Student Difficulties With Linearity and Linear Functions and Teachers' Understanding of Student Difficulties
Student Difficulties With Linearity and Linear Functions and Teachers' Understanding of Student Difficulties
by
Valentina Postelnicu
May 2011
UMI Number: 3450231
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.
UMI 3450231
Copyright 2011 by ProQuest LLC.
All rights reserved. This edition of the work is protected against
unauthorized copying under Title 17, United States Code.
ProQuest LLC
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106-1346
ABSTRACT
difficulties with aspects of linearity and linear functions, and to assess their
(MDT) on aspects of linearity and linear functions, ranked the MDT problems by
were conducted with 40 students and 20 teachers. A cluster analysis revealed the
grade level, and Group 1 enrolled in courses above their grade level. A factor
analysis confirmed the importance of slope and the Cartesian connection for
student understanding of linearity and linear functions. There was little variation
MDT increased with more advanced courses, mainly due to Group 1 student
slope from the graph of a line. That difficulty persisted across grades,
MDT, showed that students correctly identified the problems with the highest
MDT mean scores as being least difficult for them. Only Group 1 students could
identify some of the problems with lower MDT mean scores as being difficult.
i
Teachers did not identify MDT problems that posed the greatest difficulty for
their students. Student interviews confirmed difficulties with slope and the
format and length. Teachers did not identify student difficulties with slope in a
geometric context.
ii
ACKNOWLEDGMENTS
Carole Greenes for her generous and unwavering support throughout this journey.
From the beginning, she set high standards and guided me with immense
knowledge and patience, page after page, chapter after chapter, draft after draft,
towards the final draft of my thesis. This dissertation would not have been
possible without her help and encouragement. Her determination and passion for
the way with his intimate knowledge of Statistics. He has always offered generous
and critical feedback, and ideas for new directions regarding the interpretation of
years.
to have had access to his vast reservoir of knowledge, and I am grateful for his
I would also like to pay tribute to Freudenthal and von Glasersfeld. Reading
their books gave me a feeling of déjà vu, and generated the belief that there might
iii
Lastly, I would like to thank my husband Florin, my daughters Iris and
The research reported here was supported in part by a grant from the
iv
TABLE OF CONTENTS
Page
LIST OF TABLES...................................................................................................... ix
CHAPTER
Careers ........................................................................................ 7
.................................................................................................... 24
v
CHAPTER Page
Functions ......................................................................................... 30
Functions ................................................................................... 30
Questionnaires .......................................................................... 95
vi
CHAPTER Page
vii
CHAPTER Page
APPENDIX
viii
LIST OF TABLES
Table Page
Slope .................................................................................................. 99
ix
Table Page
24. Means and Standard Deviations for Scores on MDT Problems .... 136
35. Mean Scores for MDT Problems by Grade for Group 0 ............... 153
36. Mean Scores for MDT Problems by Current Mathematics Course for
x
Table Page
37. Mean Scores for MDT Problems by Grade for Group 1 ............... 154
38. Mean Scores for MDT Problems by Current Mathematics Course for
44. Group 0 Student Difficulties with MDT Problems 2 and 7b ......... 171
45. Group 1 Student Difficulties with MDT Problems 1 and 4 ............ 174
47. Final Coding Categories for Conceptual Knowledge of Slope ..... 178
48. Final Coding Categories for Student Difficulties with the Cartesian
49. Final Coding Categories for Student Difficulties with Slope ........ 180
52. Scoring Guide for MDT Problem 1 Types of Responses .............. 262
xi
Table Page
53. Scoring Guide for MDT Problem 2 Types of Responses ............. 263
54. Scoring Guide for MDT Problem 3 Types of Responses ............. 263
55. Scoring Guide for MDT Problem 4 Types of Responses ............. 264
56. Scoring Guide for MDT Problem 5 Types of Responses .............. 264
57. Scoring Guide for MDT Problem 6 Types of Responses ............. 265
58. Scoring Guide for MDT Problem 7a Types of Responses ............ 265
59. Scoring Guide for MDT Problem 7b Type of Responses .............. 266
60. Scoring Guide for MDT Problem 7c Types of Responses ............. 266
68. Conceptual Knowledge of Slope (CKS) and its Subcategories ...... 273
xii
Table Page
xiii
LIST OF FIGURES
Figure Page
1a. Histogram of MDT Scores Distribution for Grade 8 Students ...... 123
1b. Histogram of MDT Scores Distribution for Grade 9 Students ...... 124
1c. Histogram of MDT Scores Distribution for Grade 10 Students .... 124
4a. Histogram of MDT Score Distribution for Group 0 Students ....... 133
4b. Histogram of MDT Score Distribution for Group 1 Students ....... 133
xiv
Figure Page
xv
Chapter 1
1.1 Introduction
because of its place as the first course in the pipeline to the study of more
advanced mathematics, science, and technology (Katz, 2007a). Since the early
1990s and the beginning of the Algebra for All movement, student difficulties
with algebra and algebraic thinking have been studied and well documented.
nationwide (Caglayan & Olive, 2010; Greenes & Rubenstein, 2008; Kaput,
Carraher, & Blanton, 2008; Kieran & Chalouh, 1993; Kilpatrick & Izsák, 2008;
Moses B., 1999; Rowntree, 2009). Failure rates in Algebra I have ranged from
In efforts to gain insight into the nature of difficulties with algebra, some
1998; Lanier, 1999; Lobato & Siebert, 2002; Moschkovich, 1992, 1999;
Moschkovich, Schoenfeld, & Arcavi, 1993; Schoenfeld, Smith III, & Arcavi,
1993; Tourniaire & Pulos, 1985; Zaslavsky, Sela, & Leron, 2002). Other
1
(Nathan & Koedinger, 2000) particularly the source and nature of the difficulties
(Ding, 2007; Thompson A. & Thompson P., 1996; Thompson P. & Thompson
A., 1994). There are indications that teachers are not aware of the types of
difficulties students have with basic concepts of algebra, and their predictions
Statement of the Problem. The dual purpose of this study was to identify
a. Grade level?
c. Problem?
2
5. Do teachers’ speculations about the nature of students’ difficulties
a c
proportionality m. A proportion can be seen as a relationship of direct
b d
by the participants in both the pilot study and this full study (Burger et al., 2007;
Collins et al., 1998; Larson, Boswell, Kanold, & Stiff, 2004). Larson et al.
(2004) describe a model of direct variation: “When two quantities y and x have a
constant ratio k, ( k z 0 ), they are said to have direct variation. The constant k is
y
called constant of variation. If k , then y kx “ (p. 236).
x
3
Larson et al. (2004) define the linear equation in x and y as the equation that can
be written as “Ax + By = C where A and B are not both zero” (p. 209). The
same definition is used Burger et al. (2007, p. 296) and by Collins et al. (1998,
p. 281).
the origin of the Cartesian system of coordinates. When b z 0 , the line crosses
b b
x , m z 0 . In a Cartesian system of coordinates, the point ( , 0)
m m
represents the point of intersection between the graph of a function Gf and the
value of the function f (0) ; the solution is f (0) b . The point (0, b) represents
where a graph crosses the X-axis and the y-coordinate of a point where a graph
crosses the Y-axis (Larson et al., 2004, p. 222). The same definitions are found
textbooks as “rise over run” and a formula is provided for calculating slope
when the Cartesian coordinates of two points, ( x1 , y1 ) and ( x2 , y 2 ) are given; the
4
y 2 y1
slope is m (Larson et al., 2004, pp. 229-230). Some textbooks use
x 2 x1
example, Burger et al. (2007) define slope as “the constant rate of change
shown by the line,” (p. 296) and Collins et al. (1998) define slope as “the ratio
points ( x1 , y1 ) and ( x2 , y 2 ) , the slope of the line connecting the two points is
y 2 y1
m . Slope is defined algebraically as one of the parameters in the
x 2 x1
steepness of physical objects such as mountain roads, ski slopes, and wheelchair
'y
a dynamic relationship, slope represents the constant rate of change m
'x
5
(Knuth, 2000; Moschkovich et al., 1993; Schoenfeld et al., 1993). Schoenfeld et
connection, as for example, the association between the change in y and the
change in x from the slope formula, and their graphic representations as line
1.3 Rationale
with scientific and technological advances, the United States must increase its
engineering, physics and the natural sciences (Couto, Mani, Lewin, & Peeters,
2007; Lewin & Couto, 2006; National Science Board, 2008, 2010). Since the late
1990s, the United States has lost its leadership capacity in these areas. Data from
the National Science Board (2008) and the Business Higher Education Forum
engineering being awarded in China, India and Europe. The United States “has
lost ground in many areas: college degrees earned in STEM fields; scientific
journal articles authored; patents awarded; and Nobel Prizes won” (Business-
The national need for mathematicians and scientists places great responsibility
on elementary through high school and college programs to engage more students
6
Public Works, 2006). National and state organizations interested in increasing the
number of workers with 21st century skills have recommended that students take
Education Week, 2006; National Center for Public Policy and Higher Education,
2004; Partnership for 21st Century Skills, 2007). Research studies examining the
school, entrance and graduation from college, and successful entry into the
workforce, found that two courses, Algebra I and Geometry, had the highest
America (Katz, 2007a), have recognized the value of algebra to the study of
technology (Katz, 2007a; Ma, 2009; Sadler & Tai, 2007). In its Position Paper
for School Mathematics, the NCTM (2000) identifies algebra as one of five
7
major strands in mathematics for study by students Prekindergarten through
Grade 12, and highlights the importance of knowledge of algebra and abilities to
the assessment than did Grade 8 students enrolled in Pre-Algebra, and Pre-
Algebra Grade 8 students had higher NAEP scores than did students enrolled in
Given the NAEP results plus recognition of the importance of algebra in the
pipeline preparation for STEM careers, educators and others began to promote
algebra as a course for all students. In February 1994, the Algebra for All
enroll in Algebra I. From 1996 to 2005, the proportion of U.S. students taking
8
1.3.3 Addressing Failure with Algebra I. Shortly after the mandate of
Algebra for All, publications began appearing that identified huge failure rates in
Algebra I courses across the country (Gamoran & Hannigan, 2000; Kilpatrick &
Even the most successful school districts participating in the College Board
Equity 2000 program, reported failing rates of 25% (Gamoran & Hannigan,
2000). Roy Romer, Los Angeles Unified School District superintendent between
2001 and 2006, commented in January 2006 that Algebra I “triggers (school)
dropouts more than any single subject” (Helfand, 2006). In 2004 in Los
Angeles schools, the failure rate in Algebra I was 44% (Helfand, 2006). In
2005-2006, the Algebra I failure rate overall in Virginia was 19.1%. The rate
Jersey students (Carroll, 2009), and 84% of Clark County, California students
mathematics courses. For this reason, the large failure rate prompted much
research about student difficulties with key Algebra I concepts. Among these
concepts are linearity and linear functions, which have been identified by the
9
and generalization to other functions like quadratic, exponential, and
functions. Not only are these concepts important for preparation for higher level
Many educators believe that student difficulties with Algebra stem from the
consider alternative delivery methods for instruction (Rettig & Canady, 1998)
and research to gain greater insight into conceptual difficulties (Chazan, 2008;
Doerr, 2004; Greenes & Rubenstein, 2008; Kaput, Blanton, & Moreno, 2008;
Kieran, 1992; Moses, 1999) failure rates in Algebra I persisted (Kilpatrick &
Izsák, 2008).
10
1.3.4 Studies of Student Difficulties with Linearity and Linear Functions
1993; Schoenfeld et al., 1993; Smith, Arcavi, & Schoenfeld, 1989), and linear
of slope and the functional aspect as rate of change. Lobato and Siebert
concluded that students may conceive of the speed of a mobile differently from
the steepness of a ramp, despite the fact that, mathematically, speed magnitude
while interacting with the software during various tasks (Schoenfeld et al.,
11
1993), and observed during their classroom work and participation in peer
al., 1993) and revealed student difficulties with making connections among
During this time, and with an approach that delayed the introduction of formal
and the teacher’s conception of the y-intercept and his assumptions that students
contexts. The use of informal phrase “starting point” for the y-intercept
12
and graphic representations, and seemed to have a negative impact on student
the sample was small, the study showed that student difficulties with connecting
linearity and linear functions are complex topics, and student difficulties occur
Moschkovich et al., 1993; Schoenfeld et al., 1993; Smith et al., 1989). With the
with the y-intercept (Davis, 2007), all other studies addressed only student
researchers. None of the studies asked students to explain the nature of their
students’ difficulties.
13
1.3.5 Studies with other Populations. Several studies, conducted with other
populations, including elementary and middle school students, upper level high
difficulties with the concepts of slope and rate of change, connections among
De Bock et al., 2002; De Bock, Van Dooren, Janssens, & Verschaffel, 2007;
Van Dooren, De Bock, Depaepe, Janssens, & Verschaffel, 2003; Van Dooren,
De Bock, Janssens, & Verschaffel, 2007), found that students have great
school and college (Ben-Chaim et al., 1998; Post, Behr, & Lesh, 1988).
Li (2006) studied a group of 317 students in Grades 6 and 7 from two states
tasks focusing on the concepts of variable, linear equation, and linear function.
For example, when asked to solve problems like, “Look at the equation
with a process view of the linear function would use a table of values (input-
14
output table) to answer the question, while students with an object view of the
that the persistence of difficulties might be caused by the use of the recycling
curriculum in which the same topics are presented in the same way over several
years.
student difficulties with the concept of slope are robust. In physical situations,
better in functional situations, but could not connect the rate of change of a
function to the slope of the line representing that function. Students’ use of
and describing the rate of change (e.g., “The more tickets you sell the more
profit you make”) was associated with less conceptualization of slope as ratio or
rate.
the linguistic and graphic connection when modeling change with functions.
15
used by teachers in instruction or the textbooks themselves contains words and
need explanation. Yet these terms or phrases remain opaque to students. For
used. The 124 participants in their study included Grade 11 students, prospective
function, and as a geometric attribute of the line representing the graph of the
the analytic and the geometric perspectives, while others experienced difficulties
in connecting the two perspectives or favored only one of them. This study
representations, and the need for teachers and instructional materials to make
explicit the origin, sense, and the unit of measurement for a Cartesian system of
coordinates. It also points out that higher levels of mathematical knowledge are
16
associated with higher rates of success with connecting geometric and analytical
perspectives of slope.
situations, was conducted with college students, high school students and pre-
Zee, 1987). The study revealed that the nature of the participants’ difficulties
Hedberg, & Shagle, 2007) points out that solving word problems is considered
McDermott et al.’s (1987) study, the source of the trouble may be the difficulty
physics, further reinforces the need for the proposed study. Moreover,
17
common to both students and teachers, and raise the question of teachers’
concepts.
(Breidenbach, Dubinky, Hawks, & Nichols, 1992; Dubinsky & Harel, 1992;
Hansson, 2006; Norman, 1992; Sfard, 1992; Sierpinska, 1992; Slavit, 1997;
knowledge and pedagogical knowledge (Stump, 1997; You, 2006) and the
(Ding, 2007).
Using interviews and surveys, Stump (1997) examined the mathematical and
mathematics teachers. Results showed that most of the teachers had a geometric
or ratio aspects of slope were less mentioned in their descriptions of the concept.
of slope, like difficulties with the graphing of lines. What was not considered by
Stump was how the nature of teachers’ knowledge of slope correlates with their
18
students’ knowledge of the concept, nor whether teachers’ speculations about
their student difficulties with the concept of slope correspond with actual student
difficulties.
understanding of the nature or extent of these difficulties. The present study was
their students’ difficulties. The results of this study have the potential of better
to enable them to better respond to the mathematical talents and needs of their
students.
and the research questions (Section 1.1), definition of terms used to describe
linearity and linear functions (Section 1.2), and rationale for this study (Section
19
1.3). In the next chapter, research literature to support this study is discussed and
analyzed.
20
Chapter 2
REVIEW OF LITERATURE
2.1 Introduction
The purpose of this study was to investigate student difficulties with concepts
(Section 2.2) and algebra as a course or strand in the school curriculum are
studies on student difficulties with the Cartesian connection (Sections 2.4 and
2.5). The chapter continues with studies on student and teacher assessment of
problem difficulty (Section 2.6), and concludes with several remarks (Section
2.7).
In 1770, Euler defined algebra by stating its purpose, as “the science which
teaches how to determine unknown quantities by means of those that are known”
21
These definitions of algebra can be traced back more than four millennia ago
to the times of the ancient Akkadian Empire (2300 BC – 2200 BC). At that time,
formulas, but by examples. Babylonian clay tablets, dated from the twentieth
requiring the length and the width of a rectangle if certain relationships were
given (e.g., the sum and the product of the length and the width are known) (Katz,
stage (Katz, 1998; Cooke, 1997), was characterized by the use of natural language
(e.g., words, sentences) and some abbreviations to express the rules used for
the centuries by the quest for solving equations. During the third century,
Diophantus (c. 200 - c. 284) gave an algorithm for solving quadratic equations. In
the ninth century, Al-Khwarizmi (c. 780 - c. 850) wrote a manual for solving
abstract linear and quadratic equations based on the Babylonian algorithms used
for solving real-life problems. In the eleventh century, Omar Khayyam (1048-
1131) discovered a general method for solving cubic equations by employing the
published Ars Magna (1545) which contained algorithms for finding solutions to
third and fourth degree polynomial equations that were based on geometric
22
Empire until the European Renaissance, the justifications for the algorithms
The rise of algebra during the European Renaissance, between the fourteenth
and seventeenth centuries, marked the entrance of the symbolic stage with
consonants for known quantities). Over time all quantities and all operations and
The shift from the static approach for solving equations to the dynamic
function approach developed during the seventeenth century with the work of
attempts to describe the movement of projectiles and the solar system. The focus
communicate and remained geometric in nature until the work of Fermat and
a curve.
Algebra entered its abstract structural stage during the nineteenth century, and
started to depart from its aim of solving equations. The attempts of Lagrange
equations of degrees higher than four led to the development of group theory.
23
Abel used Lagrange’s idea of permutations to prove that there is no general
2.3.2).
course of study at universities, and the creation of algebra textbooks. The first
(Rey, 2006).
In the United States, Algebra as a course of study began during the eighteenth
century as a college course that focused on how to solve linear and quadratic
expressions (Kilpatrick & Izsák, 2008). The Benjamin Franklin Academy (later
time other colleges started to do so, as well. After the Civil War, geometry began
24
to be required for admission, thus setting in stone the algebra-geometry sequence
In 1827, Massachusetts passed America’s first high school law that mandated
the teaching of algebra, geometry and surveying in all high schools in every town
with more than 500 families. The content of the secondary algebra course was
The content of the high school algebra course was dominated by exercises in
During the first half of the twentieth century, the utility of algebra for the
content was revised to reflect the ideal of social efficiency (Kliebard, 2004). More
attention was given to using algebraic formulas, and interpreting graphs, and less
For a short period during the New Math era, from the mid 1950s to the mid
Sputnik crisis (1957), the new curriculum’s aim was to better prepare students for
college in order to preserve the United States’ place as world leader in space
25
and complex numbers, and the use of key unifying ideas like sets, functions and
relations, characterized the New Math algebra curriculum. Proofs also became
part of the algebra content because of their potential to help students understand
Of interest is the influence of Dieudonné and Piaget during the New Math era.
The New Math curriculum failed in the United States (Kline, 1973). The end
of the New Math era meant the end of the structural approach in the secondary
school algebra, and the end of Euclidean geometry as a course of study. The
on the “basic” algorithms applied to solving simple equations. The algebra course
was extended to two years, and separated in two parts, Algebra I and Algebra II,
Mathematics, 1997) advocated for equity and access to algebra for all students.
Moses (1995) and Mason (2008) considered that algebra should be treated as a
26
“new civil right,” and “a natural birthright of all human beings,” respectively. In
the earlier document and produced the Principles and Standards for School
major strand of the elementary through high school curriculum. Studies of student
algebra for their future academic achievement (Smith, 1996; Gamoran &
mathematics and sciences (Katz, 2007a; Matthews & Farmer, 2008), and started
to be pushed from Grade 9, where it has been positioned for a century, to Grade 8
The shift from the static equation solving to the dynamic functional approach
was the most notable conceptual change in the algebra secondary curriculum
include functions and their properties for purposes other than solving equations.
27
(Kilpatrick et al., 2007). The standards-based era culminated with the publication
of the Common Core State Standards for Mathematics (CCSSO & NGA, 2010)
accompanied with MDT problems that are illustrative of each of these four topics.
For example the pattern 12, 18, 24, 30 can be generalized to 6(n 1) or
For example:
y 2 y1
identified, and the ratio m must be calculated (see
x 2 x1
28
corresponding values for y must be calculated (see Appendix A, MDT
problem 2).
f ( x2 ) f ( x1 )
relationship between the rate of change m of the linear
x2 x1
y 2 y1
m of the line with the equation y mx b , and make the
x 2 x1
4. The study of structures, where variables are arbitrary marks on the page or
symbolic game. For example, apply the rules for simplifying algebraic
29
2.4 Research on Student Understanding of Linearity and Linear Functions
linearity and linear functions. Of interest to this study are student knowledge and
difficulties with slope or constant rate of change, and with connections among
conducted with students in Grades 8-10 and other age groups are presented
Grades 8 and 9 students enrolled in Algebra I to gain insight into their difficulties
with aspects of linearity and linear functions. Grenees et al. (2007b) designed a
7-item assessment, the Mini-diagnostic Test (MDT). The MDT is the same
instrument used by Greenes et al. (2007a) and the present study (see Appendix
A). The MDT was initially administered to 752 students in Grade 8 in five school
districts in the Northeast U.S. Subsequently, the MDT was translated and
administered to 405 Grade 8 students in Seoul, Korea, and 575 Grade 9 students
in Haifa, Israel. After completion, students were interviewed about their solution
processes. Across the three countries, the most difficult problems were MDT
problems 4, 7b, and 7c. Moreover, the same difficulties with slope, rate of change
or speed persisted across grade levels and instructional programs (i.e., different
textbooks). For MDT problem 4, the task is to calculate the slope of a line
30
graphed in a homogeneous system of coordinates. Only 33.8% of American
solved that problem. MDT problems 7b and 7c require calculation of the slope of
variable - time in hours, dependent variable - distance in miles), and the speed of
the car, respectively. Of note, MDT problems 7b and 7c refer to the same line
MDT problems 7b and 7c were 32% and 48.3%, respectively for the Unites
States, 49.9% and 39.5%, respectively for Korea, and 40% and 58.3%,
respectively for Israel. Overall, in the three countries, more students correctly
answered MDT problem 7c requiring the calculation of speed than MDT problem
7b requiring calculation of slope. American students also had difficulty with MDT
problems 1 and 6. For MDT problem 1 students had to decide whether the point
(2,8) is situated on the line y 3x 14 , and explain their reasoning. Only 52.3%
identification of the y-intercept from five choices (0, 2, 4, 6, and 12), given the
students could not connect the point ( x0 , y 0 ) with one of the solutions of the
equation y mx b , or the y-intercept with the point (0, b) from the line with the
31
equation ax by c . These findings are consistent with those noted by
MDT problem 4, students did not connect x2 x1 (change in x or “run”) with its
with positive sense. This is consistent with findings by Schoenfeld et al. (1993).
In addition, in MDT problem 4, students failed to connect the sign of the slope,
positive, with the graphed line representing an increasing function. For example,
et al. (1993). In the case of MDT problem 7b, students had difficulties calculating
the rise or the change in y. The geometric aspect of the slope prevailed for
students who answered 4/3 instead of 40, a finding in concert with the work of
Zaslavsky et al. (2002). Some students estimated 30 for the y-coordinate of the
point with the x-coordinate 1, and considered the point (1, 30) instead of the lattice
The present study was inspired by the one conducted by Greenes et al.
(2007a). Greenes et al.’s (2007a) study used only descriptive statistics, and neither
reported on the internal consistency of the MDT, nor analyzed the dimensions
across which the MDT problems aligned. Issues related to the scoring and coding
suggested by the fact that the committee of mathematics leaders who designed the
MDT speculated on the reasons for student difficulties with aspects of linearity
and linear functions, and failed to correctly identify specific difficulties 40% of
32
the time (Grenees et al., 2007b). This failure raised the question of whether
students and teachers could identify the difficulties. Consequently, in the present
study all students and their teachers were asked to rank the MDT problems by
difficulty, and comment on the nature of student difficulties with the MDT
problems.
connection, that is, the relationship between symbolic and graphic representations
of linear functions. Participants were 178 students from the same high school,
problems were designed, and each student was given a problem to solve. Thus,
each problem was solved by approximately one sixth of the participants. For
example, one problem was finding a solution to a linear equation of two variables,
given the graphic representation with two lattice points, (0,2) and (2, 4) , and
missing. Students were asked to give two solution strategies to solve the problem,
one algebraic and the other graphic, using the symbolic equation of the line, and
the graph of the line, respectively, and to explain their solutions. More than three
fourths of students used an algebraic approach, even when the graphic approach
Then, using points from the line, they calculated the slope of the line, and
33
identified the missing coefficient in the equation. Students failed to see those
solution. Knuth hypothesized that the Cartesian connection from the graph of the
line to the symbolic equation of the line is more difficult to master. Knuth argued
(2003) points out, the graphic approach may have been rejected by students (and
their teachers) because graphic representations are not precise, and could
were inferred by the researcher from student answers, without any interview input
from students or their teachers. Also, all students experienced the same
curriculum might encounter the same difficulties. In the present study, students
students (156 students in Grade 6, 115 students in Grade 7, and 146 students in
Grade 8) from the same public school, enrolled in either general mathematics
students, one focusing on ratio and proportionality, and the other on steepness.
The ratio and proportionality test had tasks of varying levels of complexity. For
example, in a task of lesser difficulty, a recipe for a soup for eight persons was
given, with the necessary quantities for each of the ingredients. Students had to
figure out quantities when the recipe was altered to four persons or six persons. In
another task, of greater difficulty, three doctors used an office two, four, and six
days per week, respectively. Together they paid $240 for lighting. Students had to
figure out the amount each doctor needed to pay so that the bill would be “fair.”
The steepness test contained tasks of comparing two lines, two roofs, and two
staircases. All lines, roofs, and staircases were represented graphically using
the variability in students’ scores on the steepness test could be explained by their
performance on the ratio and proportionality test. On the steepness test, students
two lines (80% correct answers), than in physical contexts when they were asked
to compare the slopes of two roofs (66% correct answers) or two staircases (56%
correct answers). In short, the difficulty of the task increased as the complexity of
the situation increased i.e., more lines, angles and triangles appeared in the roofs
and staircase situations than in the non-contextual lines situations. This finding is
35
not surprising, confirming previous research on proportionality/linearity (Kaput &
Apart from the shortcomings recognized by the researcher herself, like those
referring to the construct validity of the steepness test and the small sample of the
interviewed students, there are also issues in terms of the statistical methods used
about steepness, and their reasoning abilities about ratio and proportionality was
violated, given that all students were from the same school, and students in the
same class were taught by the same teacher. Thus, students shared the same
curriculum and the same instructional approaches, and the observations within
casts doubts upon any statistical significance reported in the study (Raudenbush &
Bryk, 2002). Also, the researcher did not take into consideration natural clusters,
current mathematics course (e.g., Regular Mathematics for 6th, 7th, or 8th grade,
and proportionality may be different for various groups of students (e.g., low-
present study, analyses of student performance on the MDT are made by grade,
36
Stump (2001a) studied student understanding of slope or rate of change.
for various gears. Students collected data, and constructed scatter plots and the
best-fitting line. Two months later, a test was administered and interviews were
conducted with all 22 students. In the test, students were presented with a scatter
plot, a graph of the best fitting line, and its symbolic representation
y 2.24 x 0.101 , and were asked about the slope of the line. Only eight students
mentioned a ratio or a rate and referred to the number of pedal revolutions and the
number of wheel revolutions. When asked, “What is slope?” fifteen students used
the word “angle,” nine students mentioned “rise over run,” four students referred
y 2 y1
to a formula (two students used the correct formula , and two students
x 2 x1
x 2 x1
reversed the ratio ), and eight students mentioned “steepness.” When
y 2 y1
presented with a model of a ski ramp and asked to measure its steepness, only five
students used a ratio (three students used “rise over run,” and two students used
y 2 y1
the formula ). Five other students thought of the height and the length of
x 2 x1
the ramp, but did not know how to use those to determine the slope. In another
was presented (independent variable - number of tickets sold for a dance concert,
and dependent variable- profit in dollars) and students were asked to figure out
37
the cost of a ticket and the slope of the graph. Only seven students correctly
identified the cost of a ticket, and only three of the seven students correctly
identified the slope of the line. Seven students stated that they did not know how
to find the cost of the ticket, and six students answered that they did not know
how to find the slope of the line. The remaining students gave incorrect answers.
All but four students correctly calculated the rate of growth given two sets of
values (age, height) for a child, (9, 4'3' ' ) and (11, 4'9' ' ) . In general, Stump’s
students performed better when asked to find the rate of change in functional
situations (e.g., finding the cost of a ticket given the number of tickets sold and
the profit) than when asked to find the slope as a measure of steepness (e.g.,
Teuscher and Reys (2010) studied student difficulties with identifying the rate
of change from a graph. Participants were 191 high school students enrolled in
presented (similar to MDT problem 7, Appendix A) and the task was to identify
the interval where the rate of change is 2. About 37% of students failed to
identify the interval. When asked to identify the interval on which the function
had the greatest rate of change, 57% of students disregarded the sign of the rate of
value, 2. Teuscher and Reys believed that student difficulties in finding the
greatest rate of change stemmed from difficulties with vocabulary, the meaning
38
assigned by students to the phrase, “greatest slope.” There is the possibility that,
in spite of instruction, students associated slope with the steepness of the hill or
mountain and the difficulty of climbing the hill, and disregarded the orientation of
the hill i.e., did not notice whether positive changes in the vertical height (change
in y) were associated with positive changes in the horizontal length (change in x).
Teuscher and Reys (2010) did not take into consideration the complexity of the
researchers from student solutions, with no input from students or their teachers in
including rate of change. Reported here is only the part of the study focusing on
school, ages 16-18 from four high schools, and 50 students, ages 18-22, from two
situation (water flowing in a tank at a constant rate of 2 units of depth per 1 unit
of time) and was accompanied by the corresponding table of values, with the
10 for the dependent variable y, and the graph of the function, y 2 x . The task
1
required finding the rate of change at x 2 units of time, and at x T , an
2
39
arbitraty number of units of time. For the second task, concerning the rate of
change of a linear function, the context was mathematical. The graph and the
1
participants had to figure out the rate of change at x 2 and at x T . Both
2
one item using a five-point scale {0, 1, 2, 3, 4}. The mean score for the high
school students was 2.22 , and for the future teachers, 2.02. For the first task, 20%
1
of students answered that the rate of change at x 2 is 5 (the value of the
2
1
function or the y-coordinate corresponding to x 2 ), and more than 20% of
2
students failed to respond or gave the answer “2T” for the rate of change at x T.
On the second task, with the graph of the linear function and no real-life context
1
x 2 as the value of the function y 3x 1 or the y-coordinate corresponding
2
1
to x 2 , and gave the answer “ 3x 1” for the rate of change at x T.
2
Interviews of all students about their solutions revealed that incorrect answers
of proportionality, but not to computational errors. For example, students did not
know that regardless of the “slope triangle” used, the same ratio for the slope of
40
Orton’s (1983) study demonstrated that high school and future mathematics
of change. Difficulties with rate of change occurred across both groups, regardless
and interviews were conducted with 12 of the participants. In Task 1, the function
and participants were required to find the slope of the function, and the tangent of
the angle β between the line and the positive sense of the X-axis. In Task 2, the
coordinates, and participants were also required to find the slope of the function,
and the tangent of the angle β between the line and the positive sense of the X-
geometric, and mixed. Those perspectives were common across all groups, high
perspective of slope, i.e., slope is a property of the linear function and does not
41
depend on the graphic representation, and found m 1 and tan E 1 . Ninety-one
participants held a mixed perspective of slope, i.e., they considered that the slope
1
( tan E ). Eight participants held a geometric or “visual” perspective of slope,
3
i.e., slope is an attribute of the line that varies when the system of coordinates
1 1
changes, and answered m and tan E . To illustrate those perspectives, five
3 3
reconciling the two perspectives. The conflict was generated by the conviction
that m tan E , which was contradicted by the “visual” appearances when a non-
only the Ph.D. candidates in mathematics education and holders of Ph.D. degrees
participants.
42
and linear functions emphasizes student difficulties and researchers’
Lobato and Siebert (2002) conducted a 10-day teaching experiment with nine
students in Grades 8-10 and analyzed the students’ abilities to create a “ratio-as-
measure” for slope. Tasks required students to decide what measurements were
needed to assess the steepness of a ramp, determine the steepness of the ramp, and
modify the ramp such that its steepness remains the same but its height increases.
During interviews, students were asked, “What is slope?” and were invited to
comment on other students’ ideas about measuring the steepness of ramps. They
also were asked to draw a picture of a ramp with a slope of 2. Only one student’s
conceptual development was reported, the student who made the most progress
during the teaching experiment. That student was at the end of the Grade 8 year,
and had just completed Algebra I. Although the student was acquainted with the
slope formula, at the beginning of the teaching experiment he attended only to the
height of the ramp as a measure of steepness, and treated the width of the ramp as
resulting from changes in height. Gradually, he coordinated the height with the
width of the ramp and considered them as two quantities changing in tandem, and
software simulation of a linear movement of a clown and a frog, students had the
task of assigning values for distance and time so that the frog’s walking speed
43
successfully solved the task of finding values for distance and time (e.g., 2.5 cm
in 1 second) to match the frog’s speed with the clown’s speed. However, he failed
to see how the situations were alike mathematically. Lobato and Siebert pointed
the results of these studies. For example, the situation involving slope as a
measure of the steepness of a ramp is more complex than the situation involving
linear moment with a constant speed (Lobato & Siebert, 2002). Reasoning
proportionally is most difficult for students when the geometric aspects are more
complex (Kaput & Maxwell-West, 1994; Lamon, 1993). The ramp situation
two quantities. One model of a linear function was a spool elevating system with
the independent variable being the number of rotations of the handle on the spool
system, and the dependent variable being the height of the object. The
development of the student’s knowledge of rate of change was similar to the one
described by Thompson (1994b). At the beginning, the student had an action view
of functions, concentrating on the input and output values, one a time. At the end
of the teaching experiment, the student held a process view (Sfard, 1991; Slavit,
44
1997) of linear functions, with the dependent variable and the independent
equations). Since Hines created special learning experiences for the student,
each one hour and 50 minutes in duration. Students were presented with a table of
values, and were asked to identify the pattern and continue it by adding new
entries into the table. Group 2, consisting of seven Grade 7 students, studied
characters. In one of the tasks, a table of values was given with distances and
times representing the movement of a frog (one of the software’s characters) and
students were asked if the frog was going at the same speed all of the time,
speeding up or slowing down. Ellis found that Group 2 students, who used
quantitative relationships, could more easily make ratios from quantities and
generalize correctly about linear functions, than did Group 1 students who
focused on number-patterns.
45
Ellis‘s (2009) study points to the importance of quantitative reasoning for
Siebert (2002).
school students (two students in Grade 10, three students in Grade 11, and two
students in Grade 12), enrolled in Algebra II, and their teacher. Students and
their teacher were observed during 32 lessons over a ten-week period of time. A
own terminology for the y-intercept, and they used “starting point.” The term
“starting point” could be explained by the fact that most of the real-life situations
used during instruction were modeled with linear functions that had time as the
independent variable (e.g., paying an amount of money over a period of time; the
cost of a phone call of different durations). For the post-test, participants had to
define the “starting point” and translate between different representations of the
observations and interviews revealed that students’ successful use of the “starting
point,” suggested by the real-world contexts, did not translate into successful use
negative impact on student performance on abstract tasks, i.e., tasks without real-
46
world contexts. Of note, the teacher did not question his assumptions about
students’ meaning of the “starting point,” and was unaware of the influence of the
and representations.
lasted for a long period of time, in a real classroom and with both students and
their teacher. Davis’s results confirmed those of other studies showing that
hours duration over a period of seven weeks, with a 16-year-old student to explore
Schoenfeld et al. (1993) conducted the teaching experiment using the software
containing three microworlds. The software allowed the user to input a function in
symbolic form (e.g., y 2 x 5 ) and x-values. For each x-value, the software
calculated the corresponding y-value, explained why the pair (x, y) would be on
the graph of the function, entered the coordinate pair into a table of values, and
47
plotted the point on a graph. The software also allowed explorations of the roles
and the software accordingly rotated the graph of the function through the fixed y-
software accordingly moved up or down the graph of the function. Finally, the
student could play the Targets game on a Cartesian system of coordinates. In the
game of Targets, lines appear and the student had to “shoot” them by entering the
symbolic form of the equation of the graphed line. The student explored all of the
software’s tasks, but spent most of the time playing the game of matching a line
researcher observed the student interacting with the software and asked questions
about her understanding. Analysis of student answers showed that the student
(slope m, and y-intercept). However, there was limited and unstable change in the
y 2 y1
calculating the slope using the formula m , with no connection to the
x 2 x1
value. Gradually, during the teaching experiment, she could relate the sign of m to
the orientation of the line and the magnitude of m with the inclination of the
line, but without justifying it with the Cartesian connection or having any basis
for her rule. Until the end of the experiment, the slope formula remained a “black
box” for the student. Additionally, although at the beginning of the experiment
she knew that b was called the y-intercept, she did not connect it with the fact that
(1993) mentioned that these difficulties were shared by other participants in the
study. The student’s lack of connection between the slope formula and the
connection.
students in Grade 8, and with the same software, Grapher, Moschovich et al.
functions and the object view of linear functions. During the teaching experiment
49
2) Given a graph of a group of lines with the same y-intercept (or with the
same slope), students had to write the equation of one of the lines (object-
view).
linear functions. For example, given a line and a point not situated on the line,
students had to write the equation of the line parallel to the given line and passing
Both students had difficulties moving flexibly between the process and object
perspectives of linear functions. One of the students, who had not been previously
exposed to a linear functions curriculum, also had difficulties moving within the
process and object perspectives of linear functions. The researchers explained the
not be a linear one - from process to object with the object view reserved for the
problems.
50
The Physics Education Group at the University of Washington investigated
university, and high school students enrolled in Physics classes at local high
schools. Although the degree of severity varied, student difficulties were common
across all groups. For example, in one of the tasks, a graph (position versus time)
representing the linear motion with constant speed of two objects was presented in
the speeds of the two objects at the moment t 2 . Many of the participants who
indicated the object with the greater speed. Of great interest is that the same
conceptual error was noticed in another task, when the graph position versus time
of an object in motion was given, and the participants were asked to sketch the
graph of speed versus time. Researchers hypothesized that most of the student
errors and difficulties were due to the participants’ inability to interpret graphs,
i.e., to connect between the graphic representation and the situation it represents.
slope in one situation and their abilities to transfer their knowledge of slope to
51
other situations. Students measured the areas and the heights of various
rectangles, all with the same width. Then, they completed a table of values
plots and the best-fitting line. When asked for the numerical value and the
meaning of “slope of the line,” only six students identified it as the width of the
referring to slope. The same task was repeated in different situations (e.g., the
velocity and time). The students who participated in this teaching experiment
were more likely than contrast-group students to transfer their knowledge of slope
situations, are not in concert with the findings from other studies, as for example
Zaslavsky et al. (2002) cited previously, showed that pre-service and in-service
teaching experiments (Davis, 2007; Thompson & Thompson, 1994) revealed that
teachers are not always successful in attending to their students’ thinking. For
intercept (Section 2.4.2), the teacher did not recognize his students’ difficulties
52
connecting various representations of the y-intercept. He assumed that his
difficulties.
representations of slope and their pedagogical knowledge for teaching the concept
of slope. All but two of the in-service teachers held Bachelor degrees in
teachers planned to teach at the high school level, and all but one had teaching
experience as part of their training prior to the study. An initial survey was
administered to all participants. Additionally, from each of the in-service and the
pre-service teacher groups, eight teachers, four females and four males, were
randomly selected for interviews. The initial survey and interviews included
of slope.
53
context, i.e., slope as “rise over run” or slope as change in y over change in x.
The next most often mentioned context was slope as rate of change. More pre-
service teachers (about 50%) mentioned slope as rate of change than did
in-service teachers (about 40%). Almost 40% of the in-service teachers, but only
about 15% of the pre-service teachers, mentioned slope in physical contexts, as,
pre-service teachers, and by only one in-service teacher. None of the pre-service
was mentioned by about the same percentage (20%) of in-service and pre-service
teachers.
All teachers correctly calculated the rate of growth of a child height given two
sets of values (age, height) for the child, (9, 4'3' ' ) and (11, 4'9' ' ) (Stump, 1997).
Also, all teachers correctly calculated the slope given two sets of values (duration,
cost) for a phone call, (5 minutes, $5.30), (10 minutes, $10.80), and interpreted it
as rate of change. However, two of the in-service teachers considered that the cost
of the phone call per minute was not constant, it increased with time. Both in-
service and pre-service teachers had difficulty with slope as rate of change in
movement with constant speed. For example, in one of the problems taken from
of the two objects at the moment t 2 . Two thirds (67%) of the in-service
teachers and 71% of the pre-service teachers correctly solved the problem. Four
pre-service teachers and three in-service teachers considered that the higher
greater speed. Although m1 z m2 , and therefore the speeds are different, two pre-
service teachers and one in-service teacher considered that the two objects could
have the same speed at some moment in time. Two in-service teachers did not
respond. One first-year in-service teacher mentioned that the idea of slope as rate
of change was new to her. These findings are supported by the work of
All teachers could write the linear function (x = number of hours, y = number
of points for score) for the following situation: A student initially gets 50 points
on a test, and for each hour of study, her score increases by 3 points. However,
teachers had difficulty identifying the slope and the y-intercept in linear functions
in the symbolic form y mx b , when m and b were not numbers. In one of the
problems, the formula for the expansion of the length of a metal rod under
rod, t is the temperature, is the initial length, is the initial temperature, and a
is a constant for the type of metal. The task was to express the length of the rod as
55
teachers (81%) correctly wrote the formula in linear function form (or slope-
intercept form). However, less than two thirds of them (61% of the pre-service
teachers and 62% of the in-service teachers) correctly identified the slope. Only
33% of the pre-service teachers and 57% of the in-service teachers correctly
identified the y-intercept. Two pre-service teachers and three in-service teachers
Only one-third (33%) of the pre-service teachers and two-thirds (67%) of the
in-service teachers correctly calculated the slope of a line given a point on the
line, and the measure of the angle of the line with the positive sense of X-axis.
The teachers who could not calculate the slope of the line in a trigonometric
context often referred to the need for a second point on the line.
Stump’s (1997) interviews with teachers confirmed the survey findings. Both
teachers identified more connections for slope representations than did the pre-
service teachers. For example, the in-service teachers could connect between the
most often mentioned context for slope was the geometric-analytical one.
often mentioned the physical context of slope for instruction. Their instructional
56
practices, as mentioned during surveys and interviews, reflected only slope in
physical and geometric-analytical contexts, and lacked other contexts for slope as
(1997) pointed out that teachers did not question their students’ abilities to make
difficulties with graphing lines and computing slope, while the pre-service
understanding of slope.
used for the categories of conceptual knowledge of slope in the present study (see
Chapter 3).
to the start of the methods course, a survey was administered to gauge future
with slope. The methods course included activities and class discussions on
57
various contexts and representations for slope (e.g., geometric, analytical), and
interviews with one high school student and one college student about their
difficulties with slope, interview analyses, and an algebra textbook analysis that
focused on the topic of slope, as well as the preparation of lesson plans for
teaching the concept of slope to middle school and high school students. After
the course completion, five of those participants taught a basic algebra course at
the university. Lessons on slope taught to college students during the basic
algebra course were videotaped and analyzed. To account for the development of
compared with their knowledge developed during the methods course, and with
slope of the line y x ) and physical situations (e.g., inclined plane, roof). At the
(slope as rate of change), and graphic representations of lines and their symbolic
equations. Results for only three future teachers are reported. One future teacher
one emphasized the functional situations (slope as rate of change), while the third
used both physical and functional situations. However, only one of the future
mathematics teachers used real-life situations in problems, while the other two
58
referred to real-life situations only as illustrations. In short, teachers differed in
their emphasis on slope as rate of change and their use of real-life situations in
problems. Stump hypothesized that the limited use of real-life and functional
one of the future teachers thought that student difficulties with slope stemmed
from weak procedural knowledge (e.g., not knowing the formula for slope, how to
apply it or how to correctly perform calculations). During the two interviews with
teacher, who initially thought that students would know the meaning of slope,
discovered during the interviews that students did not know what slope was or
how to calculate it. She decided to develop lessons that introduced the concept of
difficulties with slope. Only one of the future teachers’ assumptions was
59
confirmed by student interviews, and reflected only one aspect of slope as a
with the other 10 prospective teachers. The questionnaires and the interviews
functions, namely that the slope of the line representing the function varies
proportionally with the measure of the angle between the line and the positive
neither units of measurements are given, nor are coordinates marked. Should the
coordinates), then the coordinates of A may be (1, 9) meaning that the point A is
in the first quadrant, very close to the Y-axis. The task was to figure out the
equation of the line passing through the point A and the origin O of the system of
60
Well, I can use the line y x as a reference line. The slope of line AO should
be twice the slope of the line y x , which is 1. So, the slope of line AO is
about 2, and the equation is about y 2 x , let’s say y 1.9 x . (Even & Tirosh,
1995, p. 13)
The prospective teachers were asked to comment about what the student had in
mind when she responded, and if her answer was correct. Only half of the
teachers correctly identified the student error and could describe the student’s way
of thinking: Since the line y x has the slope 1 and makes a 45° angle with the
positive sense of the X-axis, and AO makes almost a 90° angle with the positive
sense of the X-axis, almost twice as much as 45°, then the slope of AO should be
almost twice the slope of the line y x . Interestingly, some of the teachers who
noticed the student’s connection between the magnitude of slope and the measure
of the angle of inclination, could not correctly solve the problem. They considered
that the line AO, being so close to a vertical line, ought to have a large slope (e.g.,
100). Some counted “rise over run,” and found a slope of 10-15, which
contradicted their idea that the slope ought to be a very large number. The other
teachers, who did not notice the student’s connection between slope and angle,
considered that the student’s error was due to an incorrect estimation of slope.
Even and Tirosh (1995) pointed out that the relationship between mathematical
both types of knowledge are necessary for teaching. The researchers found that
student reasoning about linear functions in the problem described above, and
61
some of those teachers could not solve the problem themselves. The researchers’
taught students in Grades 8-10, the survey contained only problems focusing on
(N), and graphic (G) representations. Teachers were asked to identify the number
those in the survey. For the problem requiring connecting graphic and numerical
coordinates is shown and the task was to find an ordered pair that satisfies the
linear equation of the line. The teachers mentioned that they spent the least
amount of time during instruction, and used the fewest number of problems of
that type in their assessments of the connection G to N (mean 1.35 class periods,
mean 1.22 used in assessments). For the problem requiring connecting between
equation of a line in standard form was given and the task was to complete a table
of values. Teachers mentioned that they spent 1-2 class periods teaching the
connection A to N (mean 1.71 class periods) and used 2-3 problems requiring the
62
connection A to N in their assessments (mean 2.45 used in assessments). For the
homogeneous system of coordinates and the task was to find its symbolic
teaching the connection G to A (mean 2.21) and used 2-3 problems requiring the
connections because they assumed that their students would not encounter
student difficulties with the Cartesian connection (Smith et al., 1989; Schoenfeld
et al., 1993; Moschkovich et al., 1993; Knuth, 2000). Cunningham also pointed
out that algebra teachers assume that their students will not have difficulties
Reported here is only the part of the study focusing on teachers’ mathematical
Participants were 104 elementary and middle school pre-service teachers in the
63
last stage of their preparation and enrolled in methods of teaching
values representing the value (in dollars) of a car after a numbers of years since
purchase: (0, 16800), (2, 13600), (4, 10400), (6, 7200). Teachers were asked to
write the linear function representing the depreciation of the car’s value over time.
Ninety-eight teachers did not write the correct linear function. Thirty-two teachers
did not properly attend to the change in time, i.e., did not notice that the time
interval was 2 years, and wrote V (t ) 3200t 16800 . In another problem, taken
from Knuth’s (2000) study (Section 2.4.1), only 27 teachers gave the correct
answer, and 20 used an algebraic approach. The problem required finding the
equation of a line, given the graphic representation with two lattice points, (0,2)
functions. The pre-service teachers did not successfully recognize student errors,
You’s (2006) study results also confirmed Stump’s (1997, 2001b), Even and
their strengths and weaknesses and subsequently take greater responsibility for
their own learning. The following research studies refer to measures of problem
difficulty are presented first followed by studies comparing student and teacher
problem difficulty.
education courses at university in Hong Kong. Two mathematics tests, with a total
students. Students also were asked to rate the degree of difficulty of each
problem on a scale from 1 to 5 (1-very easy, 5-very hard). Teachers were not
asked to solve the problems. Instead, they were asked to assess the level of
65
difficulty of each problem and identify the factors that might influence the level
Using a regression framework, Lee and Heyworth identified the factors predicting
the three measures of problem difficulty given by: 1) student perceived difficulty,
answers and the total number of answers, and 3) teacher assessment of problem
following four factors: 1) perceived number of difficult steps required to solve the
with the type of problem (e.g., logarithmic equations that can be brought in the
familiarity with the problem, the number of “log” occurrences in the problem, and
the numerical complexity of the problem. The difference between the factors
teachers were not asked to solve the problems. Consequently, teachers were more
likely to judge the level of problem difficulty using surface features of the
that student observations were not independent, as students came from the same
two schools, and most likely, some of them were enrolled in the same
mathematics classes, taught by the same teachers. Although the p-values reported
are small, and the results about the factors proposed as predictors for different
(Jerman, 1973; Silver & Thompson, 1984; Lane, 1991; Mason, Zollman,
Bramble, & O'Brien, 1992), more research is needed to develop a measure for
the Lee and Hayworth’s study is important because it points out that students are
better able to predict their performance than are their mathematics teachers.
Biochemistry or Physics. About 400 students, 200 from each of the Biochemistry
and Physics courses, participated in this study. Students were given an on-line test
information; “Credit” - for items requiring what a “Pass” question requires as well
a higher level of analysis and critical thinking). Students answered the questions,
67
and assigned each question a level of difficulty (Pass, Credit, and
difficulty. Overall, the more difficult questions were associated with lower student
and faculty agreement about difficulty levels, and longer response. For
about the same for all questions. For Physics students, student and faculty
agreement was 53% for “Pass” questions that were correctly answered by 71% of
questions that were correctly answered by only 47% of students). The association
statistically significant only for Physics students. Physics students and faculty
agreement of problem difficulty was significantly higher for those students who
took Physics in high school. Overall, compared with the level of problem
identified a “Pass“ question were more likely to answer it correctly. Students who
answer it incorrectly.
68
Lingard et al.’s (2009) study is important because it draws attention to the low
Gire and Rebello (2010) conducted two studies to investigate how factors,
such as problem complexity, and familiarity with the type of problem, are
both studies were college students enrolled in introductory Physics courses and
participated in the first study. Participants were asked to rate, on a scale from 1 to
In the second study by Gire and Rebello (2010), students (N=21) were enrolled
problems, as for example: The position of an abject moving along the X-axis is
given by the polynomial function x(t ) 3t 3 2t 4 ; find at least one time when
69
the velocity is zero. Subsequently, students rated the level of familiarity with the
type of problem on a scale from1 to 5 (1-most familiar, 5-least familiar), and the
difficult). Instructors (N=16) of the students were asked to rate the level of
In the first study, with one exception, students rated five of the 16 problems as
being easier than their instructors considered them to be. In the second study,
students rated four of the 10 problems as being easier than their instructors’
student ratings of familiarity with problems correlated highly with student and
with low ratings of problem difficulty; low ratings of problem difficulty were
associated with high ratings of problem familiarity. In other words, students who
correctly solved a problem were more likely to rate it as familiar and easy. In the
problem difficulty, but not with student rating of problem difficulty. In the
second study, problem complexity correlated highly with both student and
that the problem domain (Work-Energy and Kinematics) and the place where the
70
Calculus-based Physics, algebra-based Physics) may play important roles in
Zhou (2009) used teacher assessment of relative difficulty of problems and the
factors associated with those assessments, and compared them to the relative
in the Southwest. All teachers had more than four years of teaching experience.
Only two teachers held degrees in mathematics. Three of the 10 teachers had
experience teaching only mathematics (Grades 7-9); the other seven teachers
direct reading of a ruler, reading a ruler from starting positions other than zero,
unit iteration, constant units, unit conversions, and unit partition to different
fractional parts). They were asked to rank problems in order of difficulty from
their students’ perspectives. The test was also administered to elementary and
used to calculate the Spearman rank-correlations for each teacher. Only three
those three teachers taught all subjects, and one teacher taught only middle-school
mathematics. The latter teacher had the least number of years of teaching
71
experience. Overall, teachers were more likely to accurately estimate the relative
ranking of problem difficulty for easier problems, i.e., problems most students
could solve. Mathematical knowledge was not found to be associated with higher
problems. One of the two teachers with mathematics degrees overestimated easier
problems, and the other underestimated the difficulty levels of problems. Neither
the study. This raises the issue of curriculum experiences that were more likely to
be different for the two student populations, and therefore might invalidate the
difficulties.
Nathan and Koedinger conducted studies (Koedinger & Nathan, 2004; Nathan
72
problem difficulty (given by its rank order in a set of problems) with student
performance. Participants were Grade 9 students from four U.S. urban high
“Starting with some number, if I multiply it by 6 and then add 66, I get
81.90. What did I start with?” (Nathan & Koedinger, 2000, p. 170).
When Ted got home from his waiter job, he multiplied his hourly wage by
the 6 hours he worked that day. Then he added the $66 he made in tips and
found he earned $81.90. How much per hour did Ted make? (Nathan &
73
Based on student performance, “start-unknown” problems were more difficult
were given 12 problems, two problems for each of the six types, and asked to rank
them in the order of difficulty for their students, from easiest to most difficult. As
participating students were grouped within the same classes and schools.
Moreover, students were not asked about their perceptions of problem difficulty,
problem difficulty was made by a group of teachers different from those teaching
74
Lorge and Kruglov (1953) hypothesized that information about the absolute
difficulty of a subset of problems from a test, i.e., the percentage of students with
correct answers for each of the problems in the subset, would lead to improved
teacher assessment of the percentage of students with correct answers for all
this study. Two 45-problem tests, on mathematics topics taken from Grade 8 and
9 mathematics textbooks, were used for this study. The empirical difficulty of
that correctly solved the problem, was determined by administering the test
were informed about the percentage of students with correct answers for 10 of the
problems on one of the tests, and they were asked to predict the percentage of
students with correct answers for the rest of the questions in the tests. Although
75
2.7 Concluding Remarks
Solving equations was the purpose of algebra until the nineteenth century.
With the exception of the New Math era, the Algebra I curriculum was dominated
by solving equations using symbolic manipulations and formulas. After 1990, the
phenomena.
functions revealed that students had difficulty with slope or rate of change, and
line, generates more student difficulties than the analytical perspective of slope,
that is, slope as rate of change. In the geometric perspective of slope, students
of functions, most notably, the Cartesian connection, i.e., connecting between the
symbolic and the graphic representation of a linear function. Also, students have
76
for slope). Although with various degrees of intensity, research studies reveal that
point out that the same difficulties also are experienced by pre-service and in-
student difficulties with linearity and linear functions shows that teachers’ weak
their weak ability to understand their students’ thinking and attend to their
somewhat richer, the various comparison types and research objectives make it
difficult to use. The results of those studies indicate that the level of agreement
experiences, type of learners, subject matter of the problem, and content and type
of problem.
77
and linear functions, together with research on teacher understanding of student
difficulties with aspects of linearity and linear functions, and research on student
and teacher assessment of problem difficulty. In the next chapter, Chapter 3, data
78
Chapter 3
3. 1 Introduction
The purpose of this study was to identify student difficulties with aspects of
nature of the difficulties students experience. To gain insight into students’ and
developed particularly for this study, and then implemented to answer the
following questions:
a. Grade level?
c. Problem?
79
The following are described in this chapter: the instruments used to collect
data to answer the research questions, the pilot study designed to refine
instruments and develop the Preliminary Coding Categories, the nature of the
sample, the timeline for administration of the assessments, and the methods used
to analyze the data for each research question. Assessment instruments are
described in Section 3.2. The theoretical framework used for analysis is presented
in Section 3.3. The pilot study, the Scoring Guide and Preliminary Coding
students and their mathematics teachers, are described in Section 3.5. The
tests and coding interviews, and analyzing the data, is found in Section 3.6.
Section 3.7.
The following instruments were used to collect data for this study: the Mini-
Questionnaire for Students (RQS) and the Ranking Questionnaire for Teachers
Questionnaire for Teachers (IQT), and the Interview questions for students and
teachers.
linear functions is the Mini-Diagnostic Test (MDT). The MDT was developed by
80
Greenes et al. (2007b), a committee of mathematicians, mathematics educators,
and school district mathematics leaders, after analyzing five years (2001-2005) of
students’ performance on linearity and linear functions items on the Grade 8 test
The MDT, consisting of seven problems (the last problem with three sub-
problems), uses three formats: multiple choice (problems 3, 5, and 6), short
answer (problems 2, 4, and 7a, 7b, 7c) and extended response (problem 1). In the
directions, participants are encouraged to show their solution methods along with
their answers to the problems. MDT problem 1 gives the coordinates of a point
and the equation of a line. The task is to determine whether the point is contained
in the line, and to explain how the decision was made. MDT problem 2 requires
linear function, and the task is to identify the corresponding graph from a group of
four graphs. MDT problem 4 requires determination of the slope of a line from its
graph. MDT problem 5 provides a table of values and five choices of linear
functions presented symbolically. The task is to match the table of values with its
with three sub-problems or tasks. The tasks are to identify that part of the graph
when the car is traveling the slowest, determine the slope of that part of the graph,
and determine the speed of the car corresponding to that part of the graph. For
81
analysis purposes, each sub-problem of problem 7 was considered as a separate
reproduced in Appendix A.
After completing the MDT, all participants were asked to rank the MDT
developed particularly for this study (see Appendix A). The Ranking
Questionnaire for Students (RQS) asked them to order the MDT problems from
most to least difficult, and comment on the nature of the difficulty for the most
difficult problem. The Ranking Questionnaire for Teachers (RQT) asked them to
order the MDT problems from most to least difficult as perceived by their
information about gender, grade level in school, current mathematics course, and
requests highest academic degree, college major, current and previously taught
82
completion. The total time required for completion of the Information
minutes.
difficulties with MDT linearity concepts, interviews were developed for use with
did not interrupt the solution process but gave encouragement to “continue,” or
“tell more.” Once a problem was solved, probing questions were used to elicit
rationales about the choice of solution strategies or steps, or to satisfy the need for
2. How did you solve this problem? What did you do first? What did you do
next?
3. Is there any other information you would have liked to have to solve this
problem?
5. Why did you rank problem (name problem that was ranked most difficult
83
Semi-structured interviews were used with teachers, as well. All teachers
interviewed for this study were asked to provide rationales for their selection of
problems that would pose the greatest difficulty for their students, identify the
strategies to overcome those difficulties. Teachers were asked questions like the
following:
1. You chose problem (name the problem the teacher chose) as posing the
2. Problem (name the problem that caused difficulty) has been difficult for
many students.
84
situations, schemes for dealing with situations, sets of concepts and theorems
speed as in MDT problem 7). Some situations have real-life contexts (e.g.,
MDT problems 5 and 7), while others have mathematical contexts (e.g., MDT
to solve problems (e.g., MDT problems 1 and 2). Some schemes of action,
used to correctly solve a problem, are in fact algorithms for solving the
satisfies the equation of the line, the solver needs to substitute the coordinates
of the given point into the equation of the line, y mx b , and check whether
MDT problems 7a and 7c), and ways of thinking (e.g., quantitative reasoning).
85
4. Representations of linear functions (e.g., informal, natural language, or
rule in MDT 1, table of values and graphic representation in MDT problem 3).
The primary purpose of the pilot study was to gain insight into the nature of
students’ difficulties, in order to adapt the Scoring Guide (SG), and develop
Preliminary Coding Categories (PCC) for the RQS, RQT and Interviews.
Participants in the pilot study were 213 Algebra I students in Grades 8-10 and
hundred and six students were in Grade 9, and only eight students were in Grade
10. All Grade 8 students were enrolled in the same middle school. All Grade 9
and 10 students were enrolled in the same high school. Both schools were in the
same school district that is primarily middle class (less than 25% of students
eligible for free or reduced-price lunch). The MDT and Ranking Questionnaires
(RQ) were administered to all participants. Four teachers and 77 students agreed
(five students in Grade 8, four students in Grade 9, and one student in Grade 10).
Interviewed students were selected by the researcher, such that half of them had
MDT scores above the median, and half had MDT scores below the median.
86
Student MDTs were scored using the Scoring Guide. The Scoring Guide is
3.4.1 Scoring Guide (SG) for the MDT. The Scoring Guide (SG) (see
(1, 2, 3, 4, 5, 6, and 7) of the MDT was scored 20 points. The PSG modifications
correlations between the MDT problems led to considering nine MDT problems
(1, 2, 3, 4, 5, 6, 7a, 7b, 7c), rather than seven. Six of the MDT problems (3, 4, 5,
6, 7a, 7b) have dichotomous scores. The initial dichotomous scores 0 or 20 were
identification of outliers signaled by the presence of means greater than .9. It was
expected that students in advanced mathematics classes, like Algebra II, had
mastered the MDT aspects of linearity and linear functions. Thus, the MDT would
be easy for them, and consequently their MDT scores would be high (> .9),
87
Also, the Preliminary Scoring Guide was modified for use in this study to
'y
b. Eliminating the type of answer “Computing with incorrect
'x
coordinates;” and
The modified Scoring Guide, described below, will be referred to, from now
on, as the Scoring Guide (SG). Table 1 presents score possibilities for the nine
MDT problems.
88
Table 1
1 2 3 4 5 6 a b c
1 .6, .8 or 1 1 1 or 1
2 is scored 0, .2, .4, .6, .8 or 1, based on the number of correctly completed rows
(0 to 5) in the table of values. MDT problem 1 is separated into two parts, part 1a
scored 1 if the answer is “Yes,” and 0 otherwise, and part 1b is scored 1 for a
correct explanation, and 0 otherwise. The total score for MDT problem 1 is
obtained by adding the scores for parts 1a and 1b and dividing the sum by 2.
MDT problem 7c is separated into two parts, 7c1 and 7c2, one part for the
numeric value and the other for the unit, with 1 point for each correct answer and
0 otherwise. The total score for MDT problem 7c is obtained by adding the two
scores for parts 7c1 and 7c2 and dividing the sum by 2 to produce an average
score for that problem. The high correlations between the pairs of scores for the
parts 1a and 1b (.93), and parts 7c1 and 7c2 (.80), respectively, explain the
need for collapsing the two-parts scores into single scores for MDT problems 1
and 7c, respectively. Possible scores for MDT problems 1 and 7c are 0, .5, or 1.
For each of the MDT problems, 1 through 6, 7a, 7b, and 7c, scores range from 0
89
through 1. The MDT total score is obtained by adding scores for all problems 1,
Types of responses for MDT problems 1, 2, 4, 7a, 7b, and 7c with short
answers or extended responses are presented in Tables 2-7. Table 2 presents types
of answers for MDT problem 1 that requires determination if a given point (x, y)
is situated on a line.
Table 2
(b), #1-4
90
As can be seen from Table 2, the type of response for MDT problem 1a is
of the participants in this present study two problem solving strategies used by
strategies are those of the Cartesian connection and the use of a graph, and they
imply two different algorithms for solving MDT problem 1. The Cartesian
graph implies that the line y 3x 14 is graphed, and the solver determines that
the point (2,8) lies on the graphed line. Scoring MDT problem 1b requires
determination of the correctness of the answer and the solution strategy. Six non-
overlapping coding categories were identified during the coding process: correct
answer using the Cartesian Connection, incorrect answer using the Cartesian
Connection, correct answer using a graph, incorrect answer using a graph, and no
participants to construct five rows of a table of values, given the linear function
y 2 x 3 .
91
Table 3
7. No answer
answer with no success from those who did not attempt. This differentiation was
necessary to separate out those who do not answer correctly due to computational
coordinates.
92
Table 4
3. Reversed ratio
5. Irrelevant information
Types of incorrect answers for MDT problem 4 as shown in Table 4 are based
order, that is, change in x over change in y). The Type 5 code (“Irrelevant
Type 6 code is used when no answer is given, or a statement is made to the effect
Table 5 presents types of responses for MDT problem 7a. That problem
requires identification of the part of the graph when a car moves the slowest.
93
Table 5
1. Answer “R”
2. Answer “S”
5. No answer
As can be seen from Table 5, types of responses for MDT problem 7a are
grouped into five disjoint categories (“R,” “S,” “T,” an answer different from
Table 6 presents the types of responses for MDT problem 7b, that requires
system of coordinates.
Table 6
2. Incorrect answer
As can be seen in Table 6, there are three types of responses. All incorrect
answers resulting from attempts to calculate the slope are grouped together under
94
“Incorrect answer.” Other types of answers that present irrelevant information or
Table 7 presents the types of responses for MDT problem 7c that requires
Table 7
3. Incorrect answer
As can be seen in Table 7, when the numeric part of the answer is given and
either the measurement unit is incorrect or not given, the response is considered to
be of Type 2. When an attempt to calculate the speed is evident but the answer is
descriptions were short, usually only one sentence, and fairly repetitive. Teachers’
descriptions were also short, mostly one or two sentences in length. The
95
researcher read each comment and identified the major idea. For comparison
purposes, both students’ and teachers’ written comments were coded using the
presented in Table 8.
Table 8
No Comment 70 (32.9%)
about general problem difficulty using expressions like, “It’s a hard problem.”
96
Category M (“Mathematical difficulty”) was used when specific difficulties with
used when students made comments about forgetting (“I don’t remember”).
understand it.” Category W (“Whole Domain”) was used with comments about
whole domains of mathematics, like algebra, or mathematics (e.g., “I’m not good
used for difficulties that did not fit into the other categories, like references to
difficulty with vocabulary or lack of sufficient time to solve the problem. Of note,
about half of the students (45%) made comments of “Not Understanding,” about
one third of students (32.9%) did not make any comments, and frequencies of
problems for pilot study students were MDT problems 7b and 4, both requiring
identified with its coordinates is on a line, given the equation of the line. Only
28% of pilot study students solved MDT problem 7b correctly, 33% solved MDT
the interviews focused on student knowledge and difficulties with the Cartesian
Pilot study data, along with results of research studies (Lobato & Siebert,
97
Stump 1997, 2001a, 2001b; Zaslavsky et al., 2002), were used to develop
coding at the level of paragraph (Strauss & Corbin, 1998). Answers to the
not address the question were excluded from analysis. Tables 9 and 10 present the
Table 9
Connection
participants referred to the mathematical theorem, but rather, to the algorithm for
solving the problem. Three of the interviewed students missed the Cartesian
knowledge of slope.
98
Table 10
As can be seen in Table 10, there are five contexts (physical, geometric-
interviewed participants from the pilot study referred to slope mainly in physical
and geometric-analytical contexts. There was only one reference to slope in each
99
of the other contexts. One student mentioned slope as “m in y mx b ,” one
difficulties with the Cartesian connection and slope in MDT problems 1, 4 and 7b.
problem 1.
Table 11
Connection
Difficulty
statement. The Cartesian connection from point to line was observed in students’
occurred at one or more of the steps of the algorithm: substitution (one student),
or verification that the statement is true (one student). Three of the interviewed
teachers did not identify MDT problem 1 as being difficult for their students. The
fourth teacher suggested that MDT problem 1 might be difficult for students
Table 12 shows the Preliminary Coding Categories for student difficulties with
slope in the contexts of MDT problems 4 and 7b. Since the only students having
difficulty with slope were students with geometric perspectives of slope, i.e.,
oblique line using “slope triangles,” i.e., triangles with two vertices being points
( x 2 , y1 ) . When using a “slope triangle” to measure the rise and the run, the rise
and the run are oriented segments with positive or negative magnitudes
(Schoenfeld et al., 1993). Coordinating the unit of measurement with the length
101
point, or origin of measurement, and then imagining or concretely laying down as
cover the entire line segment to be measured (“exact” measurement, when the line
of the unit of measurement is rigid during the measurement process, i.e., the unit
coordinates the unit of measurement is the same for both the X-axis and the Y-
for the X-axis and the Y-axis are different. All interviewed students who
attribute of the linear function represented graphically by a line, did not encounter
difficulties. Those students calculated the slope of the line using the following
y 2 y1
on the line, and 2) calculate m . The importance of holding an analytical
x2 x1
102
Table 12
difficulty with slope used the geometric context for slope (“rise over run”). The
first interviewed student with slope difficulties, could identify the rise but
considered the run to be zero, and could not calculate the slope as “rise over run”
because she knew she could not divide by zero (MDT problem 4). The second
student incorrectly measured the run (1.5 units), while the third student correctly
103
calculated the slope in MDT problem 4 using “rise over run” (m 3) , but
4
incorrectly in MDT problem 7b ( m , instead of m 40 ). None of the teachers
3
identified MDT problems 4 as being difficult for students. Three teachers referred
to MDT problem 7 (7a, 7b, and 7c) as being difficult because of the presence of
the graph with three parts, making it difficult to interpret. One teacher pointed out
Pilot-study students had difficulty determining the rise (change in y), run
(change in x), and the ratio (slope). Their difficulties were observed as “schemes
the axes of coordinates). None of the pilot-study teachers identified the nature of
3.5 Participants
mathematics teachers from two school districts in a state in the Southwest U.S. Of
note, data analyses were conducted with 1561 students and 26 teachers.
Several characteristics of the participants and their school districts are presented
in this section. District level data were collected from the districts’ official
websites at the time of data collection. Data at the school and individual level
104
were collected from the Information Questionnaires completed by students and
teachers.
Table 13
As can be seen in Table 13, the majority of the population in both districts is
status. The percentage of students eligible for free or reduced lunch is 29% in
by district and by school (middle school, Grades 6 – 8, and high school, Grades 9
– 12) are presented in Table 14. Note that middle schools are included because
105
Table 14
From Table 14 it can be seen that there were twice as many participating
students from District 2 (N=1043) than from District 1 (N=512), that is, almost
two thirds of all students in this study (66.6%) were from District 2. About half of
the students from each district (259 from District 1, and 494 from District 2) were
8th graders in middle school, and half were 9th and 10th graders in high school (263
from District 1, and 549 from District 2). Data collected about gender revealed an
Table 15.
106
Table 15
Grade 8 9 10
Algebra II 781 78
0
Note. identifies Group 0 students; 1 identifies Group 1 students.
As can be seen from Table 15, there are two main groups of students:
Geometry (N=306) .
108
As can be seen in Table 16, seven District 1 (Middle School A and High
School B) teachers (Teachers 21-27) and 20 District 2 (Middle School C and High
participants were enrolled in courses ranging from Pre-Algebra (Teachers 1-4, 26)
to Algebra I (Teachers 5-7, 25, 27) and Geometry (Teacher 27). Some teachers,
like Teachers 8, 11, 12, 15, 27, taught both Algebra I and Geometry. Since
Teacher 13 had few students who participated in this study, three in Geometry and
one in Algebra II, data on those participants were eliminated from all statistical
analyses. Consequently, data analyses were conducted with 1561 students and 26
teachers.
study.
109
Table 17
From Table 17 it can be seen that, with the exception of District 1 Middle
School, the majority of teachers participating in this study had less than nine years
with Master degrees than District 1, both at the middle school and high school
levels, and more inexperienced teachers, with 29% of middle school teachers and
46% of high school teachers having less than four years of teaching experience.
Only one-third of the middle school teachers and three fourths of the high
were taught primarily by more experienced teachers with higher academic degrees
110
3.6 Timeline for Instrument Development, Recruitment of Participants, Data
The pilot study, development of the instruments and interview protocols, the
recruitment of participants, the data collection and the data analyses were
principals, department chairs, teachers, students, and their parents. There were
multiple site visits, discussions, phone calls, and e-mails. The recruitment
Table 18
June 2008- November 2008 Pilot study conducted and data analyzed
Development of Preliminary Coding Categories
October 2008–December 2008 Adaptation of the Scoring Guide for the Mini-
Diagnostic Test, Development of the instruments
RQS, RQT, IQS, IQT, Interview Questions
January 2009 –February 2009 Administration of the MDT, RQS, RQT, IQS,
IQT, Scoring MDTs
111
As can be seen in Table 18, the conduct of this study began in June 2008 and
The teachers who agreed to participate in this study invited their students to
participate, as well. Student participation was meritorious; overall, more than 70%
linearity and linear functions was distributed to all Grades 8, 9, and 10 students
and their mathematics teachers participating in this study, during January and
teachers who were instructed by the researcher about how to administer the test,
and obtained their verbal agreement to observe test-like conditions. The MDT was
teachers solved the MDT problems at their convenience, and all were done within
six weeks of student completion of the MDT. The Information Questionnaires and
Ranking Questionnaires were administered at the same time as the MDT. The
Mini-Diagnostic Tests were scored by the researcher using the Scoring Guide.
The written comments from Ranking Questionnaires were coded using the Final
Coding Categories (see Appendix B), and inter-rater agreement analyses of the
Interviews with participants were conducted within six weeks from the time of
the administration of the MDT. Twenty teachers from District 2, seven middle
their responses to the MDT, two students from each of the 20 teachers were
selected by the researcher and invited for interviews. In each pair of students, one
112
student had the MDT total score above the group median, and one student had the
MDT total score below the group median. Interviews were conducted by the
researcher during school time, were audio-taped, and were of 20 minutes duration.
All interviews were transcribed after the data were reduced and coded by the
researcher and another rater, using the Final Coding Categories developed for this
study (see Appendix B). Inter-rater agreement analyses of the coding were
performed.
In this section, the methods of analyses used to answer Research Questions 1-5
Research Question 1:
a. Grade level?
c. Problem?
Data were analyzed using descriptive statistics, cluster analysis and factor
standard deviations) were used. The analysis of MDT scores using descriptive
statistics led to the consideration of three possible natural clusters: by grade level,
(EFA), R-type, using the Principal Component Analysis method (Hair, Anderson,
Tatham, & Black, 1998) was employed. The R-type EFA “analyzes relationships
(factors)” (Hair at al., 1998, p. 90). An EFA using the Principal Component
Analysis method is a “factor model in which the factors are based on the total
variance.” It “implies that all variance is common or shared” (Hair et al., 1988, p.
88). The Principal Component Analysis method was preferred because the main
goal of EFA was to reduce the original set of MDT problems to a new set, with
For Research Questions 2 and 3, to assess the association between the ranking
of MDT problems by their mean scores and the ranking of MDT problems by
114
probability of concordance minus the probability of discordance” (Daniel, 1990,
p. 365).
from the Research Questionnaires were coded and analyzed, and semi-structured
interviews (Clement, 2000; Goldin, 1999) were used with both students and
teachers. The choice of the “think-aloud” procedure was justified by: 1) the need
they occur. Open coding techniques and procedures described by Strauss and
Corbin (1998) were used to code students’ and teachers’ answers from Research
refined and developed into Final Coding Categories (see Appendix B), and raters
were trained and retrained until satisfactory inter-rater agreement was achieved
particularly for use in the study were described along with their scoring, and
115
included: the MDT, the main instrument for measuring student performance on
linearity and linear functions, the accompanying Scoring Guide for scoring the
Characteristics of the sample and the general population were presented, together
with the timeline for the design and administration of the instruments, and data
questions concluded the chapter. In the next chapter, Chapter 4, data analyses and
116
Chapter 4
DATA ANALYSES
4.1 Introduction
The goals of this study were to gain insight into Grades 8, 9, and 10 students’
1,561 students (757 Grade 8 students, 416 Grade 9 students, and 388 Grade 10
students), and their 26 current mathematics teachers. Three assessment tools were
consists of nine problems (seven settings; one problem has three parts) focusing
Students (RQS) requires the ordering of MDT problems from most to least
difficult with a description of the nature of the difficulty for the problem
requests the ordering of MDT problems by difficulty from the perspective of their
students with descriptions about the nature of the difficulties for the two problems
identified as most and next most difficult. After the IQ, MDT, and RQ were
117
administered, 40 students and 20 teachers were interviewed. Data were collected
by:
a. Grade level?
c. Problem?
In this chapter the data and methods of analyses employed to answer the
questions and results of analyses are reported by research question. For Research
118
4.2 Research Question 1
a. Grade level?
c. Problem?
(N=1561) were scored by the researcher using the Scoring Guide (see Tables 1-7
in Chapter 3, and Appendix B). To establish the reliability of the Scoring Guide,
were scored by a second rater. An inter-rater agreement analysis using the Kappa
statistic (Cohen, 1960) was performed to determine consistency among raters for
the scoring of the types of student answers to the MDT problems with short or
extended answers (problems 1a, 1b, 2, 4, 7a, 7b, and 7c). MDT problems 3, 5, and
evaluation.
independent.
119
ii) The categories represented by student difficulties with MDT problems
The inter-rater reliability of scoring was measured using Cohen’s Kappa (k)
agreement:
p0 pc
k ,
1 pc
where p 0 = the proportion of units in which the raters agreed, and p c = the
The inter-rater agreement for the two raters is presented in Table 19.
Table 19
120
As can be seen in Table 19, the inter-rater agreement for MDT problems 2, 7a,
7b, and 7c is perfect (k = 1). This was expected, given the fact that the types of
answers for MDT problem 2 are based on the number of rows in the table that are
completed correctly, and all “Incorrect answers” are grouped together for MDT
problems 7a, 7b, and 7c. For MDT sub-problems 1a, 1b, and MDT problem 4, in
which incorrect responses are separated into several types, the inter-rater
agreement is high (k > .82). The lowest value of inter-rater agreement is for sub-
For data collected for this study (N=1561), the reliability of the MDT, seen as
internal consistency of the test, was measured using Cronbach’s Alpha. The
coefficient Alpha is a measure of the homogeneity of the test, i.e., if the MDT
knowledge of linearity and linear functions. The value of Alpha was .76, greater
than .7, which generally is accepted as cut-off value in the exploratory phase of
Data were analyzed using descriptive statistics, including cluster analysis and
factor analysis. A regression framework was not possible because of the nested
structure of the data, i.e., student observations nested within teachers, and thus the
linear model was not applied because of insufficient data at Level 1 (i.e., student
demographics).
121
4.2.2 Student Performance on the MDT by Grade Level. To see how
student performance on the MDT varies by grade level, the student MDT total
scores were analyzed by computing means and standard deviations by grade level.
Table 20 shows the means and standard deviations for the MDT scores by grade
level. The MDT has nine problems. Each problem has scores between 0 and 1,
thus the MDT total score ranges from 0 through 9. To better illustrate the student
Table 20
students were in Grade 8 and about one-fourth (26.6%, 24.9%) in each of the
other grades, 9 and 10 respectively. The means for the MDT score varied little by
grade level, ranging from 4.78 for Grade 8 students to 5.45 for Grade 10 students.
Overall, the mean for the MDT was 5.06. Regardless of grade level, students
122
correctly solved five of the nine MDT problems, on average. The standard
123
Figure 1b. Histogram of MDT Scores Distribution for Grade 9 Students
124
As can be seen in the histograms, MDT scores ranged from 0 through 9 with a
mean of about 5 for all grades. For Grade 8 (Figure 1a), the mode was 6. For
Grade 9 (Figure 1b) and Grade 10 (Figure 1c), the mode was 9, the maximum
score possible. It is easily noticeable in Figures 1a, 1b, and 1c that the
distributions are asymmetric due to the greater number of higher scores. More
Table 21 presents the means and standard deviations of MDT scores by current
Table 21
As can be seen in Table 21, there were about the same number of students
students in Grade 11, had the lowest enrollment (N=77). Pre-Algebra students
scored lowest on the MDT, correctly solving, on average, about four problems.
Mean scores for Algebra I and Geometry students were about the same; on
average, those students correctly solved about five problems. The standard
deviation for Pre-Algebra, Algebra I, and Geometry was about the same. Two
thirds of Algebra II students correctly solved all nine problems. The standard
deviation of the MDT scores for Algebra II students was less than 1, the least for
Figures 2a, 2b, 2c, and 2d present histograms of the MDT score distributions
respectively.
126
Figure 2a. Histogram of MDT Scores Distribution for Pre-Algebra Students
127
Figure 2c. Histogram of MDT Scores Distribution for Geometry Students
128
Figures 2a, 2b, 2c, and 2d show that the MDT score distributions are
symmetric for Pre-Algebra students, and become increasingly skewed for students
enrolled in higher level courses, Algebra I, Geometry, and Algebra II. Higher
students’ performance on the MDT was central to this study, since Algebra I is
the course in which students are expected to master aspects of linearity and linear
see how Algebra I students’ performance on the MDT varies by grade, Table 22
presents the means and standard deviations for MDT scores by grade level for
students enrolled in Algebra I. The second column of the table shows the number
Algebra I.
Table 22
129
As can be seen in Table 22, about half of the Algebra I students who
completed the MDT were in Grade 8 (49.7%) and half were in Grade 9 (48.9%).
students correctly solved about six MDT problems; Grade 9 students correctly
solved, on average, about four MDT problems; and Grade 10 students correctly
grades) correctly solved about five MDT problems. Standard deviations were
scores: by grade level, current mathematics course, and performance group. There
below their grade level. Group1 consisted of students enrolled in more advanced
mathematics courses, that is, courses above their grade level. A cluster analysis
revealed that MDT scores were better grouped (more cohesion, less separation) by
average silhouette criterion (Rousseeuw, 1987) was used for determining the best
cluster. The largest overall average silhouette indicates the best clustering
(Rousseeuw, 1987).
Figures 3a, 3b, and 3c illustrate a comparison of the quality of clusters, using
the average silhouette as a criterion for determining the best cluster from three
130
possible natural clusters, by grade, current mathematics course, and performance
group, respectively.
Clustering by Grade
131
As can be seen in Figures 3a, 3b, and 3c, the average silhouette for the three
clusters by grade, current mathematics course, and performance group are .6, .5,
and .7, respectively. The best cluster is the last one, by performance group, with a
Table 23 presents the means and standard deviations of the MDT scores by
performance level. The first column of the table shows the number and
Table 23
Group 0 students correctly solved about four MDT problems with a standard
total number of students, correctly solved about seven MDT problems with a
132
Figures 4a and 4b show histograms of the MDT score distribution by
133
Figures 4a and 4b show that the MDT score distribution for Group 0 is fairly
symmetric, while the one for Group 1 is heavily skewed due to the presence of
the variation of student performance on the MDT by problem, the nine MDT
type, using the Principal Component Analysis method (Hair et al., 1998).
i) Reduce the original set of variables (problem scores) to a new set with a
ii) Detect the structure of relationships among the variables representing the
After the data were reduced, the new reduced data were used in all subsequent
Data Types and the Choice of the Statistical Software. The variables
representing the scores for the MDT problems can be considered ordinal under the
student performance. All variables representing the scores for the MDT problems
134
scores for MDT problems 3, 4, 5, 6, 7a, and 7b have two values (0 or 1). The
variables representing the scores for MDT problems 1 and 7c have three values
(0, .5, or 1). The variable representing the score for problem 2 has six values (0,
.2, .4, .6, .8, or 1). There is growing consensus that the association between
between two unobserved continuous variables given only observed ordinal data”
(Flora & Curran, 2004, p. 467). The tetrachoric correlation is a particular case of
the polychoric correlation between two dichotomous variables. “[It] estimates the
Pearson correlation one would obtain if the two constructs were measured
About the Sample. The sample size is large (N=1561), suitable for EFA
Table 24 shows the means and standard deviations for the scores for MDT
135
Table 24
1 .51 .46
2 .63 .45
3 .78 .41
4 .35 .48
5 .81 .39
6 .51 .50
7a .59 .49
7b .29 .45
7c .57 .41
(greater than .9 or less than .1) which might create floor or ceiling effects
(McBee, 2010) and require the elimination of the respective variable from
analysis. Floor or ceiling effects occur when problems are too difficult, or too
performance, were 7b and 4, having the lowest means of .29 and .35, respectively.
Both problems 7b and 4 have short answer formats and require calculation of the
slope of a line given the graph of the line. The least difficult MDT problems, with
respect to student performance, were problems 5 and 3, with the highest means of
.81 and .78, respectively. MDT problems 5 and 3 have multiple-choice formats,
136
for all nine scores of the MDT problems ranged from .39 to .5, with little
variation.
(Hair et al., 1998, p. 99) and necessary as long as their violation might lower the
to the significance of the factors” (Hair et al., p. 99), but normality is improbable
in education and psychology, especially for the type of large sample (N=1561) in
the current study (Micceri, 1989). As such, the appropriateness of the EFA was
Table 25 shows the teatrachoric and polychoric correlations among the scores
137
Table 25
Problem 1 2 3 4 5 6 7a 7b 7c
7b 1.00 .65
7c 1.00
Table 25 shows that most of the correlations are greater than .3 (rounded),
table of values) and problem 6 (finding the y-intercept given the equation of a line
(finding the slope given the graph of a line) and problem 7c (finding the speed
time). There is also a high correlation (.59) between MDT problem 7a (comparing
speeds at three intervals of time, given the graph representing the change in
distance as a function of change in time) and MDT problem 7c. Given that MDT
problems 7a, 7b, and 7c refer to the same situation, higher correlations might have
theoretical relationships. Also high are the correlations .56 and .57 for MDT
To assist in determining if any of the problems might have been dropped from
analysis, correlations of the MDT problems with the MDT total score were
Table 26
Problem 1 2 3 4 5 6 7a 7b 7c
Correlation .62 .78 .78 .66 .73 .59 .79 .65 .70
with MDT
Total Score
true scores. The test was found reliable, with a Cronbach’s Alpha value of .76,
which is greater than .7, the accepted cut-off value (Henson, 2001).
Factors and their Interpretation. After the extraction of two initial factors,
trials with three factors were performed as well. The goal was to extract an
optimum number of factors that would lead to the simplest possible structure
139
while accounting for the largest percentage of the explained variance. More
factors extracted account for higher percentages of explained variance but lead to
structures but account for smaller percentages of the explained variance. The
criteria applied when determining the number of factors to extract (Hair et al.,
i. Kaiser (Latent Root, Eigenvalue) Criterion (Hair et al., 1998, pp. 103-104)
is the criterion used most often. Since each variable accounts for a value of
1 from the total eigenvalue, only the factors with eigenvalues greater than
1 that can account for at least one variable are considered significant. This
extracted. Table 27 shows the initial eigenvalues for each of the nine
components.
Table 27
Initial Eigenvalues
Component 1 2 3 4 5 6 7 8 9
Initial 4.35 1.09 0.76 0.63 0.57 0.48 0.45 0.38 0.29
Eigenvalues
As can be seen in Table 27, there are two eigenvalues greater than 1, (4.35 and
1.09). Therefore, according to Kaiser Criterion, two factors for the 9-variable set
should be extracted.
140
ii. Scree Test Criterion (Hair et al., 1998, pp. 104-105) uses the plot (line
graph) of the eigenvalue for each factor to determine the cut-off point
where the segment lines are almost horizontal (the slopes of the segment
lines approach 0). The factors that account for most of the variance are
those selected before the cut-off point. Figure 5 represents the scree plot
141
Figure 5 shows that only the first line segment has a slope with absolute value
greater than 1, while all other line segments are less and less steep, with slopes
According to the Scree Test Criterion, two factors account for most of the
iii. Percentage of Variance Criterion (Hair et al., 1998, p. 104) takes into
sciences), and extracts the number of factors accounting for this a priori
three. At the same time, a factor that accounts for only one or two
at least three variables. Since there are nine variables, the number of
revealed that the third factor was a weak one, accounting for at most two
variables. Table 28 shows the rotated factor loadings for three factors.
142
factors. With the VARIMAX method, high loadings are expected in each
Table 28
the first factor. MDT problems 7a, 7b, and 7c loaded significantly on two or three
factors, making the interpretation difficult. MDT problems 4 and 6 loaded on the
third factor.
variables, a solution with two factors seems preferable (Hair et al., 1998, pp.106-
113).
143
v. Heterogeneity of the Respondents Criterion (Hair et al., 1998, p. 105) is
useful in situations when the researcher suspects that the sample is not
expected that factors reflecting the variables that are more homogeneous
across the sample will be extracted first, while later factors will be loaded
Table 29 shows the means of scores of the nine MDT problems for all students
Table 29
Problem 1 2 3 4 5 6 7a 7b 7c
Overall .51 .64 .78 .35 .81 .51 .59 .29 .57
Group 0 .39 .51 .69 .24 .76 .41 .47 .16 .44
Group 1 .76 .90 .95 .58 .91 .71 .81 .53 .81
Group 1 means are greater than Group 0 means. For example, for MDT problems
7b and 4 with the lowest overall means, .29 and .35, respectively, there were
more than three times (.53/.16) as many students in Group 1 than in Group 0 who
correctly solved MDT problem 7b, and more than twice (.58/.24) as many
students in Group 1 than in Group 0 who correctly solved MDT problem 4. Since
more than 90% of Group 1 students correctly solved MDT problems 2, 3, and 5,
these variables would produce outliers and should be dropped from analysis.
144
According to Heterogeneity of the Respondents Criterion, more factors should
be extracted. In this study the desirable number of factors is three, but the number
of factors that could be extracted is restricted to two, given the small number of
variables.
a simple structure where each variable loads highly on only one factor.
Trials with two factors and then three factors were performed. Weak
factors, with high loadings for only one or two variables were excluded,
factor loading and final communalities estimates when two factors are
extracted.
Table 30
145
The columns labeled Factor 1 and Factor 2 in Table 30 show the factor
loadings (correlations between variables and the two extracted factors) for each of
the nine MDT problems. All variables, except MDT problem 6, loaded highly
(> .5) on the first factor and lowly on the second factor (< .4), which indicates
MDT problem 6, interviews with students and their teachers from one of the
middle schools revealed that Pre-Algebra students did not learn about finding the
variable indicates the reliability of the factor analysis, i.e., the variable is well
sample size is large, and in this case the sample size is large (N=1561), allowing
for lower values (< .4) of communalities to be acceptable. The total percentage of
trace (obtained by adding the final communality estimates and dividing the sum
by the number of variables) is 49%, a moderate value, which shows how different
the variables are from one another. Small communalities for MDT problems 1, 5,
6, and 7a indicate that more than half of the variance in those variables remains
The total variance explained by these two factors is split unevenly between
factors: Factor 1 accounts for 43%, and Factor 2 accounts for only 6% of the total
variance. If the factors are rotated, the same amount of variance is explained, but
is distributed more evenly between factors, 25.6% and 23.4%, respectively. The
146
aim is to obtain a matrix with either high loadings (close to 1 or -1), or low values
Table 31
1 .53 .36
2 .71 .35
3 .74 .16
4 .37 .60
5 .54 .18
6 .06 .58
7a .43 .55
7b .35 .71
7c .49 .55
As can be seen in Table 31, the rotated factor matrix offers a simple structure
on Factor 1 (factor loadings > .5). Conceptually, all of those problems are situated
problem 1 requires making the Cartesian connection from a point to the equation
of the line. MDT problem 2 requires making the Cartesian connection from the
equation of a line to points on the line. MDT problem 3 requires making the
147
connection between a table of values and a graphic representation of a linear
determining the slope of a line from the graph of the line (MDT problems 4, and
7b), finding the y-intercept of a line given the equation of the line in standard
form (MDT problem 6), finding the rate of change/speed of a car given the linear
graph of the distance-time function (MDT problem 7c), and comparing three
speeds of a car given the three-segment linear graph of the distance-time function
(MDT problem 7a). Conceptually, MDT problems 4, 6, 7a, 7b, and 7c require the
Functions.
148
Table 32
1 .53
2 .71
3 .74
4 .60
5 .54
6 .58
7a .55
7b .71
7c .55
As can be seen in Table 32, MDT problems 1, 2, 3, and 5 were grouped along
7b, and 7c were grouped along the Linear Functions dimension. To better reveal
the structure of relationships between variables for each of the factors, all low
Table 33 and Table 34 present the rotated factor loadings, using the
149
Table 33
2 .72
3 .64
4 .48
5 .50
6 .33
7a .53
7b .72
7c .51
Table 34
1 .81
4 .71
6 .60
7a .51
7b .56
150
The structure of relationships for Group 0 presented in Table 33 is identical
with the one representing the structure of relationships for the entire sample
presented in Table 32. Indeed, MDT problems 1, 2, 3, and 5 were grouped along
the Connections among Representations factor, and MDT problems 4, 6, 7a, 7b,
As can be seen in Table 34, MDT problems 2, 3, 5, and 7c were deleted from
analysis because they had mean scores above .9 and produced outliers.
between variables remained the same for both groups of students, i.e., the
Surrogate Variables for Data Reduction. The method used for data reduction
was the selection of surrogate variables to represent the entire set of variables. A
surrogate variable is a variable that represents the entire factor in further analyses.
This method of reduction was preferred being easier to interpret, though it left out
the other variables. The rotation used for obtaining a simple factor structure was
orthogonal. Therefore the factors were independent, allowing for the selection of
a) For Group 0, MDT problem 2, with loading .72 (Table 33), was
151
Representations factor. MDT problem 7b, with loading .72 (Table
33), was selected as surrogate variable for the Linear Functions factor.
b) For Group 1, the only choice for the surrogate variable representing
with loading .81 (Table 34). MDT problem 4 with loading .71 (Table
34) was selected as surrogate variable for the Linear Functions factor.
defend. Both MDT problems 1 and 2 illustrate the Cartesian connection, which is
Problems by Grade and Current Mathematics Course. After data reduction, the
Table 35 presents the means of scores for MDT problems 2 and 7b by grade
152
Table 35
As can be seen in Table 35, the mean scores for MDT problems 2 and 7b for
Group 0 students were about the same, with a very small increase (≤ .03) from
Grade 8 to Grade 9 , and a small increase (≤ .1) from Grade 9 to Grade 10.
Problem 7b was solved correctly by only 11% of Grade 8 students, 14% of Grade
Table 36 shows the means of scores for MDT problems 2 and 7b by current
Table 36
Mean Scores for MDT Problems by Current Mathematics Course for Group 0
identical to those in Table 35. These results are expected since 100% of Grade 8
153
students were enrolled in Pre-Algebra, 100% of Grade 9 students were enrolled in
for Group 0 students shows little difference between Pre-Algebra and Algebra I,
as the mean scores for MDT problems 2 and 7b increased by at most, . 03. There
was a small difference between Group 0 students enrolled in Algebra I, and those
enrolled in Geometry, but the difference between the mean scores for MDT
Table 37 presents mean scores for MDT problems 1 and 4 by grade for
students in Group 1.
Table 37
As can be seen in Table 37, the mean scores for MDT problem 1 increased by
grade, from .68 in Grade 8 to .94 in Grade 10. MDT problem 1 was successfully
lower than the mean scores for MDT problem 1, the mean scores for MDT
problem 4 also increased by grade, from .46 in Grade 8 to .77 in Grade 10. More
154
than half (54%) of Grade 8 students and about one-fourth (23%) of Grade 10
Table 38 shows mean scores for MDT problems 1 and 4 for students in
Table 38
Mean Scores for MDT Problems by Current Mathematics Course for Group 1
Table 38 indicates that Group 1 mean scores for MDT problems 1 and 4
increased by current mathematics course, with values very close to those values
shown in Table 37. This result is not surprising since 95% of Grade 8 students
and 89% of Grade 10 students were enrolled in Algebra II. MDT problem 4 was
not solved by 28% of Geometry students, and 21% of Algebra II students from
Group 1.
1. There was little variation in student performance on the MDT from grade
from Pre-Algebra to Algebra II. Mean scores more than doubled from
students correctly solved, on average, more than eight of the nine MDT
problems.
4. The most difficult problem was identifying the slope of a line from a
scores between .11 and .25) and low to moderately high across grades and
only for Group 0 students (mean scores between .47 and .58), and
increased from moderate to very high values for Group 1 students across
grades and current mathematics courses (mean scores between .67 and
.95).
156
4.3 Research Question 2
scored using the Scoring Guide (SG) described also in Chapter 3, and mean
scores by MDT problem were calculated for Group 0 and Group 1 students. Mean
scores on MDT problems were arranged from least to greatest; the most difficult
MDT problem that was ranked highest had the lowest mean. The Ranking
was the problem with the highest frequency, i.e., identified by the greatest number
of students as being the most difficult. To assess the association between the
variables representing the ranking of MDT problems by their mean scores and the
MDT and student ranking of MDT problems by perceived difficulty, for students
in Group 0 and Group 1, respectively. Roman numerals are used for ranking, with
“I“ representing the most difficult problem. The problem number is followed by,
in parentheses, mean scores for the MDT problems in the “Problem (Mean)”
157
column, and frequencies of RQS responses in the “Problem (Number of
Students)” column.
Table 39
Students
performance
I 4 (.24) I 1 (323)
II 7 (.38) II 2 (158)
IV 6 (.41) IV 4 (105)
V 2 (.50) V 6 (112)
VI 3 (.69) VI 3 (38)
As can be seen in Table 39, the most difficult MDT problem for Group 0
students was problem 4 (mean score .24), followed by problem 7 (mean score
difficult, and were identified by Group 0 students as being least difficult for them.
problem 7 as being one problem, therefore for ranking analyses, problem 7 was
158
Table 40
Students
I 4 (.58) I 7 (187)
II 6 (.71) II 2 (96)
IV 1 (.76) IV 6 (81)
V 2 (.89) V 1 (33)
VI 5 (.91) VI 3 (21)
As shown in Table 40, the most difficult MDT problem for Group 1 students
was MDT problem 4 (mean score .58), followed by MDT problem 6 (mean score
.71). MDT problem 7 was ranked most difficult by 187 Group 1 students,
3 and 5 were the least difficult for Group 1 students, and were identified by them
value of 0 for W would mean independence for the two ordinal variables
1990, pp. 365-367). The hypothesis (one-sided) that the variables representing the
H0: The variables representing the ranking of mean scores for MDT problems,
independent ( W 0 ).
H1: W ! 0 .
The test statistic, W , calculated using PASW18, was .43 and statistically non-
significant for Group 0 students, and .52 and statistically significant for Group 1
hypothesis that the variables representing the two rankings are independent could
the ranking of mean scores of MDT problems and the ranking of MDT problems
hypothesis that the variables representing the two rankings are independent (p-
value = .049 < α) could be rejected. Thus, there is a moderate agreement between
the variables representing the ranking of mean scores of MDT problems and the
problems by difficulty.
160
2. For both groups of students, MDT problems 3 and 5 (with the highest
MDT problems 4, 6, and 7 (with the lowest mean scores) were not
3. All students could correctly identify the least difficult problems, but only
students (Group 0 or Group 1). The lack of distinction was necessary because the
of students. The mean scores of MDT problems were ranked from lowest to
highest, the most difficult MDT problem ranked the highest had the lowest mean.
Frequencies obtained from RQT were ranked from highest to lowest, the MDT
problem with the highest frequency being the one considered by teachers as the
most difficult for their students. The ranking of student MDT mean scores was
compared with the ranking obtained from RQT. The relationship between the
variables representing the ranking of MDT problems by their mean scores and the
161
ranking by teachers of MDT problems by difficulty as perceived by their students,
MDT and rankings of problem difficulty by teachers. Roman numerals are used
for ranking (e.g., “I” is the most difficult). The problem number is followed by
parentheses with mean scores for the MDT problems in “Problem (Mean)”
Teachers)” column.
Table 41
Teachers
I 4 (.35) I 7 (15)
II 6 (.51) II 6 (6)
Note. *Tied Rank (MDT problems 1, 2, and 4 were ranked most difficult by
the same number of teachers; therefore they share the same rank, IV).
162
Like their students, when ranking MDT problems by difficulty, teachers also
As can be seen in Table 41, the most difficult MDT problem for students was
problem 4 (mean score of .35). Only one teacher identified MDT problem 4 as
being most difficult for students. Teachers considered MDT problem 7 as most
difficult for their students, which ranked fourth in the hierarchy of student
performance on the MDT. Another difficult problem was MDT problem 6, ranked
second in difficulty based on student performance on the MDT (mean score .51).
MDT problem 6 was ranked second by teachers with respect to its difficulty.
MDT problem 5, the least difficult problem (mean score .81) for students, was
calculated, and the hypothesis (one-sided) that the variables representing the two
H0: The variables representing the ranking of mean scores for MDT problems,
independent ( W 0 ).
H1: W ! 0 .
The test statistic W was calculated using PASW18, and found to be .10, and
representing the ranking of mean scores of MDT problems and teachers’ ranking
2. Most teachers (15) considered MDT problem 7 as being most difficult for
their students. However, when describing the nature of difficulty they did not
notice that the most difficult part was 7b. They mentioned 7a as being
difficult because the graph has three parts and three speeds for the car are
required. They also mentioned 7c because students do not associate slope with
students. Twice as many students correctly solved 7a and 7c, than 7b.
3. Teachers did not consider MDT problem 4 difficult for their students.
by 7b. Both problems 4 and 7b require the calculation of the slope of a line
4. Teachers did not consider MDT problem 1 to be difficult for their students.
Six teachers considered MDT problem 6 difficult because their students did
not learn about the y-intercept. Student performances on MDT problems 1 and
164
4.5 Research Question 4
Questionnaire for Students (RQS). The MDTs were scored using the Scoring
Guide (SG), and students’ responses from Ranking Questionnaires were coded
using the Final Coding Categories (see Appendix B). To gain greater
teacher who also agreed to be interviewed. In each pair of students, one student
had an MDT total score above the group median, and one had an MDT total score
below the group median. All interviews were conducted by the researcher. The
Interviews data were reduced after the Exploratory Factor Analysis was
performed, transcribed, coded using the Final Coding Categories (see Appendix
4.5.1 Coding and Identifying the Nature of Student Difficulties from RQS.
of the nature of the difficulty for the MDT problem that posed the greatest
difficulty for them. Their comments were coded and analyzed. Using the
165
Preliminary Coding Categories. Table 42 presents descriptions of the Final
Table 42
Category Description
As can be seen in Table 42, there are six categories used for coding the
participants’ comments about the nature of difficulties. However, they are not the
same as the preliminary six categories presented in Table 8 (Chapter 3). The
comment”, e.g., “Algebra is hard”) and O (“Other”, e.g., “The wording was not
166
clear”); advancing new categories L (“Never Learned/Taught”) and T (“Thinking
were redefined to include the comments from the old category W. The remaining
The coding for reduced data (MDT problems 2 and 7b were coded for Group 0
students, and MDT problems 1 and 4 were coded for Group 1 students) was
analysis using the Kappa statistic (Cohen, 1960) was performed to determine
consistency among raters. The same assumptions were made for calculating the
independent.
167
ii) The categories representing the nature of student difficulties are of
The inter-rater agreement for the two raters is presented in Table 43.
Table 43
grade level or below their grade level, and Group 1 students who were enrolled in
advanced mathematics courses, that is, courses above their grade level in the
responses of both groups of students, Group 0 and Group 1, was very high, with k
= .94 (p < .05), 95% CI (.91, .98), and k = .90 (p < .05), 95% CI (.81, .99),
respectively.
performed for the reduced data, i.e., difficulties with MDT problems 2 and 7b
were analyzed for Group 0 students and the nature of difficulties with MDT
168
problems 1 and 4 were analyzed for Group 1 students. Tables 44 and 45 present
Group 0 student difficulties with MDT problems 2 and 7b, and Group 1 student
example, the comment “’I thought that one was the hardest because those types
are just hard to me.’[2]”, is a comment about MDT problem 2. In both tables, the
column labeled “Frequencies by Problem” shows, for each problem, the number
169
Problem 2
y 2x 3
Create a table of values for the equation. Complete 5 rows of the table.
Problem 7 (All parts are presented, although the focus of analysis is on 7b.)
The graph below represents the distance that a car traveled after different
numbers of hours.
(7a) Which part of the graph (R, S, or T) represents the hours when the car
moved the
slowest?
(7b) What is the slope of part R of the graph?
(7c) What is the speed of the car in part R?
170
Table 44
Problem 2 Problem 7b
(N=158) (N=136)
171
As can be seen in Table 44, about two-thirds of the 158 students from Group 0
who identified MDT problem 2 as being most difficult made no comment about
the nature of their difficulty, or stated that they did not understand the problem.
Only 19 students who identified MDT problem 2 as most difficult described their
difficulty as the identification of values for x (e.g., “I couldn't figure out what
More than half of the 136 students from Group 0 who ranked MDT problem 7
numbers to use in the formula to compute the slope (e.g., “In part b I didn't know
Note that about 16% of the 1040 students from Group 0 correctly solved MDT
students’ success in solving the problems: 18% of Group 0 students who correctly
solved MDT problem 7c that required finding the speed of the car (a real-life
context) were not able to solve MDT problem 7b that required finding the slope of
172
Table 45 presents difficulties with MDT problems 1 and 4 by category of
difficulty for Group 1 students. For convenience, the texts of MDT problems 1
Problem 1
(a) Is the point with coordinates (2,8) on the graph of the line y 3x 14 ?
(b) How did you decide?
Problem 4
In the graph above, what is the slope of the line? Show your work.
173
Table 45
Problem 1 Problem 4
(N=96) (N=82)
H (Hard Problem) “All the problems were easy.” [1] 1( 1%) 2(2.4%)
“I was never good with slopes, or
knowing how to find them, it's
hard for me.”[4]
L (Never “I don't know how to find the 0( 0%) 1(1.2%)
Learned/Taught) slope. We're never taught it
properly.”[4]
In Table 45 it can be seen that almost half of the 96 students from Group 1 who
identified MDT problem 1 as most difficult made no comment about the nature of
missed the Cartesian connection from point to line, had difficulty identifying
174
(2,8) as coordinates of a point ( x0 , y 0 ) , and checking whether the point (2,8) is
were supposed to be, but I drew out a graph and then it just clicked in my head
the x and y axis”). Interesting to note is that 14% of the 521 students from Group
1 who missed MDT problem 1 were able to correctly solve MDT problem 2,
Also of note is that about one-fourth of the 82 students from Group 1 who
ranked MDT problem 4 as most difficult made no comment about the nature of
should start (e.g., “Didn't really give a place to start”), and assigning a magnitude
in the absence of a grid line (e.g., “It wasn't a grid to help”). Students also had
difficulty identifying the slope as “rise over run” (e.g., “I don't know what
rise or run, and then computing with those (e.g., “With the decimal it made it
included only 472 students, those Group 0 students who identified MDT problems
2 or 7 as being most difficult, and those Group 1 students who identified MDT
problems 1 and 4 as most difficult. The following are the results of the analyses of
difficulties.
computational difficulty.
5. For MDT problem 2, students had difficulty identifying x-values for the
table of values.
6. For MDT problem 4, students had difficulty forming a ratio for slope.
7. For MDT problem 7b, students had difficulty determining the rise or
and their mathematical difficulties with the Cartesian connection and slope,
interviews were conducted with 40 students (28 students from Group 0 and 12
students from Group 1). Using the Preliminary Coding Categories from Tables 9-
12 (see Chapter 3) the researcher and the second rater coded together five student
176
interviews and three teacher interviews, and agreed on definitions and boundaries
for the Final Coding Categories. Agreement occurred when the same paragraph
was coded by both raters using the same categories. Then using the Final Coding
Categories the 40 student interviews were coded and analyzed (see Tables 46-49
and Appendix B). Tables 46 describes the Final Coding Categories for Conceptual
Table 46
As can be seen in Table 46, one more category was added to the preliminary
coding category from Table 9 (see Chapter 3) to account for the Cartesian
Slope (CKS).
177
Table 47
In Table 47, there are only two final categories for conceptual knowledge of
contexts (CKS-P) (Bennett et al., 2004; Larson et al., 2004; Rubenstein, et al.,
1995). Also, the slope context for MDT problems 4 and 7b is geometric-analytical
could not account for knowledge of slope in those contexts (e.g., mentioning “m
178
in y mx b ” was not considered evidence of conceiving of m as a parameter).
Consequently, the final categories for conceptual knowledge of slope were limited
Table 48 presents the Final Coding Categories for Student Difficulties with the
Table 48
Connection
difficulty identifying the coordinates of a given point with values for the
variables x and y from the equation of the line, substituting the coordinates of the
point into the equation of the line, and checking if the equality statement is true.
start completing a table of values for the function, and calculating the
corresponding y-values.
Table 49 provides the Final Coding Categories for Student difficulties with
Table 49
180
The categories identifying student difficulties with slope presented in Table 49,
7b.
An inter-rater agreement analysis using the Kappa (k) statistic (Cohen, 1960)
Student Difficulties (SDC, SDS) were performed after the data were reduced, for
the 40 student interviews, and was high, from k = .80 (p < .05), 95% CI ( .63, .97)
to k = 1.
and its converse. When students described their solutions to MDT problems 1 and
2, they did not refer to the mathematical theorem, but rather identified their
schemes of action for solving the problems. Those schemes of action were not
always successful.
Problem 1. Nine of the interviewed students from Group 1 solved MDT problem
181
The checking step was clearly stated before deciding whether the point was on
the line (e.g., “…and 8 does equal 8 , so it does lie on the line”).
Three students from Group 1 had difficulty with the Cartesian connection from
point to line in MDT problem 1. One student plotted the point (2,8) and graphed
the line y 3x 14 (see Figure 8). Since the point (2,8) was not on her line, the
student concluded that the point could not be on the line y 3x 14 , without
182
Figure 8. Example 1 of Student Answer to MDT Problem 1
(see Figure 9). From his representation of the equation, he could not decide
183
None of the schemes tried by the third student were successful (see Figure 10).
to check whether the statement 8 6 14 is true. She also tried making a table
of values, graphing the line y 3x 14 using the table of values, and plotting the
184
I draw a graph first. I put in the numbers y 3x 14 ; 8 3(2) 14 ;
you plug all the numbers in and you draw a line, this is going up. So
Problem 2. All interviewed students from Group 0 (N=28), with the exception of
One student explains how she solved MDT problem 2 (see Figure 11):
“I just picked random numbers for x and then I plugged the numbers I chose into
Only one of the interviewed students remarked on the pattern “subtract 2 every
time” and used it to complete the last row of his table (e.g., “…4 times 2 is 8
185
plus 3 is 5 and it’s going by 2 so this is 7 ”). All other students calculated each
of the y-values.
“Because it didn't have any numbers, patterns to start off with, then I
y 2 x 3 .
given x 0 there exists a unique y-value that satisfies the equation of the line
used the scheme of calculating the y-intercept of a line given the point (x, 0) , as
I just put 1, 0, 1, 2, 3, and to find y just put it in, put the x value in the
186
plus b and 0 and … then do the same thing for 1…Negative 2 times 1 plus
2. Students who missed the Cartesian connection from point to line in MDT
187
y 3x 14 or to check whether the statement 8 3(2) 14 is true. They
corresponding y-values.
( x1 , y1 ) and ( x2 , y 2 ) , the slope of the line connecting the two points is the “rise
y 2 y1
over run,” or m . As mentioned in Chapter 3, students with a
x 2 x1
triangles,” counted the rise and the run, and calculated slope as “rise over run.”
y 2 y1
function) calculated the slope using the formula m .
x2 x1
used by students for calculating the slope in MDT problem 4 has two steps:
188
Step 1. Identify rise (change in y, or y 2 y1 ) and run (change in x, or x2 x1 ) .
y 2 y1
Step 2. Calculate “rise over run,” or .
x 2 x1
The following are two examples of students’ algorithms used to calculate the
To find the slope of a line I would do the rise over the run …I actually
don’t remember how to find the slope with just an equation or formula
right now so I’m just going to go ahead and count… the rise is 1, 2, 3, 4,
y 2 y1
You need to know the slope formula which is (writing (0, 4) ,
x 2 x1
54 9
(3,5) , 3 ). I took the two points that they gave us and I
30 3
put the y-s on top and I subtracted them and put the x-s on the bottom and
I divided.
For Student 1, the geometric aspect of line prevailed and the student found the
rise and run by counting the number of units. For Student 2, the analytical aspect
of linear function prevailed and the student found the change in y and the change
Ten of the interviewed students from Group 1 correctly solved MDT problem
4. They identified the coordinates of two points from the graphed line and used
189
y 2 y1
the formula to compute the slope. Only two of the interviewed students
x 2 x1
had difficulty finding the slope in MDT problem 4. Both students constructed
used them to figure out (by counting) the rise and the run. One of the students
who constructed a “slope triangle,” drew two line segments, “rise” and “run”
together with the line segment joining the two points of the given line, (3,5)
and (0, 4) (see Figure 13). He correctly identified and calculated the rise and the
The student identified the slope of the line as the ordered pair (3, 9) since he could
190
The other student who had difficulties calculating the slope in MDT problem 4
drew a vertical line segment from (3,5) to (3, 0) and identified it as the slope
of the line (see Figure 14). The length of that line segment might be associated
with the rise in a “slope triangle” with the vertices (3,5) , (3, 0) , and
(3, 0) situated on the X-axis with the measure for the rise; he “counted over 3
from 0” and took into consideration the negative sense of the X-axis:
The student did not mention any second quantity that might play the role of the
run, and concluded that the slope of the line had the magnitude of 3 .
191
In conclusion, the geometric perspective of slope prevailed in both schemes of
4.5.2.2.2 Student Difficulties with Slope in MDT Problem 7b. The importance
slope. They calculated the slope of part R of the graph as “rise over run,” using
the same unit of measurement on both coordinate axes, despite the fact that there
was a different unit of measurement on the Y-axis. They found that the run, the
line segment from the origin of the system of coordinates to (3, 0), had a length of
3 units on the X-axis, and the run, the line segment from the origin of the system
of coordinates to (0, 120), had a length of 4 units on the Y-axis. They concluded
that the slope of the line segment in MDT problem 7b is 4/3 (rather than 120/3).
One student explained how he calculated the slope in MDT problem 7b (see
Figure 15):
192
Figure 15. Example of Student Answer to MDT Problem 7
As can be seen in Figure 15, the student was able to figure out the speed of the
193
homogenous systems of coordinates, but not for non-homogeneous systems of
The following are results from interviews about student difficulties with slope:
failed to identify the rise, the run, or the ratio “rise over run.”
3. In MDT problem 7b, students failed to take into consideration the values
instead counted the “tick marks” on the Y-axis (1, 2, 3). When the analytic
y 2 y1
perspective of slope prevailed, and students used to calculate the
x 2 x1
slope, the interviewed students did not encounter difficulties, except for
that students had difficulty with the Cartesian connection and slope. Students
missing the Cartesian connection from point to line had difficulty identifying a
given ordered pair as Cartesian coordinates of a point, and checking whether the
point is situated on a line, given its equation. Students missing the Cartesian
connection from line to points had difficulty identifying x-values for a table of
values and calculating the corresponding y-values, given the equation of the line.
194
forming a ratio for slope. Also, students had difficulty identifying the unit of
only those students with a geometric perspective of slope had difficulty figuring
All participating teachers (N=26) solved the MDT problems, completed the
Ranking Questionnaire for Teachers (RQT) for which they ordered the MDT
problems from most to least difficult from the perspective of their students, and
commented on the nature of the difficulty for the two problems they considered to
be most difficult for their students. For comparison purposes, teachers’ written
comments about student difficulties with the MDT problems were coded using the
same Final Coding Categories as the ones used for coding students’ written
comments from the Ranking Questionnaires for Students (Table 42). The inter-
Twenty teachers, seven from middle school and 13 from high school, were
interviewed. Teachers’ interviews were coded using the same Final Coding
Categories as the ones used to code student Interviews (Tables 46-49). The inter-
rater agreement was also high, from k = .80 (p < .05), 95% CI (.59, 1) to k = 1.
195
In the sections that follow, teachers’ assessments of problem difficulty are
problems 1 and 2) and slope (MDT problems 4 and 7b), along with student
Table 50
H (Hard Problem) 0
196
As shown in Table 50, teachers did not consider any of the MDT problems as
being “hard” for their students. Ten teachers (38%) identified the lack of
taught/learned) as being the main factor for problem difficulty. Three teachers
(9%) identified the amount of time necessary to solve a problem because of its
features like, the length of the problem (e.g., MDT problem 7) or the format of the
explanation of the solution procedure). Only one teacher (4%) mentioned that
students have difficulty remembering the procedure for calculating the slope of a
line. Also, only one teacher (4%) referred to student difficulty with the vocabulary
difficulty, none of those teachers described an actual student difficulty with slope.
Only one teacher identified MDT problem 1 as being the most difficult for
students missing the Cartesian connection, and only one teacher identified MDT
problem 2 as being most difficult for students because of the vocabulary used.
Connection.
MDT, 867 (55.5%) students did not correctly solve problem 1. Only one (3.8%)
197
of the 26 teachers ranked MDT problem 1 as most difficult. She described the
If the students plug in the numbers into the equation, they might not know
what to do next, simplify, then they may not know how to check if it is
‘true’ or not. If the students graphed the equation, a hand drawn graph,
may be hard to find the answer because hand drawn graphs can be hard to
read.
solve MDT problem 1, and plotted the point and graphed the line, then difficulty
may arise because hand drawn graphs may be inaccurate or difficult to interpret.
Interviews revealed that students had difficulty identifying the ordered pair
true, as well as deciding whether the point is on the graph after they plotted the
All interviewed teachers (N=20), except for the teacher mentioned above, were
connection from point to line and difficulties with MDT problem 1. Six teachers
described their knowledge of the Cartesian connection from point to line as both a
198
If the coordinates (2,8) are in the graph of that line, then they should
fulfill that equation, so if you put them in, it should balance it, it should be
Fourteen (70%) teachers described the Cartesian connection from point to line
to the equation of two variables because it makes the sentence true, right?
their students who could not solve MDT problem1, or tried to solve it using a
graphing approach:
sure that it’s a true statement. If they don’t have, can make that connection
switching the x-value with the y-value in (2,8) when substituting those values
pair. Knowing that x comes first, y comes second and knowing that they
in MDT problem 1 for their students, and referred to the Cartesian connection
199
from point to line as an algorithm to solve the problem, they failed to suggest that
MDT, 685 (43.9%) students did not correctly solve problem 2. Only one (3.8%)
of the 26 teachers ranked MDT problem 2 as most difficult “mostly because of the
identified the mathematical difficulty as the absence of any given x-values to help
It’s sort of open-ended in that they can fill in any value for x and finding y
is going to be difficult, filling out that table, just because of how open-
ended it was. If the problem included values for x, it might be a little bit
Two (10%) of the interviewed teachers referred to the novelty of the format of
were always given to their students, and the format of the table was different from
200
Teacher 1:
the values of 2 to 2 . I say that all the time. I let them know what
Teacher 2:
Our district testing, our state testing, and our books all do the tables the
same way. I can see them looking at this table and going ‘What?’ because
they’re used to having to put the equation in the middle, plug it in, and
then solving it. So, just the format, I can see them being confused by the
format. Sometimes if they don’t recognize the format, they just shut down.
values once x-values are identified, although this type of difficulty arose for two
these points together and so if they don’t understand that, they wouldn’t
know what to do. I’ll show a line, but what I’ll do is I’ll go in and I’ll just
put points, I’ll take my marker and just make dots all the way along it, and
I’ll say: “See all these points here? All these points make this equation
true. So, any of these points are values of x and y that make it true.”
201
As with problem 1, most teachers referred to the Cartesian connection from
However, they failed to point out difficulties their students may encounter at each
Analytical Context
the MDT, 1009 (71.4%) students did not correctly solve problem 4. Only one
(3.8%) of the 26 teachers ranked MDT problem 4 as most difficult because “no
encountered difficulties with figuring out the slope. Those students attempted to
construct “slope triangles” and failed to identify the rise, the run, or the ratio “rise
over run.”
Well, in this case, there’s various ways. You can pick any two points and
do the change in y over the change in x. And then of course there’s always
the rise over the run thing, so you basically you use just a method of
counting or if you actually use that formula. I think for most students […]
making the segment a rate-triangle format and doing the rise over the run
202
Student interviews also revealed measurement difficulties when identifying the
“run” for calculating the slope. This difficulty occurred when students chose
points for the “run” that were not integers. Instead, they approximated the x-
intercept of the line. When prompted by the researcher, four (20%) of the
If we refer to the point where this line from problem 4 intersects X-axis
they could give you many different values. They could give you 0, they
would tell you that it’s (0, 1.5), they’d tell you that it’s (0, 1) because it’s
close enough to… they’ll think that it’s close enough, then they’ll figure
Two (10%) of the interviewed teachers suggested that the lack of gridlines or
labels for the points in MDT problem 4 may have been the cause of students’
measurement difficulties. However, student interviews did not reveal this type of
Also another error that they make is they flip the x and y coordinates …
they’ll write down the coordinate right but when they go to plug it into
y 2 y1
their , they’ll use the x and y of one coordinate, of the same
x 2 x1
coordinate, and put that on top and they put the x and y on the bottom. I
can write that out for you so that you kind of get an idea. They’ll have
203
y 2 x2 instead. They’ll use just one point and be like, O.K., I’m starting
with this point and that’s my second point so I’m going to do that one on
y 2 y1
All interviewed students who employed the formula to compute the
x 2 x1
students’ difficulties with slope, teachers did not mention any of the difficulties
Coordinates. On the MDT, 1115 (71.4%) students did not correctly solve problem
7b. Fifteen (55.6%) teachers identified problem 7 (7a, 7b, and 7c) as being most
difficult for their students. Twelve (44.4%) teachers described the major difficulty
as both misreading and interpreting the graph, particularly interpreting the graph
formed by the three line segments (7a), and the computing of the speed of the car
(7c). Only when prompted by the researcher did 12 (60%) of the interviewed
They’re counting the rise over the run. So they’re going over, they know
how many spaces they’re going over, but they’re counting the spaces,
204
they’re not counting the quantity. So they’re going over 3 like they should,
they know that they’re going over 4, but they’re not associating 4 with this
value.
Only one of the interviewed teachers (5%), and only when prompted,
system of coordinates:
I would explain it to them that that tick mark isn’t just one. What does it
say next to it? Is it 30? So if that one tick mark is 30 is it really that I’m
students, when prompted by the researcher, they could describe the nature of the
205
missing the Cartesian connection (MDT problem 1) and considering slope
x-values for the table in MDT problem 2, four (20%) teachers referred to
4, and four (20%) teachers described the difficulty measuring the run or
Chapter Summary. In this chapter the data, analyses of data and results of the
analyses were presented by research question. The main results are the following:
i. There was little variation in student performance on the MDT from Grade
ii. All students knew which of the problems were correctly solved, and
therefore they could correctly identify the problems with the highest mean
scores as being the least difficult problems. Only Group 1 students could
identify some of the problems with lower mean scores as being difficult
206
for them. Teachers could not identify the MDT problems that were most
iii. The most difficult problems for participating students were those requiring
identification of the slope of a line from its graph, MDT problems 4 and
7b. That difficulty persisted across grades and mathematics courses. The
failed to identify the rise, the run, or the ratio “rise over run.” In MDT
problem 7b, students failed to take into consideration the values of the
function for rise and instead, counted the “tick marks” on the Y-axis.
y 2 y1
When the analytic aspect of slope prevailed and students used to
x 2 x1
calculate the slope, the interviewed students did not encounter difficulties,
iv. Teachers’ comments about the nature of their students’ difficulties did not
the MDT and Interviews. The agreement remained low, even when
difficulties. Conclusions that can be drawn from the results of the data analyses,
207
along with limitations for generalizing results and recommendations for further
208
Chapter 5
5.1 Introduction
To meet the national need for more mathematicians, scientists, engineers and
highly skilled workers, high schools are now requiring completion of the Algebra
I, Geometry, and Algebra II sequence for entrance into college and the workplace.
Of those courses, Algebra I is singled out as fundamental preparation for the study
of higher-level mathematics and the sciences (Gamoran & Hannigan, 2000; Katz,
2007; Matthews & Farmer, 2008; Smith, 1996; Speilhagen, 2006). Among the
key concepts developed in Algebra I are linear equations and functions (National
Mathematics Advisory Panel, 2008). During the last two decades, student
difficulties with aspects of linearity and linear functions have been studied by
Cheng, 2010; Davis, 2007; Ellis, 2009; Hines, 2002; Knuth, 2000; Lamon, 1993;
Lobato & Siebert, 2002; Moschkovich et al., 1993; Newburgh, 2001; Orton, 1983;
Schoenfeld et al., 1993; Smith et al., 1989; Stump, 2001a; Zaslavsky et al., 2002).
Despite their efforts to gain insight into student difficulties and assist mathematics
difficulties with linearity and linear functions persist (American Diploma Project,
2010).
In prior research studies, with few exceptions (Even & Tirosh, 1995; You,
2006), student understanding and difficulties were assessed without input, in the
form of descriptions, from students or their teachers (Cheng, 2010; Davis, 2007;
209
Knuth, 2000; Orton, 1983; Stump, 2001a; Zaslavsky et al., 2002). By contrast, the
present study not only tested students on concepts of linearity, but also
those results with respect to previous research findings are presented by research
a. Grade level?
c. Problem?
210
5. Do teachers’ speculations about the nature of students’ difficulties correspond
and interviews?
a. Grade level?
c. Problem?
(IQS) and the Mini-Diagnostic Test (MDT) were administered to all students.
Data from IQS (grade level, current mathematics course) and MDT scores were
analyzed.
correctly solved five of the nine MDT problems. However, analyses of the MDT
grade level. The presence of higher scores at higher grade levels may be due to
mathematics courses.
211
above their grade level in the mathematical instructional sequence, outperformed
sequence. Performance gains on the MDT across courses were due mainly to
relationship between the type of learners participating in this study and their
to students in Grades 8 and 10, and found that those who completed Algebra I
later or did not take Algebra I . The American Diploma Project (ADP) reported
grade level. That is, students who completed Algebra I in earlier grades
those two research studies had results in concert with the results of the present
study, their assessments differed. In both Gamoran and Hannigan’s and ADP’s
linearity) and were tied to course completion (i.e., administered at the end of an
212
instructional sequence). By contrast, the present study measured student
during the instructional period and not at the end of a course, and was not
greatest difficulty with MDT problems focusing on slope (MDT problems 4 and
7b, identifying the slope given the graph of the line), and the Cartesian connection
(MDT problems 1 and 2, Cartesian connection from point to line, and line to
points, respectively). Student difficulties with slope and the Cartesian connection
persisted across grades, mathematics courses, and types of learners (Group 0 and
enrollment in more advanced mathematics courses, like Geometry and Algebra II.
One possible explanation for higher MDT scores for Geometry and Algebra II
advanced levels. Findings of the present study are in concert with the three-
country study using the MDT conducted by Greenes et al. (2007a) that found
that U.S., Israeli and Korean students had difficulties with slope (MDT problems
4 and 7b). Furthermore, U.S. students had greater difficulty with the Cartesian
connection (MDT problems 1 and 2) than did the Korean and Israeli students.
Those differences in performance, even at the same grade level, suggest that
U.S. curricula, versus curricula and pedagogical practices in Korea and Israel,
213
may place less emphasis on making connections among various representations of
functions.
In conclusion, student performance on the MDT varied little across grades, but
students who were enrolled in courses above their grade level expectations in the
instructional sequence. These findings are consistent with the results of the studies
conducted by Gamoran and Hannigan (2000), and the American Diploma Project
(2010). The most difficult problem across grades and mathematics courses was
the problem of identifying the slope of a line, given the graph of the line (MDT
problems 4 and 7b). Results from previous studies with similar or same problems,
analytical contexts (Greenes et al., 2007a; Knuth, 2000; Orton, 1983; Stump,
Ranking Questionnaire for Students (RQS) were administered to all students. The
MDT mean score for each of the MDT problems was used to obtain the empirical
problem difficulty from the RQS were used to obtain student subjective ranking
214
performance on the MDT and the subjective ranking based on student perception
Group 0 students and moderate agreement for Group 1 students between their
students could correctly identify the least difficult problems, but only Group 1
students could identify some of the difficult problems. In other words, all students
knew what they knew well, but only the high-achieving students somewhat knew
what they did not know on MDT. Overall, for both groups of students, MDT
problems 3 and 5, with the highest mean scores, were identified as the least
difficult, whereas MDT problems 4, 6 and 7 with the lowest mean scores, were
(Lingard et al., 2009) found a similar result of moderate agreement between the
215
learner, and the problem content and context may play important roles not only in
Student performance, given by the MDT mean score for each of the MDT
students’ performance showed that teachers were not aware of their students’
difficulties with slope (MDT problems 4 and 7b) and the Cartesian connection
(MDT problem1). For example, none of the teachers identified MDT problem 4 as
MDT and teacher ranking of MDT problems by difficulty. By contrast, with the
216
exception of the study by Lingard et al. (2009), other studies showed low or
moderate agreement (Koedinger & Nathan, 2004; Lee & Heyworth, 1996, 2000;
Lorge & Kruglov, 1953; Nathan & Koedinger, 2000; Zhou, 2009), or high
agreement (Gire & Rebello, 2010). However, even in the case of the study by
Lingard et al. (2009), only the level of difficulty of the problem on a scale with
three values was identified. In all of these studies, since the populations, contents,
conducted with 40 students (see Chapter 3). Students’ written comments from the
RQS about the nature of their difficulties, and students’ interviews were analyzed.
In their written comments, about 51% of the 472 students did not describe
their difficulties, and 28% of students referred to the resources necessary to solve
student assessment of problem difficulty and amount of time needed to solve the
problem also was identified by Lingard et al. (2009). Taking into consideration
the association between problem characteristics and the number of solution steps,
217
the findings from the present study are in accord with Lee and Heyworth’s
(1996; 2000) findings about student perceived problem difficulty that could be
predicted by, among other factors, the number of algorithmic steps and the
Only 20% of the 472 students in the present study made comments about their
comments and interviews revealed that students had difficulty with the Cartesian
identifying the Cartesian coordinates of a point and checking whether the point is
situated on a line, given the coordinates of the point and the equation of the line
(MDT problem 1); and identifying x-values for a table of values and calculating
corresponding y-values, given the equation of the line (MDT problem 2).
(2007a), Knuth (2000), Moschkovich et al. (1993), and Schoenfeld et al. (1993).
Student difficulties with determining the slope of a line, given the graph of the
line (MDT problems 4 and 7b, included: 1) measuring (e.g., not connecting the
rise and run of the slope of the line with the corresponding oriented segments,
positive or negative); 2) forming a ratio for slope (e.g., attending only to the rise,
of coordinates (e.g., counting the “tick marks” on the Y-axis and disregarding the
or rise) with their graphic representations and with their corresponding positive
218
or negative sign, is also reported by Greenes et al. (2007a), Moschkovich et al.
(1993), Schoenfeld et al. (1993), and Teuscher and Reys (2010). Results of the
present study as well as results from previous studies (Cheng, 2010; Kaput &
Maxwell-West, 1994; Lamon, 1993; Lobato & Siebert, 2002) suggest that
more difficult task than identifying the rate of change of a linear function when
explanation may be that the graphing context for slope implies knowledge of
toward the positive sense of the Y-axis, its magnitude has a positive sign),
difficulty in identifying the slope of a graphed line lies with the connection with
perspective of slope, (i.e., students who calculated slope as “rise over run”) had
difficulty figuring out the slope of the graphed line in MDT problems 4 and 7b.
Another explanation may be that the complexity of the problem increased for
1994; Lamon, 1993; Lobato & Siebert, 2002). Indeed, calculating the slope of a
identifying the rise and the run from a “slope triangle.” By contrast, students with
by Lobato and Siebert (2002), Orton (1983), and Stump (2001a) suggest that the
difficulty taking into consideration and coordinating both the rise (change in y)
and the run (change in x). As well, students in the present study had difficulty
identifying the unit of measurement when asked to determine the slope of a line
note is that findings of the present study do not support the explanation suggested
by Teuscher and Reys (2010) about student difficulties with slope or rate of
change, specifically that difficulty with the sign associated with the rise or change
in y stems from difficulty with the language used in the problem. Only five of the
1561 students in the present study mentioned difficulties with vocabulary, i.e., not
referring to slope or rate of change varied across studies, the findings of the
present study and other studies (Orton, 1983; Lobato and Siebert, 2002;
Schoenfeld et al. 1993; Stump, 2001a; Zaslavsky et al., 2002) point toward
220
difficulties with the concept itself, with the mathematical conventions behind
representations.
Another explanation for student difficulties may lie in the way that linear
functions, and in particular the concept of slope and the Cartesian connection, are
be associated with the nature of the context that precedes the formal introduction
of slope (e.g., ski slopes, roof tops). For example, in a textbook used in one of the
slope defines the “slant” of a line.[…] For a line on the coordinate plane,
Some roofs are steeper than others. In mathematics, a number called slope
rise to run for any two points on the line (Rubenstein et al., 1995, p. 361).
The geometric perspective of slope as “rise over run” precedes the functional
mathematics classes, like Precalculus and Calculus. The results of this study
221
showed that only students with a geometric perspective of slope had difficulties
conventions behind graphing, may also lie in the way graphing is presented in
U.S. mathematics curricula. For example, the graph of a function is not defined
Larson et al. (2004), students are reminded in Chapter 1 about the procedure to
When you are given the rule for a function, you can prepare to graph the
corresponding output values […] Let the horizontal axis represent the
input t (in minutes). Label the axis from 0 to 5. Let the vertical axis
represent the output h (in feet). Label the axis from 0 to 400. Plot the data
points given in the table. Finally, connect the points […] (Larson et al.,
2004, p. 49).
axes (e.g., scale for the Y-axis). Difficulty with identifying the scale of the Y-axis
Students attempted to calculate the “rise” by counting the tick marks “by ones”
for the Y-axis, instead of recognizing that each interval between two consecutive
222
In summary, possible explanations of student difficulties with slope and the
Cartesian connection may lie in the way these topics are presented in the school
mathematics curriculum.
conducted with 40 students and 20 teachers. Teachers’ written comments from the
RQT about the nature of their students’ difficulties and teachers’ interviews were
analyzed and compared with students’ performance on the MDT and students’
interviews.
Teachers, like their students, did not consider the MDT problems as being
“hard.” About half of the teachers identified the nature of the difficulty as being
the “look of the problem,” that is the lack of familiarity with the problem (e.g.,
MDT problem 2), the number of sub-problems in problem (e.g., MDT problem 7
has three sub-problems), or the format of the problem (e.g., MDT problem 1
requires an extended response). These findings are in concert with Lee and
223
difficulty. Of interest is that teachers in the present study solved all the problems
before identifying problem difficulty. By contrast, the teachers from Lee and
Heyworth’s (1996, 2000) study commented on the difficulty without first solving
the problems. Despite this difference, teachers’ responses were of the same type.
showed that students had difficulty with MDT problems 1 and 2 referring to the
speculations about the nature of their students’ difficulties did not correspond to
and student interviews. Only six teachers (23%) mentioned an actual student
problem 1, and connecting slope in MDT problem 7b with the rate of change in
Only four (20%) of the interviewed teachers mentioned actual student difficulties,
224
b) difficulty finding x-values to complete a table of values for the linear function
of linearity and linear functions (Cunningham, 2005; Davis, 2007; Even & Tirosh,
1995; Stump, 1997, 2001b). When asked to identify student difficulties with
slope, some of the teachers from Stump’s (1997, 2001b) and Even and Tirosh’s
formula and applying it, and difficulty performing numerical calculations. When
among various representations (Cunningham, 2005; Davis, 2007; Even & Tirosh,
1997, 2001b; McDermott et al., 1987; Orton, 1983; You, 2006; Zaslavsky et al.,
2002). For example, teachers had difficulty determining the slope of a line
teachers correctly solved the MDT problems. In others studies (Even & Tirosh,
225
1995; Zhou, 2009) although involving small numbers of teachers, teachers with
some student difficulties and predict student performance. Those studies (Even &
Tirosh, 1995; Zhou, 2009) together with the present study, point out that the
associated with being able to predict student difficulties with the problem.
of problem difficulty, students from Group 1 were better able to describe the
assessment of problem difficulty was based on the “look of the problem,” that is,
student familiarity with the problem, number of sub-problems in the problem, and
prompting, teachers did not attend to the mathematical nature of actual student
slope or rate of change is even more problematic considering the great amount of
instructional time spent teaching the concept. “I beat that topic to death,”
commented one teacher, surprised that students still had difficulties with slope
226
5.3 Limitations of the Study
Limitations of the study focus on the instruments, the study sample, the
The Instruments. When the MDT was designed, there were two major
considerations:
1) The focus should be on a few aspects of linearity and linear functions that,
and 2) Students must be able to complete the MDT, IQS, and RQS in one class
linear functions, the MDT is not comprehensive. For example, only two problems
problems address slope as speed (MDT problems 7a and 7c), and three problems
focus on the Cartesian connection (MDT problems 1, 2, and 6). It may be that
those few problems did not provide a clear picture of the extent of students’
knowledge of those concepts. Furthermore, the MDT mean score for Grade 10
that the problems were too easy for those students, thus creating a ceiling effect
were invited to participate in this study. About two thirds of the participants were
227
from a school district that had only 2% of students eligible for free or reduced-
price lunch. As such, the student population for this study is not representative of
the majority of school districts in the United States. For example, in 2007-2008,
only 29.6% of the U.S. students were enrolled in secondary schools with less than
25% of students eligible for free or reduced-price lunch (National Center for
advanced mathematics courses (i.e., courses above their grade level). Less than
1% of students were enrolled in courses below their grade level. Thus, all results
For the reasons stated above, the findings of this study cannot be generalized
beyond the particular population, and the educational context (cultural settings
example, Pre-Algebra students from one middle school had not yet studied ways
to determine the y-intercept of a line from the equation of the line (MDT problem
6). If the MDT was administered at a later time, student performance might have
been different.
228
The MDT was administered to students by their teachers. Although they
administration took place. Whether or not teachers provided some assistance (e.g.,
questions were subject to the researcher’s interest. If, instead, the interviews were
difficulties, and the theoretical framework used for analysis, results might have
varied.
was not used because of the hierarchical or nested structure of the data, i.e.,
student observations nested within teachers, and thus the violation of the
reported in the study (Raudenbush & Bryk, 2002, pp. 3-10). A hierarchical linear
model could not be applied because of insufficient data collected at Level 1 (i.e.,
229
5.4 Recommendations for Future Research
The Instruments. With respect to the MDT, more problems focusing on the
same aspects of linearity and linear functions, but in different contexts (e.g., slope
Moschovich et al. (1993), would alleviate the ceiling effects observed in the
present study. To avoid ceiling and flooring effects (McBee, 2010), different tests
levels, i.e., low-achieving (students enrolled in courses below their grade level),
level), and high-achieving (students enrolled in courses above their grade level).
only rank problems by difficulty, but also, for each problem, identify factors that
student and teacher familiarity with the problems and problem types. For
example, for students, one might ask, “Which of these problems have you seen
230
For teachers, one might ask, “Which of these problem types have you already
explored with your students?”and “Which of these problem types will you explore
The Population. Using the instruments and procedures but with a greater
present study.
rate of change in functional contexts (e.g., constant speed, constant water flow),
The present study provided a snapshot of student knowledge, and student and
231
student knowledge, as well as changes in their teachers’ perception of their
pre-service and in-service programs for middle and high school teachers of
mathematics.
232
REFERENCES
ACT. (2006). Ready for College and Ready for Work: Same or Different? Iowa
City, IA: Author.Retrieved March 25, 2011, from
http://www.act.org/research/policymakers/pdf/ReadinessBrief.pdf
Bennett, J., Chard, D. J., Jackson, A., Milgram, J., Scheer, J. K., & Waits, B. K.
(2004). Middle School Math Course 3: Algebra Readiness. Austin, Texas:
Holt, Rinehart nd Winston.
233
Bracey, G. (2000, October 25). The Malevolent Tyranny of Algebra. Education
Week , 20, 47, 50. Retrieved May 19, 2009, from
http://www.edweek.org/ew/articles/2000/10/25/08bracey.h20.html?qs=Brac
ey,+g.
Breidenbach, D., Dubinky, E., Hawks, J., & Nichols, D. (1992). Development of
the Process Conception of Function. Educational Studies in Mathematics ,
23 (3), 247-285.
Burger, E. B. , Chard, D. J., Hall, E. J., Kennedy, P. A., Leinwand, S. J., Renfro,
F. L., …Waits, B. K. (2007). Algebra 1. Orlando: Holt, Rinehart and
Winston
Caglayan, G., & Olive, J. (2010). Eighth grade students' representations of linear
o equations based n a cups and tiles model. Educational Studies in
Mathematics , 74 (2), 143-162. doi: 10.1007/s10649-010-9231-z
Carroll, K. (2009, October 30). Most New Jersey students failed new Algebra
tests. NorthJersey.com. Retrieved November 20, 2010, from
http://www.northjersey.com/news/education/Most_New_Jersey_students_fa
iled_new_Alegebra_tests.html
Cavanagh, S. (2008, April 23). Catching Up on Algebra. Education Week , 27, pp.
25-28. Retrieved August 12, 2009, from
http://www.edweek.org/ew/articles/2008/04/23/34algebra_ep.h27.html
234
Cohen, J. (1960). A coefficient of agreement for nominal scales. Educational and
Psychological Measurement 20 (1): 37–46.
doi:10.1177/001316446002000104
Collins, W., Foster, A. G., Winters, L. J., Swart, W. L., Cuevas, G. J., Rath, J., ...
Gordon, B. (1998). Algebra 1: Integration Applications Connections. New
York: Glencoe/McGraw-Hill.
Cooke, R. (1997). The History of Mathematics: A Brief Course. New York: John
Wiley & Sons.
Couto, V., Mani, M., Lewin, A. Y., & Peeters, C. (2007). The Globalization of
White-Collar Work: The Facts and Fallout of Next-Generation Offshoring
(Booz Allen Hamilton and Offshoring Research Network, Duke
University). Retrieved March 25, 2011, from
https://offshoring.fuqua.duke.edu/pdfs/gowc_v4.pdf
235
Darling-Hammond, L. (2007, October 14). High-Quality Standards, a Curriculum
Based on Critical Thinking Can Enlighten Our Students. San Francisco
Chronicle. Retrieved June 23, 2009 from http://articles.sfgate.com/2007-10-
14/opinion/17265943_1_standards-and-tests-assessments-multiple-choice-
tests
De Bock, D., Van Dooren, W., Janssens, D., & Verschaffel, L. (2002). Improper
Use of Linear Reasoning: An In-Depth Study of the Nature and the
Irresistibility of Secondary School Students' Errors. Educational Studies in
Mathematics , 50 (3), 311-334.
De Bock, D., Van Dooren, W., Janssens, D., & Verschaffel, L. (2007). The
Illusion of Linearity:From Analysis to Improvement . New York: Springer.
De Bock, D., Verschaffel, L., & Janssens, D. (1998). The Predominance of the
Linear Model in Secondary School Students' Solutions of Word Problems
Involving Length and Area of Similar Plane Figures. Educational Studies in
Mathematics , 35 (1), 65-83.
Dietz, S. (2010). State High School Tests: Exit Exams and Other Assessments.
Washington, DC: Center on Education Policy. Retrieved March 25, 2011,
from http://www.cep-dc.org/index.cfm?DocumentSubTopicID=8
Dubinsky, E., & Harel, G. (1992). The Nature of the Process of Function. In G.
Harel, & E. Dubinsky (Eds.), The Concept of Function :Aspects of
Epistemology and Pedagogy (pp. 85-107). Washington, DC: Mathematics
Association of America.
236
Education Week. (2006). Diplomas Count: An Essential Guide to Graduation
Policy and Rates. Retrieved March 25, 2011, from
http://www.edweek.org/ew/toc/2006/06/22/index.html
Euler, L. (1984). Elements of Algebra (Reprinted from the 1840 ed.). (J. Hewlett,
Trans.) New York: Springer-Verlag.
Even, R., & Tirosh, D. (1995). Subject-matter knowledge and the knowledge
about students as sources of teacher presentations of the subject-matter.
Educational Studies in Mathematics , 29 (1), 1-20.
237
Greenes, C., Chang, K. Y., & Ben-Chaim, D. (2007). International Survey of
High School Students’ Understanding of Key Concepts of Linearity.
Proceedings of the 31st Conference of the International Group for the
Psychology of Mathematics Education, 2, pp. 273-280. Seoul, Korea.
Retrieved March 25, 2011, from http://eric.ed.gov/PDFS/ED499417.pdf
Greenes, C., Rosenberg, S., Bodie, K., Chevaire, D., Garabedian, C., Wulf, D., …
Herlihy, E. (2007). The Professional Development of Leaders and
Teachers of Mathematics. National Council of Supervisors of Mathematics
Journal of Mathematics Education Leadership , 10 (1), 16-24. Retrived
March 25, 2011, from
http://www.mathedleadership.org/docs/resources/journals/NCSMJournalVo
l10Num1.pdf
Greenes, C., & Rubenstein, R. (Eds.). (2008). Algebra and Algebraic Thinking in
School Mathematics: Seventieth Yearbook. Reston, VA: National Council
of Teachers of Mathematics.
Hair, J. F., Anderson, R. E., Tatham, R. L., & Black, W. C. (1998). Multivariate
data analysis (5th ed.). Englewood Cliffs, NJ: Prentice-Hall.
Helfand, D. (2006 , January 30). A Formula for Failure in L.A. Schools. Los
Angeles Times. Retrieved May 19, 2009 from
http://www.latimes.com/news/education/la-me
dropout30jan30,1,2605555.story.
238
Hoffer, T. B., Venkataraman, L., Hedberg, E. C., & Shagle, S. (2007). Final
Report on the National Survey of Algebra Teachers for the National Math
Panel. Retrieved March 25, 2011, from
http://www2.ed.gov/about/bdscomm/list/mathpanel/final-report-algebra-
teachers.pdf
Holgado –Tello, F. P., Chacón –Moscoso, S., Barbero –García, I., & Vila–Abad,
E. (2008). Polychoric versus Pearson correlations in exploratory and
confirmatory factor analysis of ordinal variables. Quality and Quantity , 44
(1), 153-166, doi:10.1007/s11135-008-9190-y
Jacobson, L. (2008, July 30). California Board Mandates Algebra 1 for All 8th
Graders. Education Week , 27 (43), p. 9. Retrieved March 25, 2011, from
http://www.edweek.org/ew/articles/2008/07/16/43algebra.h27.html?qs=jaco
bson,+L.
Kaput, J. J., Blanton, M. L., & Moreno, L. (2008). Algebra From a Symbolization
I Point of View. n J. J. Kaput, D. W. Carraher, & M. L. Blanton (Eds.),
Algebra in the Early Grades. Mahwah, NJ: : Lawrence Erlbaum Associates.
Kaput, J. J., Carraher, D. W., & Blanton, M. L. (Eds.). (2008). Algebra in the
Early Grades. Mahwah, NJ: Lawrence Erlbaum Associates.
239
Katz, V. J. (2007). Algebra: Gateway to a Technological Future. Washington,
DC: The Mathematical Association of America. Retrieved March 25, 2011,
from http://www.maa.org/algebra-report/Algebra-Gateway-Tech-Future.pdf
Kieran, C., & Chalouh, L. (1993). Prealgebra: The Transition from Arithmetic to
Algebra. In D. T. Owens (Ed.), Research Ideas for the Classroom: Middle
Grades Mathematics (pp. 179-198). Reston, VA: National Council of
Teachers of Mathematics.
Kilpatrick, J., & Izsák, A. (2008). A History of the Algebra in the School
Curriculum. In C. Greenes, & R. Rubenstein (Eds.), Algebra and Algebraic
Thinking in School Mathematics: Seventieth Yearbook (pp. 3-18). Reston,
VA: National Council of Teachers of Mathematics.
Kilpatrick, J., Mesa, V., & Sloane, F. (2007). U.S. algebra performance in an
international context. In T. Loveless (Ed.), Lessons learned: What
international assessments tell us about math achievement (pp. 85-126).
Washington, DC: The Brookings Institution Press.
Kline, M. (1973 ). Why Johnny Can't Add: The Failure of the New Mathematics.
St. Martin's Press.
Koedinger, K., & Nathan, M. (2004). The Real Story Behind Story Problems:
Effects of Representations on Quantitative Reasoning. Journal of the
Learning Sciences , 13 (2), 129-164.
240
Lane, S. (1991). Use of Restricted Item Response Models for Examining Item
Difficulty Ordering and Slope. Journal of Educational Measurement , 28
(4), 295-309.
Larson, R., Boswell, L., Kanold, T. D., & Stiff, L. (2004). Algebra I: Concepts
and Skills. Evanston, IL: McDougal Littell.
Lingard, J., Minasian-Batmanian, L., Vella, G., Cathers, I., & Gonzalez, C.
(2009). Do Students with Well-Aligned Perceptions of Question Difficulty
Perform Better? Assessment & Evaluation in Higher Education , 34 (6),
603-619. doi: 10.1080/02602930802287249
241
Loveless, T. (2009). The 2008 Brown Center Report on American Education:
How Well Are American Students Learning? Washington, DC: Brookings
Institution Press. Retrieved March 25, 2011, from
http://www.brookings.edu/reports/2009/0225_education_loveless.aspx
Mason, E., Zollman, A., Bramble, W. J., & O'Brien, J. (1992). Response Time
and Item Difficulty in a Computer-Based High School Mathematics Course.
Focus on Learning in Mathematics , 14 (3), 41-51.
Matthews, M., & Farmer, J. (2008). Factors Affecting the Algebra I Achievement
of Academically Talented Learners. Journal of Advanced Academic , 19 (3),
472–501. doi: 10.4219/jaa-2008-810
Micceri, T. (1989). The Unicorn, the Normal Curve, and Other Improbable
Creatures. Psychological Bulletin , 105 (1), 156-166.
doi: 10.1037/0033-2909.105.1.156
242
Moschkovich, J. N., Schoenfeld, A. H., & Arcavi, A. (1993). Aspects of
Understanding: On Multiple Perspectives and Representations of Linear
Relations and Connections among Them. In T. A. Romberg, E. Fennema, &
T. P. Carpenter (Eds.), Integrating research on the graphical representation
of functions (pp. 69-100). Hillsdale, NJ: Erlbaum.
Moses, B. (1999). Algebraic Thinking Grades K-12: Readings from the NCTM's
School-Based Journals and Other Publications. Reston,VA: National
Council of Teachers of Mathematics.
National Center for Public Policy and Higher Education. (2004). Policy Alert: The
Educational Pipeline: Big Investment, Big Returns. San Jose, CA: Author.
Retrieved March 25, 2011, from
http://www.highereducation.org/reports/pipeline/pipeline.pdf
243
National Council of Teachers of Mathematics. (1989). Curriculum and
Evaluation Standards for School Mathematics. Reston, VA: Author.
Retrieved March 25, 2011, from
http://www.nctm.org/standards/content.aspx?id=16909
National Science Board. (2008). Science and Engineering Indicators 2008. Two
volumes.Arlington, VA: National Science Foundation (volume 1, NSB 08-
01; volume 2, NSB 08-01A). Retrieved March 25, 2011, from
http://www.nsf.gov/statistics/seind08/pdfstart.htm
Partnership for 21st Century Skills. (2007). Beyond the Three Rs: Voter Attitudes
toward 21st Century Skills. Tucson, AZ: Author. Retrieved March 25,
2011, from http://www.p21.org/documents/P21_pollreport_singlepg.pdf
Pelavin, S., & Kane, M. (1990). Changing the Odds: Factors Increasing Access to
College. New York: The College Board.
244
Pennsylvania Council of Teachers of Mathematics. (1997). Position Paper:
Algebra for ALL Students. Retrieved November 28, 2008, from
Pennsylvania Council of Teachers of Mathematics Position Paper: Algebra
for ALL Students: http://www.pctm.org/AlgebraForAll.html
Post, T. R., Behr, M. J., & Lesh, R. (1988). Proportionality and the Development
of Prealgebra Understanding. In A. Coxford, & A. Schute (Eds.), The Ideas
of Algebra K–12, 1988 Yearbook of the National Council of Teachers of
Mathematics (pp. 78–90). Reston, VA.
Richmond, E. (2009, February 4). Students show math gains despite high failure
rates. Las Vegas Sun. Retrieved November 20, 2010 from
http://www.lasvegassun.com/news/2009/feb/04/students-show-math-gains-
despite-high-failure-rate/
245
Rubenstein, R. N., Craine, T. V., Butts, T. R., Cantrell, K., Dritsas, L., Elswick,
V. A., ... Walton, J.C (1995). Integrated Mathematics 1 . Evanston, Illinois:
McDougal Littell.
Sadler, P. M., & Tai, R. H. (2007, July 27). The Two High-School Pillars
Supporting College Science. Science , Vol. 317 (5837), pp. 457-458.
doi: 10.1126/science.1144214
Schoenfeld, A. H., Smith III, J. P., & Arcavi, A. (1993). Learning: The
Microgenetic Analysis of One Student's Evolving Understanding of a
Complex Subject Matter Domain. In R. Glaser (Ed.), Advances in
Instructional Psychology (Vol. 4, pp. 55-174). Hillsdale, NJ: Lawrence
Erlbaum Associates.
Shettle, C., Roey, S., Mordica, J., Perkins, R., Nord, C., Teodorovic, J., … Brown,
J. (2007). America’s High School Graduates: Results from the 2005 NAEP
High School Transcript Study (NCES 2007467). U.S. Department of
Education, National Center for Education Statistics. Washington, DC: U.S.
Government Printing Office.
Smith, J. (1996). Does an Extra Year Make Any Difference? The Impact of Early
Access to Algebra on Long-Term Gains in Mathematics Attainment.
Educational Evaluation and Policy Analysis , 18 (2), 141-153.
246
Smith, J., Arcavi, A., & Schoenfeld, A. H. (1989). Learning Y-intercept:
Assembling the Pieces of an "Atomic" Concept. Proceedings of the Annual
Conference of the International Group for the Psychology of Mathematics
Education, 3, pp. 174-181. Paris, France.
Strauss, A., & Corbin, J. (1998). Basics of Qualitative Research: Techniques and
Procedures for Developing Grounded Theory. Thousand Oaks, CA: Sage.
Teuscher, D., & Reys, R. E. (2010). Slope, Rate of Change, and Steepness: Do
Students Understand These Concepts? Mathematics Teacher , 103 (7), 519-
524.
247
Tourniaire, F., & Pulos, S. (1985). Proportional Reasoning: A Review of the
Literature. Educational Studies in Mathematics , 16 (2), 181-204.
Van Dooren, W., De Bock, D., Depaepe, F., Janssens, D., & Verschaffel, L.
(2003). The Illusion of Linearity: Expanding the Evidence towards
Probabilistic Reasoning. Educational Studies in Mathematics , 53 (2), 113-
138.
Van Dooren, W., De Bock, D., Janssens, D., & Verschaffel, L. (2007). Pupils’
Over-Reliance on Linearity: A Scholastic Effect? British Journal of
Educational Psychology , 77, 307-321. doi: 10.1348/000709906X115967
Yaffee, R. A. (2003). Common Correlation and Reliability Analysis with SPSS for
Windows. Retrieved October 5, 2009, from
http://www.nyu.edu/its/statistics/Docs/correlate.html
248
Zaslavsky, O., Sela, H., & Leron, U. (2002). Being sloppy about slope: The effect
of changing the scale Educational Studies in Mathematics. Educational
Studies in Mathematics , 49 (1), 119-140.
249
APPENDIX A
INSTRUMENTS
250
1. The Mini-Diagnostic Test (MDT)
Problem 1
(a) Is the point with coordinates (2,-8) on the graph of the line y = 3x -14?
Problem 2
y = - 2x + 3
Create a table of values for the equation. Complete 5 rows of the table.
251
Problem 3
Which of the graphs below contains the points given in the table?
252
Problem 4
In the graph above, what is the slope of the line? Show your work.
253
Problem 5
The owner of a car dealership noticed a pattern in the weekly car sales, as shown
Which of the following equations represents the relationship between the number
Problem 6
Given this linear equation 2x + 3y = 12, which of the following is the y-intercept?
254
Problem 7
The graph below represents the distance that a car traveled after different
numbers of hours.
(a) Which part of the graph(R, S, or T) represents the hours when the car moved
the
slowest?
255
2. Information Questionnaire for Students
1. Name (print)
_____________________________________________________
____________________________________________________
____________________________________________________
256
3. Ranking Questionnaire for Students
Least difficult
4. For the problem you named the most difficult, tell us why you chose it.
257
4. Information Questionnaire for Teachers
2. College major?
6. Name (print)
7. School:
8. Phone:
9. Email:
258
5. Ranking Questionnaire for Teachers
2. Which of the problems will be most difficult for students? Least difficult?
3. For the two problems that you identified as most difficult for your students,
a) Most Difficult:
259
APPENDIX B
260
PART I: Scoring Guide (SG)
To score the MDT, use Table 51. Score each of the parts of MDT problem 1
(part 1a and part 1b) with 1 for a correct answer and 0 otherwise. The total score
for MDT problem 1 is obtained by adding the scores for parts 1a and 1b and
dividing the sum by 2. Score MDT problem 2 with 0, .2, .4, .6, .8 or 1, based on
Score MDT problems 3, 4, 5, 6, 7a, and 7b with 1 for a correct response and 0
otherwise. MDT problem 7c is separated into two parts, 7c1 and 7c2, one part
for the numeric value and the other for the unit, with 1 point for each correct
answer and 0 otherwise. The total score for MDT problem 7c is obtained by
adding the two scores for parts 7c1 and 7c2 and dividing the sum by 2.
Table 51
Use Tables 52-60 to record the frequencies (number of students) for the
261
Table 52
Students
262
Table 53
Students
7. No answer
Table 54
1. a
2. Correct answer b
3. c
4. d
263
Table 55
Students
3. Reversed ratio
5. Irrelevant information
Table 56
1. a
2. b
3. Correct answer c
4. d
264
Table 57
1. a
2. b
3. Correct answer c
4. d
Table 58
1. Answer “R”
2. Answer “S”
5. No answer
265
Table 59
2. Incorrect answer
Table 60
3. Incorrect answer
After solving the MDT problems, all students identify the most difficult
problem and explain why it is difficult. Teachers identify the two most difficult
problems for their students, and describe their suspicions about why students
266
Code student and teacher comments on the Ranking Questionnaires using the
description of the category and examples of what constitutes evidence for its
Table 61
Comments Comments
267
Table 62
Comments Comments
“Problem number 1 was the most difficult “Haven’t seen the material.”
because we have not learned this yet.”
“Graph interpretation , application
“I have never seen a problem like that of skills for graphing (not learned
before.” art this time).”
“I was absent when the teacher taught it.” “We hadn’t explained how to find
slope.”
268
Table 63
Comments Comments
“I didn't understand at first what x and y “If the students plug the #s into the
were supposed to be. I drew out a graph equation, they might know what to
and then it just clicked in my head the x do next –simplify, then they may
and y axis.” not know how to check if it is ‘true’
or not.”
“2 was the most difficult because you had
to figure out what numbers you had to put “Students struggle with getting to
in because it doesn't tell you.” make up values for x…they are used
to being given the numbers and then
“4, it wasn't a grid to help.” solving.“
Table 64
Comments Comments
“I couldn't remember the formula for “They don’t know where to start.”
slope.”
269
Table 65
Comments Comments
Table 66
concept, topic or domain, how to solve a problem, or the vocabulary used, without
Comments Comments
“I didn't know it.” “My students can solve it but the wording
is not clear.”
“Your wording was weird and I just
didn't understand it!” “Students never get it.”
270
Part II: Coding the Interviews
Students and teachers solve the MDT problems aloud, and comment on the
strategies used by students to solve MDT problems 1 and 2, and from teachers’
answers to two questions: “What do your students need to know in order to solve
this problem?” and “How would you help your students overcome difficulties
271
Table 67
Subcategories
CKC-C From Line to Point(s) A.“I just picked random numbers for x and
If a point is ( x 0 , y 0 ) on the graph of the then I plugged the numbers I chose into the
line l then its coordinates satisfy the equation and got the y values.”
equation of l, y mx b .
B.“You have a situation where you have an
There is an infinite number of points infinite amount of solutions and those
( x 0 , y 0 ) on the line l satisfying the solutions are all ordered pairs.”
equations of the line l, y mx b .
to the question, “What do you mean by slope?“ and from their explanations given
while solving problems 4, and 7b. Teachers’ knowledge of slope is inferred from
interview questions: “How would you explain what slope means to your
students?” and “What would you do to help your students overcome these
273
Student Difficulties with the Cartesian Connection from Point to Line (SDC-D)
are inferred from student schemes and teacher comments, and occur at each of the
Table 69
(MDT Problem 1)
274
Student Difficulties with the Cartesian Connection from Line to Points (SDC-
C) are inferred from student scheme and teacher comments, and occur at each of
Table 70
(MDT Problem 2)
Student Difficulties with Slope (SDS) are inferred from student answers and
geometric context, in MDT problem 4 and MDT problem 7b, are coded.
275
Table 71 presents the subcategories for Student Difficulties with Slope in an
Table 71
SDS-HM
Identifying the origin from A. “Slope is the rise over run it goes up four and it
where the measurement doesn’t go in any direction left or right, it’s 4.“
should start (e.g.,
considering the origin of the B.”Students don’t know where to start counting for the
system of coordinates as rise. They think they should start from the origin.”
starting point in measuring
the rise)
Identifying the unit of A. “And the run, I just counted how many it went over
measurement (e.g., until the line intersected it. So it would be one and a
estimating non-integer half. “
coordinates like the x-
intercept≈ 1.5) B.“ If we refer to the point where this line from
problem 4 intersects X-axis, they could give you many
different values. They could give you 0, they would tell
you that it’s (0, 1.5)…”
277
APPENDIX C
RECRUITMENT LETTERS AND IRB APPROVALS
278
1. Letter to Students
Dear Student:
We are conducting a study about students’ understanding of some topics in
algebra in order to discover better ways to support learning.
I am inviting you to participate in the study by completing a short 7-item quiz
about beginning ideas from algebra. The quiz will take about 30-40 minutes to
complete.
Your participation in this study is voluntary. The results will not affect your
grade. If you choose not to participate or to withdraw from the study at any time,
your course grade will not be affected. Your work on the quiz will remain
anonymous.
If you agree to participate in this study, please complete the information
below. You also will need to have your parent or guardian sign an agreement
form as well.
Return the signed part of the form to your teacher.
If you have any questions about this study, please contact me at the email or
phone number listed below.
Thanks in advance for your help.
Valentina Postelnicu
ASUP Phys Chem Applied Math
(480)727-1977
e-mail: [email protected]
If you have any questions about your participation in this research, or if you feel
you have been placed at risk, you can contact the Chair of the Human Subjects
Institutional Review Board, through the ASU Research Compliance Office at
(480) 965-6788.
------------------------------------------------------------------------------------------------
________________________________ ____________
(Signature ) (Date)
279
2. Parental Consent for Interviewing Minors
Valentina Postelnicu
Instructor, ASUP Phys Chem Applied Math
(480)727-1977, e-mail: [email protected]
If you consent to your child’s participation in this study, please complete the
section below and have your child return this form to the teacher as soon as
possible. A copy of this signed form will be returned to you. Thank you.
---------------------------------------------------------------------------------------------------
Student Name
280
If you have any questions about you or your child's rights as a subject/participant
in this research, or if you feel you or your child have been placed at risk, you can
contact the Chair of the Human Subjects Institutional Review Board, through the
ASU Research Compliance Office at (480) 965-6788.
281
3. Letter to Students for Interview
Dear Student:
We are conducting a study about students’ understanding of some topics in
algebra in order to discover better ways to support learning.
I am inviting you to be one of 10 students from your school to talk about how
you solved some of the problems on the algebra quiz that you already completed.
The interview will last about 20 minutes, will take place during school time, and
will be audio taped.
Your participation in this part of the study is voluntary and the results will not
affect your course grade. You may choose not to participate in or to withdraw
from the taping at any time. Again, that decision will not affect your course grade.
Transcriptions of the taping will not use your name. The audiotape will be
destroyed at the end of this study.
If you agree to be tape recorded, please complete the information below. You
also will need to have your parent or guardian sign an agreement form as well.
Return the signed part of the form to your teacher.
If you have any questions about this study, please contact me at the email or
phone number listed below.
Thanks in advance for your help.
Valentina Postelnicu
ASUP Phys Chem Applied Math
(480)727-1977, e-mail: [email protected]
If you have any questions about your participation in this research, or if you feel
you have been placed at risk, you can contact the Chair of the Human Subjects
Institutional Review Board, through the ASU Research Compliance Office at
(480) 965-6788.
----------------------------------------------------------------------------------------------
(Printed Name)
________________________________ ____________
(Signature) (Date)
282
4. Letter to Teachers
Dear Colleague,
We need your help. The School of Educational Innovation and Teacher
Preparation at Arizona State University, Polytechnic campus, is focusing on ways
to enhance student achievement in mathematics. As a first step, we examined
student performance on state and other standardized tests. This analysis revealed
that students in Grades 8-10 have particular difficulty with some fundamental
topics in algebra. To determine how well our instructional programs address these
important concepts, we developed Mini-Diagnostic Tools, (MDT). The first of
these, MDT -1, assesses students’ understanding of lines, how lines are
represented with words, tables, graphs and symbols. MDT-1 consists of 7
problems with formats much like those used in the AIMS test.
MDT-1 can be administered to students in one class period. MDT-1 should
take no more than 35 minutes for its completion. We would like you to administer
MDT-1 to your Grade 8, 9 or 10 students.
To gain clearer understanding of students’ solution methods, after the
administration of MDT-1, we would like to conduct interviews with you and 10
students from your school . Interviews will be approximately 20-30 minutes in
length and can be conducted at the school sites. Interviews will require students
who completed the quiz to orally recount their solution processes to some of the
items. Interviewers will be members of the project staff, all of whom are
experienced teachers and are either faculty or graduate students in mathematics
education in the School of Educational Innovation and Teacher Preparation.
Prior to the administration of the MDT-1, signed permission forms must be
received from all students and their parents or guardians. For the audio-taped oral
recounting of solution processes, we will need a separate form signed by parents
or guardians. No teacher or student names will be collected. Audio tapes will be
shared only with the project staff and will be destroyed after the completion of the
project.
If you would like additional information, please contact me or Valentina
Postelnicu. Email is the best way to reach us.
With great appreciation and best wishes,
Carole Greenes, Ed.D.
Dean and Professor Mathematics Education, School of Educational Innovation
and Teacher Preparation, Arizona State University at the Polytechnic campus
(480) 727-1820,
e-mail: [email protected]
Valentina Postelnicu
Instructor, ASUP Phys Chem Applied Math, (480)727-1977,
e-mail: [email protected]
283
5. IRB Approvals
284
285
286