Beruflich Dokumente
Kultur Dokumente
https://doi.org/10.1007/s11858-019-01083-3
ORIGINAL ARTICLE
Abstract
Solving systems of linear equations is of central importance in linear algebra and many related applications, yet there is
limited literature examining the symbolizing processes students use as they work to solve systems of linear equations. In this
paper, we examine this issue by analyzing final exam data from 68 students in an introductory undergraduate linear algebra
course at a large public research university in the United States. Based on our analysis, we expanded our framework (Larson
& Zandieh, 2013) for interpretations of matrix equations to include augmented matrices and symbolic forms commonly used
in solving linear systems. We document considerable variation in students’ symbolization processes, which broadly occurred
along two primary trajectories: systems trajectories and row reduction trajectories. Row reduction trajectories included at
least five symbolic shifts, two of which students executed with a great deal of success and uniformity. Students’ symbol-
izing processes varied more in relation to the other three shifts, and these variations were often linked to trends of variable
renaming, variable creation, or imagined parameter reasoning. Students were more flexible in their solution strategies when
solving systems involving lines than for systems involving planes.
13
Vol.:(0123456789)
M. Zandieh, C. Andrews‑Larson
entailed in symbolizing the steps in a solution trajectory to complete a row reduction process. In such cases, as in the
as well as in symbolizing and interpreting a solution to the data for this paper, student symbolization then focuses on
system. In cases where there are infinitely many solutions, moving between systems of equations, augmented matrices
characterizing the solution set can be particularly chal- and even vector equations as part of solving systems and
lenging and involves creating and interpreting a variety of interpreting those solutions.
symbolizations.
13
Symbolizing while solving linear systems
nuances involved in student symbolizing when solving and we determined that the framework of Larson and Zandieh
reasoning about systems of linear equations. (2013) would be useful for describing and documenting stu-
dents’ transitions between a range of inscriptions in solving
systems. In addition, as we analyzed the data and wrote the
3 Conceptual framing paper, we realized that additional aspects would be needed to
more fully describe the nuances in symbolizing that students
For this paper we are interested in student participation were engaged in. In what follows we describe the original
in the mathematical practice of symbolizing. Our fram- Larson and Zandieh (2013) conceptual framework and our
ing is based in Freudenthal’s (1991) notion of mathemat- extensions of it for this paper.
ics as a human activity. Individual student participation in Larson and Zandieh (2013) developed a framework
mathematical activity is one of the four ways of describing for making sense of student thinking by identifying three
mathematical progress posited by Rasmussen, Wawro, and important interpretations of the matrix equation Ax = b
Zandieh (2015). Although there are many aspects of student where A is an m × n matrix, x is in R n and b is in R m.
participation that may be documented, for this paper, we Namely, Ax = b can be interpreted as a system of equa-
follow Rasmussen, Zandieh, King, & Tempo’s (2005) focus tions, a linear combination of column vectors, or as a
on the “notion of advancing mathematical activity as acts transformation from Rn to Rm (see Fig. 1). In this paper,
of participation in different mathematical practices” (p. 53). we expand this framework to also include the interpre-
These disciplinary practices of mathematics include proving, tation of augmented matrices. Figure 1 emphasizes that
defining, algorithmatizing (creating algorithms) and symbol- we see algebraic and geometric as two contexts within
izing (Rasmussen et al., 2005; Zandieh & Rasmussen, 2010; which a person may interact with each interpretation. As
Zandieh, Wawro, & Rasmussen, 2017). in Zandieh (2000) we acknowledge that there may be addi-
We take symbolizing broadly to mean how students tional contexts such as a physical setting or an applica-
reason with any inscriptions including analytic as well as tion context, but we focus on these two primary contexts
graphical inscriptions. In approaching the data for this paper, for the purpose of this paper. In Fig. 1 each row is an
13
M. Zandieh, C. Andrews‑Larson
interpretation and each of the eight boxes represents an Figure 2 illustrates each of these algebraic interpretation-
interpretation–context pair, with each pair being illus- form pairs for one of the systems students solved in our data:
trated for the case where x and b are vectors in R 2 and three planes intersecting in a line. One thing to note is that
there is one unique solution. the augmented matrix equivalent form (AM-E) is what is
There are two things of particular interest in terms of often called in textbooks (e.g., Lay, 2012) the reduced row
the Augmented Matrix interpretation. First is that the literal echelon form (RREF). In the case of infinitely many solu-
symbol x disappears completely from the algebraic context tions (as in Fig. 2), there is no augmented matrix solution
[A|b]. This condensing of expression is helpful for efficiency form (AM-S) that has a direct correspondence to the other
but has the potential for loss of information in communica- solution forms listed in the final column of Fig. 2. On the
tion (Zandieh & Knapp, 2006, 2018). As such, we conjecture other hand, in the case of a system with a unique solution,
that transitions from the Augmented Matrix interpretation the reduced row echelon form would correspond to other
into other interpretations are more difficult because the sym- solution forms in the sense that the values for each of the
bols of the vector x have to be reconstructed. Second, to variables in the solution would be explicitly listed.
our knowledge there is no generally accepted geometrical Given the lack of research on students’ symbolizing while
representation associated with augmented matrices. In other solving linear systems, as well as the complexities and sub-
words, there is not a geometric picture that can be drawn for tleties involved, our research questions are: How do students
the augmented matrix interpretation that is different from engage in symbolizing when solving consistent systems of
the geometric contexts drawn for the other interpretations. equations involving lines and planes? More specifically,
In addition, although a person might transition directly from what interpretations, contexts, and forms do students draw
an augmented matrix into any of the geometric contexts, on as they proceed through their solution process?
students in our data always transitioned into a different We explore examples that have a single solution as well as
algebraic interpretation first before relating to the problem’s examples that have infinitely many solutions. In the Sect. 5,
geometry. we illustrate the trajectories that students used in transition-
In analyzing our data, we realized that the interpreta- ing through various interpretations, contexts, and forms. In
tion–context pairs in Fig. 1 were important but did not cap- doing so we shed light on the nuances involved in solving
ture all aspects of student symbolizing when solving systems these systems of equations. In Sect. 6 we reflect on the simi-
of equations. Sometimes students transitioned between two larities and differences in student symbolizing trajectories
different interpretations or between two different contexts, when solving four different consistent systems.
whereas in other cases students transitioned between sym-
bolizations that are within a particular interpretation–context
pair. Therefore, Fig. 2 lists three different algebraic forms 4 Data sources and methods of analysis
within each of the algebraic interpretations occurring most
often in our data. Original (O) forms have the same set of To answer our research questions, we draw on data taken from
constraints as the given problem. Equivalent (E) forms have the final exams of 68 students enrolled in two sections of an
the same solution set as the original form, but a different set introductory linear algebra classes at a large public research
of constraints. Solution (S) forms have the same constraints university in the United States. Both sections were taught by
as the Equivalent forms but are formatted so that the values the same instructor who was an experienced linear algebra
of the unknowns can be read off directly. instructor at the institution. Most students in the class were
engineering majors in their junior or senior year of college,
13
Symbolizing while solving linear systems
and had completed three semesters of calculus including dif- two on each of two versions of the course’s written final
ferential, integral, and multi-variable calculus prior to enroll- examination. One version of the exam asked about equations
ing in this linear algebra course. About half of the students in of lines (Version L) and the other asked about equations of
the classes were non-traditional students (e.g. not continuously planes (Version P). The questions are shown in Fig. 3. Both
enrolled in post-secondary education since graduating from versions include a system with a unique solution (part a on
high school), about twenty percent of the students spoke a each) and a system with infinitely many solutions that cor-
native language other than English, and 8 of the 68 students respond to a line of intersection (part b on each). There are
were women. more equations than unknowns in all parts except Version
Course topics included systems of linear equations, span P part b, in which there are the same number of equations
and linear independence, linear transformations, determi- and unknowns. Note that 31 students took Version P of the
nants, eigenvectors, eigenvalues, and diagonalization. The exam, and 37 students took Version L.
course textbook was Lay (2012). The primary deviation Our analysis took place in three main phases: we first
from the course textbook sequence of topics was that span broadly categorized interpretations used in students’ solving
and linear independence were introduced prior to methods processes, then coded the chain of inscriptions in individual
for solving systems of linear equations. Two 75-min class students’ solving trajectories, and finally aggregated solution
sessions were spent on explicit instruction of row reduction trajectories across all students taking each version of the
methods, and these were regarded as part of standard instruc- exam. During each phase of analysis, both authors of this
tion (not the focus of any particular instructional interven- paper analyzed the data to come to consensus for each stu-
tion). Instruction on solving systems included conventions dent on how their response should be categorized or coded.
for converting between systems of equations and augmented In order to get a sense of students’ broad use of interpreta-
matrices, row operations, their origin in manipulating sys- tions, in the first phase of analysis, we categorized students’
tems of equations, reduced row echelon form, representation responses according to their primary solution strategy, which
of solutions in general and parametric vector equation forms, fell into two broad categories. If a student rewrote the system
and interpretations of solutions in real-world and geometric as an augmented matrix and fully row reduced, we catego-
contexts (e.g. intersections of lines, planes, etc.). rized the solution strategy as a reduced row echelon form or
In this report, we examine student responses to four ques- RREF strategy. If a student solved the system using meth-
tions that offer insight into students’ symbolizing processes, ods of elimination or substitution or otherwise analyzed the
Fig. 3 Exam questions to be
analyzed
13
M. Zandieh, C. Andrews‑Larson
equations without row reduction, we categorized the solution made inferences about the order of inscriptions based on
strategy as a Systems strategy. a combination of physical organization (e.g. left to right,
In the second phase of analysis, we documented the tra- top to bottom), students’ use of arrows and annotations, and
jectory of each student’s solution process. To do this, we our own understandings of which inscriptions were com-
developed codes for the variety of inscriptions that students mon manipulations of others when solving systems. We also
used as they worked toward a solution. Code Names and documented any written explanations students made, which
descriptions are detailed in Fig. 4. The Code Names in the were particularly helpful in seeing how they made sense
first column of Fig. 4 are accurate to our analysis process of the solutions to the systems (and at times in helping us
and align with common ways of talking about these forms in understand their reasoning during the solving process).
mathematical language. However, as our analysis of student To illustrate our process, Fig. 5 shows the work of student
symbolizing evolved to include the forms in Fig. 2, we began A14 and a portion of A14’s row of the spreadsheet with
using the descriptions corresponding to the interpretation- Code Names. The student first correctly rewrote the system
form abbreviations listed in the third column of Fig. 4. Note as an augmented matrix (AM), then correctly rewrote it in
that P and V do not have a designation as an interpretation- row reduced form (RREF), then correctly rewrote it as a
form from Fig. 2 because they do not refer to an equation. system of equations in which x was written in terms of z
After the Code Names had been developed, they were (SE’’), then correctly rewrote this as a vector equation (VE).
recorded in a spreadsheet (1 row per student) with anno- Note that the third code name is SE’’ because it is not a
tations added to note any idiosyncratic details related to direct rewriting of the RREF as a system but also involved
a particular inscription. The cells in the row were color some manipulation to write x = 2 − z. As illustrated with
coded according to whether each inscription was correct this student, the Code Names were our first window into the
or included an error (neon green for correct, yellow for interpretations and forms used by students and thus into the
incorrect). For each student row, the column entries were complexities of the symbolizing process.
populated chronologically with the code names as they In the third and final phase of analysis, we sorted the
appeared in each student’s solution process. To do this, we spreadsheet from the second phase of analysis by the first
column of code names, then the second column, and so on
Code Description Interpretation -
Name Form so that students with similar initial codes were grouped
AM
RREF
Student writes the system as an augmented matrix
Student writes the system in reduced row echelon
AM-O
AM-E
together. The resulting visualization helped us decide to
form (or row echelon form) aggregate the trajectories of the students into a new spread-
SE’ Student writes the RREF form of the matrix as a SE-E or SE-S
system of equations sheet in which each row corresponded to a single aspect
SE’’ Student writes the system of equations in another SE-E or SE-S of the trajectory (identified by Code Name) and each col-
form
VE Student writes the system as a vector equation LC-S umn corresponds to a student (e.g., Figure 6). Due to minor
P Student writes solution as a point none
V Student writes solution as a vector none
variation in individual student trajectories, we collapsed
(or expanded) some cells to align individual trajectories
Fig. 4 Code names and descriptions with common milestones observed as students progressed
toward a solution. For example, student A14 in Fig. 5 did
13
Symbolizing while solving linear systems
13
M. Zandieh, C. Andrews‑Larson
13
Symbolizing while solving linear systems
Fig. 12 Correct RREF rewritten as incorrect system due to variable Fig. 13 Correct solution with imagined parameter interpretation (stu-
creation and renaming dent A33, planes part a)
13
M. Zandieh, C. Andrews‑Larson
In the lines case, student trajectories varied more widely. We exemplify the group of students who used an RREF
Some students symbolized using variations on the stand- strategy for the intersection of lines questions by drawing on
ard RREF trajectory as in the planes case. Other students the work of student B11. In part a, (see Fig. 16a) the student
leveraged multiple various symbolizations of the original solved the system using a standard RREF trajectory by sym-
system of equations instead of using an augmented matrix bolizing an augmented matrix (AM/AM-O), the row reduced
(i.e., they used a Systems trajectory). In addition, many stu- form of the augmented matrix (RREF/AM-E), the solution
dents employed the geometric context to reason about their as a system of equations (SE′/SE-S), and the solution as a
solutions in ways that were notably different from the planes point (P). In part b (see Fig. 16b) the student completed a
case. Figure 15 presents an Aggregated Trajectory diagram similar RREF trajectory symbolizing an augmented matrix
of student responses across both parts of the Lines question. (AM/AM-O), the row reduced form of the augmented matrix
For example, the leftmost column represents the response of (RREF/AM-E), and the solution as a system of equations
student B11 for task La (top portion) and task Lb (bottom (SE′/SE-E). Of interest is that B11 rewrote the solution as a
portion). different system (SE″) and explained that the intersection is
Fig. 15 Aggregated trajectories for student solutions on lines task parts a and b (La, Lb)
13
Symbolizing while solving linear systems
this line. Note that the particular rewrite of SE″ in this case columns of zeros when interpreting their solution. However,
expressed the solution in the form y = mx + b, from which some students struggled to use this method. Figure 17a, b
students can directly read off the slope and y-intercept. illustrate these struggles.
This SE’’ is a form of SE-S that is different from a typical Figure 17a shows the work of student B13, who appeared
linear algebra solution form as explicated in Fig. 2 and in to want to row reduce the matrix but wrote “not inverti-
the RREF strategies used by students in the Planes task. We ble” in the brackets where the RREF would have appeared.
discuss below that this form was common in the solutions Although it is true that the matrix is not invertible, this
of students who used a Systems strategy for solving Version student concluded incorrectly that the system has no inter-
L. In addition, we note that student B11 did not indicate that section and never stated a row reduced form. In Fig. 17b,
there is a free variable. Of the 9 students who started and student B2 attempted the workaround method by adding a
ended part b with a correct RREF solution, 5 of them did not column of zeroes to row reduce. However, B2 then inter-
explicitly indicate that there was a free variable. In this way preted the row reduction as if the 4th column of zeroes was
these students (like B11 in Fig. 16b) may have switched at what each equation is ‘equal to’ with the third column being
the end from a more formal linear algebra interpretation of treated as a third variable column. This contributed to B2’s
the solution to one based in their familiarity with equations engagement in variable creation and renaming.
of lines. When the students in this group moved on to Lb, there
was some divergence in how students proceeded. Of the 8
5.2.2 Incorrect RREF strategies for both La and Lb (8 of 37 students, 3 students wrote down an incorrect augmented
students) matrix (light green boxes in first row of Lb in Fig. 15) and 1
student (B29) stopped after writing the correct augmented
The work of students who exclusively used an RREF strat- matrix. The remaining 4 students all changed the names of
egy but were not able to complete this trajectory success- the variables but were split in how they reasoned.
fully had similarities with the work of students with incor- Figure 18 follows the work of students B13 and B2 to
rect RREF trajectories in the Planes tasks, particularly in provide examples from this group’s solutions. Although B13
terms of variable creation and renaming. A compounding was not able to row reduce the matrix in part a (Fig. 17a),
factor was students not knowing how to manage a technical B13 did row reduce the matrix in part b (Fig. 18a) and cor-
difficulty with their calculator. The most commonly used rectly symbolized the new system, indicating one free vari-
calculators in the class were the Texas Instruments, TI-83 able. However, this student circled the word point instead
and TI-84 calculators. Neither of these calculators will row of line. We do not know why the student circled point, but
reduce a matrix that has more rows than columns. In class we note that renaming the variables from x, y to x1, x2 may
students were taught a workaround of adding columns of have kept the student from recognizing a familiar equation
zeros to create a square matrix, and then ignoring those of a line.
13
M. Zandieh, C. Andrews‑Larson
Student B2 followed up on the work in Fig. 17b by still solve. Students’ explanations of these solution forms (SE-
using the extra column of zeros. Thus the one free variable S) sometimes emphasized the geometric context and other
expected in a correct solution to part b turned into two free times focused more on algebraic similarity.
variables for student B2 (Fig. 18b). This variable creation Four of the five students correctly circled that the inter-
and renaming suggests that B2 was not connecting to the section of the three lines in Lb was “a line”. As shown in
geometric context of the intersection of lines in R2. Fig. 19b, B25 explained this interpretation by saying “They
are on top of each other. There are inf [sic] many solutions.”
5.2.3 Systems strategies for both La and Lb (5 students) On the other hand, one of the five students, B3, had a differ-
ent interpretation for Lb. After correctly using substitution
Systems trajectories relied on creating new forms of the to solve for (4/3, 2/3) in La, B3 answered Lb with, “They are
original system of equations and making connections to the scalars A.K.A. same equation. THUS there is no intersec-
geometric context. Although only 5 students used a Systems tion!” B3 indicated in La that the first three equations were
trajectory for all aspects of La and Lb, some of the strategies “scalars thus same as 2x − y = 2.” However, B3 interpreted
they used were also leveraged by the students who switched this in Lb to mean that the three “same” lines had no inter-
to a Systems strategy after an initial attempt to use an RREF section. This is consistent with findings in Oktaç (2018), and
strategy (see Sect. 5.2.4). Of the 5 students using only a we imagine this may be based on a concept image or natu-
Systems trajectory, two graphed the lines on their page and ral language interpretation of intersection as being a single
stated the intersection of those lines as their solution to La. point where two non-parallel lines meet or cross. One of the
Unfortunately, each of these students had sign errors that students in Sect. 5.2.4 also used this same lines means no
caused them to have incorrect intersections. The other three intersection reasoning.
students leveraged the fact that the first three equations were
“the same” and used substitution or elimination to correctly 5.2.4 Switching from an RREF strategy to a systems
find the intersection point (4/3, 2/3). Figure 19a shows the strategy (12 students)
work of student B25.
On part b, all five students noted “same line” or “same Switchers are illustrated in Fig. 15 by columns that have one
equation” referring to the first two or three equations. Some or more green boxes at the top of the column indicating an
students symbolized the y = mx + b format while other RREF strategy with blue boxes underneath the green boxes
students created new versions of the ax + by = c format. indicating a switch to a Systems strategy. Eight students in
Note that B25 symbolized the latter format for La, but y = this category switched to a Systems strategy after symbolizing
mx + b for Lb (see Fig. 19). Both notations leverage previ- only the augmented matrix or the augmented matrix and the
ous knowledge of algebra, but this is especially true for y RREF. These students appear in Fig. 15 in columns 21–27
= mx + b which B25 uses to emphasize that “They are the and 29. In each of these cases, after switching to the Systems
same line with the same slope and y intercept.” In this way, strategy, the students used methods similar to the students in
students leveraged SE-E forms to note similarities and to Sect. 5.2.3 (e.g., Fig. 19b). For example, student B4 wrote
13
Symbolizing while solving linear systems
the correct augmented matrix, then marked it out and rewrote shifts using the lens of Larson and Zandieh’s (2013) inter-
each of the equations in the form y = 2x − 2 or y = 1/2x. Noting pretations, their algebraic and geometric contexts, and the
that the first three equations were “same”, B4 used substitu- various forms delineated in the expanded framework pro-
tion to find the correct intersection point. For Lb, B4 wrote, posed in this paper. This allowed us to do three things:
“They intersect at all points because they form a line.” identify those symbolic shifts where students were highly
Two students (columns 10 and 11 of Fig. 15) completed successful, identify those symbolic shifts that were trouble-
part a with an RREF trajectory just as in Fig. 16a, but then some, and identify particular student reasoning patterns (e.g.
switched to a Systems strategy similar to student B4 or stu- variable creation and renaming, imagined parameter) that
dents in 4.2.3. For example, A15 stated, “The first three lines occurred around troublesome symbolic shifts.
are scalar multiples of each other, therefore they have infinite In the Planes case, most students relied on RREF strat-
intersection along y = 2x − 2.” egies and were highly successful rewriting the system as
Columns 19 and 20 of Fig. 15, illustrate two additional an augmented matrix (AM-O) and using technology to find
students (B10, B30) who made a switch to a Systems strategy the RREF (AM-E). However, students were less successful
between La and Lb, but they adopted a (correct) “modified” reconstructing systems (SE-E) from the RREF and using
RREF strategy for La, denoted in mint green in Fig. 15. Fig- this information to identify and interpret solutions (SE-S or
ure 20 illustrates that student B10 wrote “Row 1, 2, 3 are sca- LC-S). In particular, the condensed RREF form led some
lar multiples” and then row reduced a smaller matrix using students to engage in variable creation and renaming when
only row 1 and row 4. This modified RREF strategy lever- shifting from an AM interpretation (AM-E) to a systems
ages an interpretation of the first three equations as being of equations interpretation (SE-S) in ways that were incon-
the same line, so that the intersection of the four lines is just sistent with the original system. In addition, some students
the intersection of one of the “same” lines with the differ- seemed to engage in imagined parameter reasoning when
ent line. Consistent with this reasoning, both students who transitioning from algebraic solution forms into the geomet-
used a modified RREF strategy for La completed Lb using a ric context.
Systems strategy. For example, B10 stated, “the first 3 lines In the lines case, a sizable portion of students began with
overlay and intersect on all points on the line y = 2x − 2.” or switched to a Systems strategy. Students’ previous expe-
All 12 switchers showed considerable flexibility in sym- riences with lines in both algebraic and geometric contexts
bolizing various forms of the system of equations and in gave them a flexibility in symbolizing the Lines task that
coordinating those algebraic forms with the geometric con- was not demonstrated by students solving the Planes task. In
text of lines intersecting. A particularly powerful strategy particular, same line reasoning allowed for simplified Sys-
was same line reasoning, i.e., the ability to note when equa- tems trajectories. In what follows, we use Figs. 21 and 22 to
tions represent the same line using either a slope-intercept summarize each of the two main strategies and to structure
symbolization or a scalar multiple symbolization. our discussion about the conceptual subtleties involved when
students solve systems of equations. We then conclude with
recommendations for future research and instruction.
6 Discussion Figure 21 highlights the main symbolic shifts in student
trajectories when using an RREF strategy. Within the box
In this analysis, we examined how students engaged in the we also note the interpretation-form pair (as in Fig. 2) for
disciplinary practice of symbolizing while solving systems the first five boxes where a symbolic form was used. The last
of linear equations. The primary symbolizing trajectories we box indicates the switch to a geometric context required by
observed were organized around RREF strategies, Systems part c. Clearly this process requests students to have consid-
strategies, or transitions from the former to the latter. In the erable flexibility in moving between various interpretations,
context of these trajectories, we analyzed students’ symbolic contexts, and forms.
In addition to the transitions between the boxes indicated
by the right-pointing arrows, below the boxes we empha-
size two particular groupings of the forms and contexts.
Although each of the interpretation-form pairs in Fig. 21
has the same solution set, we can think of them as actually
referring to two different systems of equations. The first two
boxes indicate two interpretations (SE and AM) of the same
system, the original system. The next three boxes indicate
interpretation-form pairs for a different system. This new
system is equivalent to the original system in that it has the
Fig. 20 Modified RREF strategy (student B10, lines part a) same solution set, but these new interpretation-form pairs
13
M. Zandieh, C. Andrews‑Larson
cannot be determined from the original system by transi- We note two main differences between the trajecto-
tioning between the interpretations as described in Fig. 1. In ries for an RREF strategy and that for a Systems strategy
other words, Fig. 1 describes interpretations of a particular observed in our study. One is partially accounted for in the
system. Transitioning among the interpretations and con- second box of Fig. 22. Since the systems of equations that
texts in Fig. 1 allows one to describe that same system in the students were given for this study included some redun-
multiple ways. However, moving from box two to box three dancy, students often noticed that redundancy (e.g., same
in Fig. 21, is a change in the system that is being described. line reasoning) and rewrote equations into a new form of
This change requires movement to what could be considered SE-O to account for that. The second difference that is not
a new set of relationships between interpretations and con- highlighted in Fig. 22, is that students often used words ref-
texts that is relevant to this new system. In essence, there are erencing the geometric context (e,g., lines or intersection)
two versions of Fig. 1 when solving, the set of relationships when progressing through the milestones of this Systems
for the original system and the set of relationships for the trajectory. In contrast, students using an RREF strategy
new, equivalent system. rarely mentioned the geometric context until they were asked
The main trajectory for the Systems strategy, as highlighted directly about it for part c.
in Fig. 22, has both similarities and differences with the RREF
trajectory. As shown in Fig. 22, the first two boxes describe 6.1 Implications for future research and instruction
the original system and the last two boxes describe a new,
equivalent system that is created by the algebraic processes The primary contribution of our analysis is twofold. First, we
of substitution and/or elimination. This parallels the trajectory highlight the considerable nuance and variation entailed in
in Fig. 22 in which the equivalent system is created by a row students’ symbolizing processes as they work to solve sys-
reduction process. One difference is that, for the students in tems of linear equations. Second, the expansion of our con-
this study, the row reduction step was done on a calculator ceptual framework in Figs. 1 and 2 provides a structure and
which obscured the elements involved in the transition. Note language for detailing these processes through interpretations,
that each step in a row reduction or a substitution and elimi- contexts, and forms. Students must come to coordinate these
nation process potentially gives rise to a different equivalent interpretations, contexts, and forms as they work with systems
system. In Figs. 21 and 22, these multiple equivalent systems that are the same, and that are different but equivalent in the
are all grouped together in one box. Once a simplification pro- sense that they have the same solution set. In looking at the
cess (e.g. elimination, substitution, and/or row reduction) has various transitions indicated in Figs. 21 and 22, we note that
been completed, the final equivalent system may be rewritten some transitions were much easier for students than others. An
in various ways to denote a solution form. important area for future research is to explore in more detail
which transitions are more or less challenging for students and
13
Symbolizing while solving linear systems
what causes some transitions to be easier than others. Although Harel, G. (2017). The learning and teaching of linear algebra:
we have a focus on particular transitions that our students Observations and generalizations. The Journal of Mathemati-
cal Behavior, 46, 69–95.
attempted for this solving process, transitions across the many Hillel, J. (2000). Modes of description and the problem of represen-
interpretations, contexts and form pairings could be relevant. tation in linear algebra. In J.-L. Dorier (Ed.), On the teaching
Our data was limited to students’ written responses, so studies of linear algebra (pp. 191–207). Dordrecht: Kluwer Academic
that incorporated interview data could be particularly helpful Publishers.
Larson, C., & Zandieh, M. (2013). Three interpretations of the matrix
in gaining more detailed insights into students’ reasoning. equation Ax = b. For the Learning of Mathematics, 33(2), 11–17.
We argue that the prevalence of variable creation and Lay, D. (2012). Linear algebra and its applications (4th ed.). Hobo-
renaming observed in our data set indicates a need for more ken: Pearson.
careful attention to the notion of equivalent systems. This Oktaç, A. (2018). Conceptions about system of linear equations and
solution. In S. Stewart, C. Andrews-Larson, A. Berman, & M.
argument is consistent with Steinberg et al.’s (1991) finding Zandieh (Eds.), Challenges and strategies in teaching linear
that secondary students who were most successful at reason- algebra (pp. 71–101). Berlin: Springer.
ing about solutions to equations were those who transformed Possani, E., Trigueros, M., Preciado, J. G., & Lozano, M. D. (2010).
equations to determine equivalence of equations. It is also Use of models in the teaching of linear algebra. Linear Algebra
and its Applications, 432(8), 2125–2140.
consistent with Harel’s (2017) finding that it is not immedi- Rasmussen, C., Wawro, M., & Zandieh, M. (2015). Examining indi-
ately obvious to learners that some transformations of linear vidual and collective level mathematical progress. Educational
systems will in fact leave solution sets unchanged. Thus, we Studies in Mathematics, 88(2), 259–281.
argue that further exploration of students’ understanding of Rasmussen, C., Zandieh, M., King, K., & Teppo, A. (2005). Advanc-
ing mathematical activity: A view of advanced mathematical
equivalent systems is fertile ground for future research, as thinking. Mathematical Thinking and Learning, 7, 51–73.
well as an important point to which instructors and instruc- Sandoval, I., & Possani, E. (2016). An analysis of different repre-
tional designers should attend more carefully. sentations for vectors and planes in ℝ3. Educational Studies in
We also argue that students’ success leveraging geometry Mathematics, 92(1), 109–127.
Sierpinska, A. (2000). On some aspects of students’ thinking in lin-
to reason about solutions in the lines case—paired with stu- ear algebra. In J.-L. Dorier (Ed.), On the teaching of linear
dents’ difficulties making sense of solutions geometrically algebra (pp. 209–246). Dordrecht: Kluwer.
in the planes case—carries implications for instructors and Steinberg, R. M., Sleeman, D. H., & Ktorza, D. (1991). Algebra
instructional designers. Our data in the lines case points to students’ knowledge of equivalence of equations. Journal for
Research in Mathematics Education, 22(2), 112–121.
the power of imagery when algebraic and geometric con- Stewart, S., & Thomas, M. O. J. (2009). A framework for mathemati-
texts are well-coordinated. Our data in the planes case sug- cal thinking: The case of linear algebra. International Journal
gests that many students struggle to coordinate symbolic of Mathematical Education in Science and Technology, 40(7),
and geometric contexts in R3. Sandoval and Possani (2016) 951–961.
Trigueros, M. (2018). Learning linear algebra using models and
identify tasks that press students to explicitly coordinate the conceptual activities. In S. Stewart, C. Andrews-Larson, A.
geometry of planes and their intersections with parametric Berman, & M. Zandieh (Eds.), Challenges and strategies in
vector equations. Perhaps experience working on tasks such teaching linear algebra (pp. 29–50). Berlin: Springer.
as these would help students develop more flexibility in tran- Trigueros, M., & Possani, E. (2013). Using an economics model for
teaching linear algebra. Linear Algebra and its Applications,
sitioning between interpretations, contexts and forms in R3 438(4), 1779–1792.
in ways that could be leveraged to enhance their understand- Zandieh, M. (2000). A theoretical framework for analyzing student
ings of linear systems and their solution sets. understanding of the concept of derivative. CBMS Issues in
Mathematics Education, 8, 103–127.
Zandieh, M. J., & Knapp, J. (2006). Exploring the role of metonymy
in mathematical understanding and reasoning: The concept of
Funding Funding was provided by National Science Foundation (Grant derivative as an example. The Journal of Mathematical Behav-
no. DUE 1712524). ior, 25(1), 1–17.
Zandieh, M., & Knapp, J. (2018). Metonymy and metaphor: How lan-
guage can impact understanding of mathematical concepts (Part
References I). MathAMATYC Educator, 9(2), 23–27.
Zandieh, M., & Rasmussen, C. (2010). Defining as a mathematical
activity: A framework for characterizing progress from informal
Dorier, J. L., Robert, A., Robinet, J., & Rogalski, M. (2000). The to more formal ways of reasoning. The Journal of Mathematical
obstacles of formalism in linear algebra. In J. L. Dorier (Ed.), Behavior, 29(2), 57–75.
On the teaching of linear algebra (pp. 85–124). Dordrecht: Zandieh, M., Wawro, M., & Rasmussen, C. (2017). An example of
Kluwer. inquiry in linear algebra: The roles of symbolizing and brokering.
Dorier, J.-L., & Sierpinska, A. (2001). Research into the teaching and PRIMUS, 27(1), 96–124.
learning of linear algebra. In D. Holton (Ed.), The teaching and
learning of mathematics at university level: An ICMI study (pp.
Publisher’s Note Springer Nature remains neutral with regard to
255–273). Dordrecht: Kluwer Academic Publishers.
jurisdictional claims in published maps and institutional affiliations.
Freudenthal, H. (1991). Revisiting mathematics education.
Dordrecht: Kluwer Academic.
13