has been changed.

Figure 2. A selection from the first and second test questions.

PME31―2007 4-203
Stewart & Thomas

Scalar multiple
Question 1 on the test asked the students, given a vector v, to show how to construct
scalar multiples. On the first test, Figure 3 shows that student F has a wrong
embodied conception of scalar multiple, not appreciating that kv is parallel to v for
all k. However, by the second test this has improved and he displays an action
conception of scalar multiple in the embodied world. While only two of his vectors
are shown here (he also drew 2v) he constructed each of them as a separate entity,
each with their own distinct action. In contrast student Y has combined all three
multiples into the same straight line. This gives evidence of the generalization of the
scalar multiple kv of the vector v, and may be described as the embodiment of the
process of scalar multiplication. The straight line itself can be seen as the
encapsulation of this process into an object-like ‘kv’. Each of the students was able
to use the symbolic world to represent the vectors as 3v, 1/3v, and –3/2v, etc.

Student F (Pre) Student F (Post) Student Y (Post)

Figure 3. Action and process embodied perceptions of scalar multiple.

Linear combination
In question 3 the students were asked to describe in their own words what they
thought a linear combination is. Student J is unsure and describes it (first test) as “A
vector can be present as a relationship between other 1 or more vectors”, and in the
second test the ‘relationship’ is expounded as “One vector is the combination of the
others”. In her interview she also seemed confused, and when asked what a linear
combination is, she said “Ok, linear combination, for example, ah, eh.. is a hmm, it’s
kind of vector equation and I think the linear combination is like that. Is one or two
vectors are independent, they are form a plane or space”. When asked for an example
she gave two vectors (1, 0, 1) and (–1, 0, –1) with the second a multiple, –1, of the
first. Student F also had problems. Unable to answer in the first test he resorts to a
procedural, or action, explanation in the second test, “Linear combination is looking
for whether the last column is composed by another two columns after reduction
raw.” In contrast, Student Y uses the symbolic world for his first test answer, in a
structural, proceptual form xv + yu (Figure 4a), and in the second test gives the only
glimpse of an embodied view, saying that “several linearly independent vectors
combined together form a line, plane”. In his interview he struggled to try and recall
both a symbolic form and a learned definition:
Y: I can’t quite remember the definition, I can just remember those forms something like
b = x1v1+ x2v2… Linear combination is an object class in a space formed by the two
vectors and x, y are scalars, this is my understanding of linear combination.

4-204 PME31―2007
 Stewart & Thomas
The post-doctoral respondent was clearly thinking in the formal world and gives a
‘standard’ kind of definition (Figure 4b).

Figure 4. Symbolic and formal descriptions of linear combination.

Asked to construct a diagram to represent a linear combination in question 4,
students J and Y had no problem, showing the parallelogram and construction lines
(Y’s in Figure 5a). This is at least an action conception in the embodied world, and
suggests a process view. Student F on the other hand has some idea of what to do but
has apparently focussed on a recalled embodied relationship rather than the
construction. In the first test (Figure 5b) he didn’t keep the vectors in the same
direction as those given, but this was corrected in the second test (Figure 5c),
although his vectors don’t form a parallelogram with c, and it seems this information
is not part of his embodied schema for linear combination.

a b c
Figure 5. Symbolic and formal descriptions of linear combination.
Question 11 was designed to see whether the students could link a concept across the
visual and symbolic representations. A process perspective on linear combination in
the symbolic world should enable one to take the generalised symbolic form of linear
combination c1v1+ c2v2 given in this question and reason on its possible embodied
implications. This proved too difficult for student F, who we have already seen is
mostly at an action-process conception of linear combination. Student J also failed to
answer the question, but did try to link the information to the matrix representation
that she presumably felt more comfortable with (Figure 6a). Student Y was able to
link to the embodied, visual representation (see Figure 6b). In the first test he
explained that the ‘space’ formed by v1 and v2 has to contain the z-axis.

a b
Figure 6. Symbolic and embodied perspectives for linear combination.
The three cases in the diagrams show an embodied understanding of what is required
even though precise ideas of span or subspace are not used, and no line or plane is

PME31―2007 4-205
Stewart & Thomas
mentioned. This approach contrasts with that of the post-doc, who used the higher
level concept of span to write that “Non-trivially if Span(v1, v2) contains the z-axis”.
A key identifying characteristic of an object view of a symbolism (procept) is
whether one can operate on the object, as symbolised. In question 13 both J and Y
tackled k(c1v1+c2v2)+m(c3v1+c4v2) using process knowledge in the symbolic world,
multiplying out the brackets and collecting terms (J’s in Figure 7). However, in the
post-test, Y stated that the result was “Still a linear combination of v1 and v2.”, seeing
the structure of the result of his symbolic manipulations.

Figure 7. Process working in the symbolic world for linear combination.

Span and subspace
Student F struggled to describe the concepts, writing nothing in the first test and for
span in the second test he wrote “Span is used to collect vectors that those vectors
have taken E-value.” J similarly did not write anything the first time but in the
second test she displayed an embodied notion of span, saying “the plane that vectors
form.” Y wrote “Span is some vectors form a basis. i.e. {v, u}”, but then wrote
nothing for basis, and this was added to slightly in the post-test with “span: several
linearly independent vectors form the basis.” In the interviews J was unclear, saying
“Span, ..hmm…I think if there are independent vectors formed a plane, it’s uh..
infinite, like if there are 3 or 4 more two vectors form a basis yeah, yeah.” but Y
seemed to have a better understanding of span, not based on a definition, but an
embodied view:
Y: I forgot the definition but my understanding of span is…it’s like little module of that
space, subspace, 3-D like this, and the two vectors form a little plane, and those two
vectors in this plane. This plane is a subspace of R3, …Span is a little module of the
subspace. The general form, something like linear combination.
It was clear that the concept of span was not well understood, and only the post-doc
was able to link span to linear combination, and to generalise, stating that it “…is the
set of all linear combinations of v1…vn”. This shows the power of his formal world
thinking, which was linked to a symbolic form. Only Y and the post-doc attempted
subspace, with Y giving some embodied ideas, “Subspace is something like, if in a
3-d space, a plane, a line, 0, or itself, all can be a subspace”, but did not mention the
need for the lines or planes to pass through the origin.
Question 7 was directed at identification of a subspace in R3, although interpretation
of the third picture proved problematic. Student F wrote nothing in the first test for
this question, but in the second gave the second picture as a subspace since “Those
vectors being contained by same plane.” J described both of the planes as subspaces
in the first test but changed this in the second to say, correctly, that the second was

4-206 PME31―2007
 Stewart & Thomas
not a subspace; no reason was given in the test, but in the interview this was clarified
“And this one before the tutorial I think is a subspace after the tutorial I think is not a
subspace… because it looks like this plane is not like this one [referred to the first
picture] from the origin.” Y chose only the first plane as a subspace in both tests,
saying “Those vectors form a plane in a 3-D space. A plane is a subspace of 3-D
space.”, and “this is a subspace. It is a plane in a 3-D space.” It is not clear whether
this confirms his idea above that any line or plane could be a subspace, since he
rejected the second plane without reasons. When asked about the question in the
interview his reply showed confusion over vectors not from the origin “Subspace is
from the.. It’s formed by those two vectors. And the second graph.. this looks odd,
because they are not from the origin. Only the post-doc correctly identified the
answers, with reasons in the test: “Subspace, since it is a plane containing the origin”
and “Not a subspace – since it does not contain the point (0, 0, 0)”, but it seems that J
had the embodied idea after the tutorials. The purpose of question 10 was to
ascertain if links between concepts were being made. Students F and J wrote nothing
in either test for this, although when pressed in the interview J showed she is moving
towards some embodied understanding “Linear.. it means the two vectors are linearly
independent, if you look at span like a subspace and the set of all linear combinations
lies on the span and also lies on the subspace, or they form”.
Figure 8 shows Y’s answers in both tests. In the first test (top) he still links span to
basis, and has an embodied view of both subspace and the set of linear combinations,
but in the second test he is much more able to link these concepts together, although
not as succinctly as the post-doc who wrote “(a) and (c) are obviously the same since
they both describe the plane generated by v1, v2.” In his interview Y was able to
express the problem he was having with relating the three concepts:
Y: Span and subspace…those two are related together, and if we call subspace W…and
those vectors are the span, because those vectors formed that subspace…Linear
combination, yeah, this question confuses me. How do we distinguish between linear
combination and subspace?

Figure 8. Y’s improving links between span, subspace and linear combination.
He sees the span of three vectors in R3 as always forming a subspace, since position
vectors always go through the origin, but can’t link to linear combination.

PME31―2007 4-207
Stewart & Thomas

CONCLUSIONS
This research confirms the idea that some students struggle with basic linear algebra
concepts such as linear combination, span and subspace. It seems to us, on the basis
of limited data, that the use of embodied notions in the tutorials helped. We asked
the students if this was the case and how they would explain some of the ideas to
others. J was very clear that a visual, embodied approach had greater value for her
than beginning with definitions:
J: Basically when I was in lecture I mixed up. All the relationship between definitions of
subspace linear comb….But when I came to your tutorial there was some graphs and also
very clear explanation that helped me to understand…And if I become a tutor I teach as
your way, first I...graph them, not the definition, I think its too difficult to understand,
makes them confused.
Y agreed “First give them a picture and start from something in the real life, not from
the maths because students are just started studying maths, they couldn’t understand
definitions.” Commenting on appreciation of the tutorials F and J wrote “F: It was
really helpful to understand the basic of concept. When I began reviewing without
tutorials, I could do any question about that part.”, “J: It’s quite helpful. It clarifies
my confusion of theories…such as basis, subspace, span-- which I confused through
the semester.” We continue to construct the framework, based on APOS theory and
the three worlds of thinking, which presents embodied, symbolic and formal
experiences that students could have with linear algebra concepts. Further research is
under way to examine the value of the framework in learning.
REFERENCES
Carlson, D. (1993). Teaching linear algebra: Must the fog always roll in? In D. Carlson, et al.
(Eds.), Resources for teaching linear algebra, (Vol. 42, pp. 39-51). MAA Notes.
Dubinsky, E. & McDonald, M. (2001). APOS: A constructivist theory of learning. In D.
Holton (Ed.) The Teaching and Learning of Mathematics at University Level: An ICMI
Study (pp. 275–282). Dordrecht: Kluwer.
Gray, E. & Tall, D. O. (1994). Duality, ambiguity & flexibility: A proceptual view of simple
arithmetic, Journal for Research in Mathematics Education, 26, 115–141.
Gray, E. & Tall, D. O. (2001). Relationships between embodied objects and symbolic
procepts: An explanatory theory of success and failure in mathematics, In M. van den
Heuvel-Panhuizen (Ed.) Proceedings of the 25th Conference of the International Group
for the Psychology of Mathematics Education, Utrecht, Netherlands, 3, 65–72.
Hillel, J. (2000). Modes of description and the problem of representation in linear algebra,
In J.-L. Dorier (Ed.) The Teaching of Linear Algebra in Question (pp. 191–207),
Dordrecht, Netherlands: Kluwer Academic Publishers.
Tall, D. O. (2004). Building theories: The three worlds of mathematics. For the Learning of
Mathematics, 24(1), 29–32.
Tall, D. O. & Mejia-Ramos, J. P. (2006). The long-term cognitive development of different
types of reasoning and proof, Conference on Explanation and Proof in Mathematics:
Philosophical and Educational Perspectives, Essen, Germany.
Tall, D., Thomas, M. O. J., Davis, G., Gray, E. & Simpson, A. (2000). What is the object of
the encapsulation of a process? Journal of Mathematical Behavior, 18(2), 223–241.

4-208 PME31―2007