Student Difficulties with Linearity and Linear Functions and

Teachers' Understanding of Student Difficulties

by

Valentina Postelnicu

A Dissertation Presented in Partial Fulfillment

of the Requirements for the Degree
Doctor of Philosophy

Approved April 2011 by the

Graduate Supervisory Committee:

Carole Greenes, Chair

Victor Pambuccian
Finbarr Sloane

ARIZONA STATE UNIVERSITY

May 2011
 UMI Number: 3450231

All rights reserved

INFORMATION TO ALL USERS

The quality of this reproduction is dependent upon the quality of the copy submitted.

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

The focus of the study was to identify secondary school students'

difficulties with aspects of linearity and linear functions, and to assess their

teachers' understanding of the nature of the difficulties experienced by their

students. A cross-sectional study with 1561 Grades 8-10 students enrolled in

mathematics courses from Pre-Algebra to Algebra II, and their 26 mathematics

teachers was employed. All participants completed the Mini-Diagnostic Test

(MDT) on aspects of linearity and linear functions, ranked the MDT problems by

perceived difficulty, and commented on the nature of the difficulties. Interviews

were conducted with 40 students and 20 teachers. A cluster analysis revealed the

existence of two groups of students, Group 0 enrolled in courses below or at their

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

in student performance on the MDT across grades. Student performance on the

MDT increased with more advanced courses, mainly due to Group 1 student

performance. The most difficult problems were those requiring identification of

slope from the graph of a line. That difficulty persisted across grades,

mathematics courses, and performance groups (Group 0, and 1). A comparison of

student ranking of MDT problems by difficulty and their performance on the

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

Cartesian connection. Teachers' descriptions of problem difficulty identified

factors such as lack of familiarity with problem content or context, problem

format and length. Teachers did not identify student difficulties with slope in a

geometric context.

ii
 ACKNOWLEDGMENTS

Foremost, I would like to express my deepest gratitude to my advisor Professor

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

improving mathematics education for every student continue to motivate me and

provide an inspirational model for me as a mathematics educator.

I am forever grateful to Professor Finbarr Sloane who helped me every step of

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

my results. I shall always remember his warm encouragement throughout the

years.

I am indebted to Professor Victor Pambuccian who helped me with the

mathematics and the history of mathematics content of my thesis. I feel privileged

to have had access to his vast reservoir of knowledge, and I am grateful for his

constant input during the course of this thesis.

It is an honor for me to thank all my professors from Arizona State University.

I learned so much from all of them.

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

be a place for me in mathematics education.

iii
 Lastly, I would like to thank my husband Florin, my daughters Iris and

Eveline, and my mother for their unconditional love and trust.

The research reported here was supported in part by a grant from the

National Science Foundation (NSF 0833760).

iv
 TABLE OF CONTENTS

Page

LIST OF TABLES...................................................................................................... ix

LIST OF FIGURES .................................................................................................. xiv

CHAPTER

1 OVERVIEW OF THE STUDY ............................................................ 1

1.1 Introduction .................................................................................. 1

1.2 Definition of Terms ...................................................................... 3

1.3 Rationale ....................................................................................... 6

1.3.1 The Need for More Scientists ........................................... 6

1.3.2 Algebra, the Primary Course in the Pipeline to STEM

Careers ........................................................................................ 7

1.3.3 Addressing Failure with Algebra I .................................. 9

1.3.4 Studies of Student Difficulties with Linearity and Linear

Functions in Grades 8-10 ......................................................... 11

1.3.5 Studies with other Populations ....................................... 14

1.3.6 Studies of Teacher Understanding ................................. 18

2 REVIEW OF LITERATURE ............................................................. 21

2.1 Introduction ................................................................................. 21

2.2 Algebra as a Body of Knowledge: Historical Perspective ....... 21

2.3 Algebra in the School Curriculum ........................................ 24

2.3.1 Algebra in the School Curriculum: Historical Perspective

.................................................................................................... 24

v
CHAPTER Page

2.3.2 School Algebra from a Curriculum Perspective ........... 28

2.4 Research on Student Understanding of Linearity and Linear

Functions ......................................................................................... 30

2.4.1 Studies on Student Understanding of Linearity and Linear

Functions ................................................................................... 30

2.4.2 Teaching Experiments on Student Learning of Linearity

and Linear Functions ................................................................. 42

2.4.3 Studies on Student Understanding of Slope or Rate of

Change in Other Content Areas ................................................ 50

2.5 Research on Teacher Knowledge of Student Understanding of

Linearity and Linear Functions ........................................................ 52

2.6 Student and Teacher Identification of Problem Difficulty ....... 65

2.7 Concluding Remarks ................................................................. 76

3 DATA COLLECTED AND METHODS OF ANALYSES .............. 79

3.1 Introduction ................................................................................. 79

3.2 Study Instruments ....................................................................... 80

3.3 Theoretical Framework .............................................................. 84

3.4 Pilot Study................................................................................... 86

3.4.1 Scoring Guide (SG) for the MDT ................................... 87

3.4.2 Preliminary Coding Categories for Ranking

Questionnaires .......................................................................... 95

3.4.3 Preliminary Coding Categories for Interviews .............. 97

vi
CHAPTER Page

3.5 Participants ........................................................................... 104

3.5.1 Student Demographics .................................................. 105

3.5.2 Teacher Demographics ................................................. 109

3.6 Timeline for Instrument Development, Recruitment of

Participants, Data Collection and Analyses ................................... 111

3.7 Methods of Analyses ................................................................ 113

4 DATA ANALYSES .......................................................................... 117

4.1 Introduction ............................................................................... 117

4.2 Research Question 1 ............................................................... 119

4.2.1 Inter-Rater Agreement and Instrument Reliability ..... 119

4.2.2 Student Performance on the MDT by Grade Level .... 122

4.2.3 Student Performance on the MDT by Current

Mathematics Course ............................................................... 125

4.2.4 MDT Student Scores by Performance Group .............. 130

4.2.5 Student Performance on the MDT by Problem ........... 134

4.3 Research Question 2 ............................................................... 157

4.4 Research Question 3 ............................................................... 161

4.5 Research Question 4 ................................................................. 165

4.5.1 Coding and Identifying the Nature of Student Difficulties

from RQS Written Comments ................................................ 165

4.5.2 Coding and Identifying Students’ Mathematical

Difficulties from Interviews .................................................... 176

vii
CHAPTER Page

4.6 Research Question 5 .............................................................. 195

4.6.1 Teachers’ Assessment of the Nature of Problem

Difficulty ................................................................................. 196

4.6.2 Teachers’ Comments on Student Difficulties with the

Cartesian Connection .............................................................. 197

4.6.3 Teachers’ Comments on Student Difficulties with Slope

in a Geometric-Analytical Context ........................................ 202

5 DISCUSSIONS AND CONCLUSIONS ......................................... 209

5.1 Introduction ............................................................................... 209

5.2 Results and Discussion .......................................................... 210

5.2.1 Research Question 1 Results and Discussion............... 211

5.2.2 Research Question 2 Results and Discussion............... 214

5.2.3 Research Question 3 Results and Discussion............... 216

5.2.4 Research Question 4 Results and Discussion............... 217

5.2.5 Research Question 5 Results and Discussion............... 223

5.3 Limitations of the Study ......................................................... 227

5.4 Recommendations for Future Research ................................. 230

REFERENCES ...................................................................................................... 233

APPENDIX

A INSTRUMENTS ............................................................................. 250

B SCORING GUIDE AND CODING MANUAL ............................ 260

C RECRUITMENT LETTERS AND IRB APPROVALS................ 278

viii
 LIST OF TABLES

Table Page

1. MDT Scores by Problem .................................................................... 89

2. MDT Problem 1 Types of Responses ................................................. 90

3. MDT Problem 2 Types of Responses ................................................. 92

4. MDT Problem 4 Types of Responses ................................................ 93

5. MDT Problem 7a Types of Responses ............................................. 94

6. MDT Problem 7b Types of Responses ............................................. 94

7. MDT Problem 7c Types of Responses ................................................ 95

8. Preliminary Coding Categories for RQ Written Responses .............. 96

9. Preliminary Coding Categories for Conceptual Knowledge of

Cartesian Connection ........................................................................ 98

10. Preliminary Coding Categories for Conceptual Knowledge of

Slope .................................................................................................. 99

11. Preliminary Coding Categories for Student Difficulties with the

Cartesian Connection ...................................................................... 100

12. Preliminary Coding Categories for Student Difficulties with

Slope .............................................................................................. 103

13. Demographics by School District ................................................... 105

14. Student Distribution by District and School.................................... 106

15. Student Distribution by Grade and Current Mathematics Course .. 107

16. Participating Students by Current Mathematics Course, School, and

Teacher ............................................................................................. 108

ix
Table Page

17. Participating Teachers’ Experience and Degree ............................ 110

18. Timeline for Instrument Development, Recruitment of Participants,

Data Collection and Analyses ......................................................... 111

19. Inter-Rater Agreement on Scoring of Responses ........................... 120

20. MDT Performance by Grade Level ................................................ 122

21. MDT Student Performance by Current Mathematics Course ....... 125

22. Algebra I Students’ Performance on the MDT by Grade ............... 129

23. MDT Scores by Performance Level ............................................... 132

24. Means and Standard Deviations for Scores on MDT Problems .... 136

25. Correlations among Scores on the MDT Problems (MicroFACT) 138

26. Correlations of Problems with MDT Total Score .......................... 139

27. Initial Eigenvalues ........................................................................... 140

28. VARIMAX Rotated Factor Loadings (Three Factors) .................. 143

29. Means of Scores of MDT Problems by Group of Students ........... 144

30. Unrotated Factor Loadings ............................................................. 145

31. VARIMAX Rotated Factor Loadings (Two Factors) ..................... 147

32. Labeled Factor Matrix ................................................................... 149

33. Rotated Factor Loadings for Group 0 ............................................. 150

34. Rotated Factor Loadings for Group 1 ............................................. 150

35. Mean Scores for MDT Problems by Grade for Group 0 ............... 153

36. Mean Scores for MDT Problems by Current Mathematics Course for

Group 0 .......................................................................................... 153

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

Group .............................................................................................. 155

39. Group 0 Performance on the MDT vs. Ranking of Problem Difficulty

by Students ....................................................................................... 158

40. Group 1 Performance on the MDT vs. Ranking of Problem Difficulty

by Students ........................................................................................ 159

41. Student Performance on the MDT vs. Ranking of Problem Difficulty

by Teachers ...................................................................................... 162

42. Final Coding Categories for RQ Written Responses ..................... 166

43. Inter-Rater Agreement on RQ Written Responses ......................... 168

44. Group 0 Student Difficulties with MDT Problems 2 and 7b ......... 171

45. Group 1 Student Difficulties with MDT Problems 1 and 4 ............ 174

46. Final Coding Categories for Conceptual Knowledge of Cartesian

Connection ...................................................................................... 177

47. Final Coding Categories for Conceptual Knowledge of Slope ..... 178

48. Final Coding Categories for Student Difficulties with the Cartesian

Connection ...................................................................................... 179

49. Final Coding Categories for Student Difficulties with Slope ........ 180

50. Teacher Assessment of Problem Difficulty ................................... 196

51. Scoring Guide for MDT Scores by Problem .................................. 261

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

61. Category H: Hard Problem ............................................................. 267

62. Category L: Never Learned /Taught ............................................... 268

63. Category M: Mathematical Difficulty ............................................. 269

64. Category R: Not Remembering ...................................................... 269

65. Category T: Thinking Time ............................................................ 270

66. Category U: Not Understanding ..................................................... 270

67. Conceptual Knowledge of the Cartesian Connection (CKC) and its

Subcategories .................................................................................. 272

68. Conceptual Knowledge of Slope (CKS) and its Subcategories ...... 273

69. Student Difficulties with Cartesian Connection from Point to Line

SDC-D (MDT Problem 1) .............................................................. 274

70. Student Difficulties with Cartesian Connection from Line to Point

SDC-C (MDT Problem 2) .............................................................. 275

xii
Table Page

71. Student Difficulties with Slope in Homogeneous Systems of

Coordinates SDS (MDT Problem 4) ............................................... 276

72. Student Difficulties with Slope in Non-Homogeneous Systems of

Coordinates SDS-N(MDT Problem 7b) ......................................... 277

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

2a. Histogram of MDT Scores Distribution for Pre-Algebra

Students ........................................................................................... 127

2b. Histogram of MDT Scores Distribution for Algebra I

Students ........................................................................................... 127

2c. Histogram of MDT Scores Distribution for Geometry

Students ............................................................................................ 128

2d. Histogram of MDT Scores Distribution for Algebra II

Students ........................................................................................... 128

3a. Silhouette Measures of Cohesion and Separation - Clustering by

Grade ............................................................................................... 131

3b. Silhouette Measures of Cohesion and Separation - Clustering by

Current Mathematics Course .......................................................... 131

3c. Silhouette Measures of Cohesion and Separation - Clustering by

Performance Group .......................................................................... 131

4a. Histogram of MDT Score Distribution for Group 0 Students ....... 133

4b. Histogram of MDT Score Distribution for Group 1 Students ....... 133

5. Scree Plot of the Initial Eigenvalues ................................................ 141

6. MDT Problems 2 and 7 ..................................................................... 170

xiv
Figure Page

7. MDT Problems 1 and 4 .................................................................... 173

8. Example 1 of Student Answer to MDT Problem 1 .......................... 183

9. Example 2 of Student Answer to MDT Problem 1 .......................... 183

10. Example 3 of Student Answer to MDT Problem 1 .......................... 184

11. Example 1 of Student Answer to MDT Problem 2 ........................... 185

12. Example 2 of Student Answer to MDT Problem 2 .......................... 187

13. Example 1 of Student Answer to MDT Problem 4 .......................... 190

14. Example 2 of Student Answer to MDT Problem 4 .......................... 191

15. Example of Student Answer to MDT Problem 7 ............................. 193

xv
 Chapter 1

OVERVIEW OF THE STUDY

1.1 Introduction

Algebra I has been referred to as the “gateway to a technological future”

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.

Those difficulties have persisted despite an increase in remediation efforts

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

16% to 54%, with higher percentages associated with socio-economically

disadvantaged students (Bracey, 2000; Dietz, 2010; Gamoran & Hannigan,

2000; Helfand, 2006 ; Johnson, 2008).

In efforts to gain insight into the nature of difficulties with algebra, some

researchers have studied students’ understanding of the fundamental concepts,

linearity and linear functions (Ben-Chaim, James, Fitzgerald, & Benedetto,

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

researchers have studied teachers’ understanding of students’ algebraic thinking

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

of difficult problems for students do not match students’ perceptions of

difficulty or their performance (Greenes et al., 2007b).

Statement of the Problem. The dual purpose of this study was to identify

the difficulties students in Grades 8, 9, and 10 experience while solving

problems that require understanding of aspects of linearity and linear functions,

and to assess mathematics teachers’ understanding of the nature of the

difficulties experienced by their students.

Research Questions. This study addressed the following questions:

1. Does student performance on the Mini-Diagnostic Test (MDT) vary by:

a. Grade level?

b. Current mathematics course?

c. Problem?

2. Does student ranking of problems by difficulty, after completing the

MDT, correspond to their performance on the MDT?

3. Does teacher ranking of problems by difficulty, after completing the

MDT, correspond to student performance on the MDT?

4. What is the nature of students’ difficulties?

2
 5. Do teachers’ speculations about the nature of students’ difficulties

correspond to actual student difficulties as identified from student

performance on the MDT and interviews?

1.2 Definition of Terms

Linearity. To a mathematician, linearity refers to the property of a function

to be homogenous and additive, i.e., f (mx) mf ( x), f ( x1  x2 ) f ( x1 )  f ( x2 ) .

The function f : ƒ o ƒ, f ( x) mx, m z 0 (or y mx ) fulfills both conditions

and can be seen as a relationship of direct variation with a constant rate of

change m, or a relationship of proportionality with the constant of

a c
proportionality m. A proportion can be seen as a relationship of direct
b d

variation, or linearity between two variables (Freudenthal, 1983). Linearity

appears as direct variation or direct proportionality in Algebra I textbooks used

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

Linear Function. A linear function is represented in Algebra I programs in

the form y mx  b , and is referred to as a linear equation of two variables, the

slope-intercept form of the equation of a line, or simply the equation of a line.

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).

Graph of a Linear Function. The graph of a linear function f ( x) mx  b

or y mx  b (the slope-intercept-form) is a line. When b 0 , the line contains

the origin of the Cartesian system of coordinates. When b z 0 , the line crosses

the Y-axis at (0, b) , a point different from the origin.

X-intercept, Y-intercept. The x-intercept (or the “zero”) of the function

f ( x) mx  b is found by solving the equation f ( x) 0 ; the solution is

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

X-axis. The y-intercept of the function f ( x) mx  b is found by calculating the

value of the function f (0) ; the solution is f (0) b . The point (0, b) represents

the point of intersection. In Algebra I textbooks, definitions for the x-intercept

and y-intercept refer to their graphic representation; the x-coordinate of a point

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

in Burger et al. (2007, p. 296) and Collins et al. (1998, p. 346).

Slope or Rate of Change. The slope of a line is defined in Algebra I

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

“ratio,” “change,” and “constant rate of change,” in their definitions. For

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

of the change in the y-coordinate to the corresponding change in the x-

coordinate” (p. 326). Stump (1997) identifies five different contextual

definitions of slope: geometric, algebraic, physical, trigonometric, and

functional. Geometrically, slope is defined as a property of a line. Given two

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

linear equation; slope is the m in the equation y mx  b . Slope describes the

steepness of physical objects such as mountain roads, ski slopes, and wheelchair

ramps. Slope is defined trigonometrically as the tangent of the angle of

inclination T ; slope is m tan T . When linear functions, of the form

y mx  b , are understood to be composed of two variables, x and y, that have

'y
a dynamic relationship, slope represents the constant rate of change m
'x

(Stump, 1997, p. 4).

Cartesian Connection. The Cartesian connection can be stated

mathematically with the following theorem: A point ( x 0 , y 0 ) is on the graph of

the line l if and only if its coordinates satisfy the equation of l, y mx  b

5
(Knuth, 2000; Moschkovich et al., 1993; Schoenfeld et al., 1993). Schoenfeld et

al. (1993) consider all mathematical conventions associated with graphic

representations in a Cartesian system of coordinates to be part of the Cartesian

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

segments with oriented magnitudes, rise and run.

1.3 Rationale

1.3.1 The Need for More Scientists. To compete successfully internationally

with scientific and technological advances, the United States must increase its

innovation capacity and produce more experts in mathematics, technology,

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

(2005) show an increasing number of advanced degrees in science and

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-

Higher Education Forum, 2005, p. 3).

The national need for mathematicians and scientists places great responsibility

on elementary through high school and college programs to engage more students

in the study of these subjects (Committee for Economic Development, 2003;

6
Public Works, 2006). National and state organizations interested in increasing the

number of workers with 21st century skills have recommended that students take

increasingly challenging mathematics courses throughout their school programs,

and complete, at minimum, Algebra I, Geometry, and Algebra II before high

school graduation (ACT, 2006; American Diploma Project, 2004; Arizona

Community Foundation, 2008; Bozick & Lauff, 2008; Darling-Hammond, 2007;

Education Week, 2006; National Center for Public Policy and Higher Education,

2004; Partnership for 21st Century Skills, 2007). Research studies examining the

relationship of high school course-taking histories to graduation from high

school, entrance and graduation from college, and successful entry into the

workforce, found that two courses, Algebra I and Geometry, had the highest

correlation (Pelavin & Kane, 1990; Shettle et al., 2007).

1.3.2 Algebra, the Primary Course in the Pipeline to STEM Careers.

Several professional associations, including the National Council of Supervisors

of Mathematics (NCSM) (1977), the National Council of Teachers of

Mathematics(NCTM) (1980; 1989; 2000) and the Mathematical Association of

America (Katz, 2007a), have recognized the value of algebra to the study of

higher level mathematics, physics, the natural sciences, engineering, and

technology (Katz, 2007a; Ma, 2009; Sadler & Tai, 2007). In its Position Paper

on Basic Mathematical Skills, the NCSM (1977) identifies algebra as a pivotal

discipline in the school mathematics curriculum. In its Principles and Standards

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

think algebraically to the successful study of other mathematical domains

including geometry, probabilities and statistics, calculus.

In 1992, the National Assessment of Educational Progress (NAEP) reported

that Grade 8 students enrolled in Algebra I had higher scores in mathematics on

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

general mathematics courses (National Center for Education Statistics, 1996).

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

movement was born (Pennsylvania Council of Teachers of Mathematics, 1997).

Throughout the United States, all students in Grades 8 or 9 were required to

enroll in Algebra I. From 1996 to 2005, the proportion of U.S. students taking

Algebra I in Grade 8 rose from 24% to 34% (Cavanagh, 2008). By 2007,

nationwide 31% of Grade 8 students were enrolled in Algebra I (Loveless,

2009). These percentages are expected to rise as more states, following

Minnesota’s and California’s examples, requiring all Grade 8 students to enroll

in Algebra I (Jacobson, 2008; Loveless, 2009).

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 &

Izsák, 2008; Dietz, 2010).

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

was 10% higher for economically disadvantaged (26.6%) versus economically

non-disadvantaged students (16.7%) (Johnson, 2008). In 2009, 55% of New

Jersey students (Carroll, 2009), and 84% of Clark County, California students

failed end-of-semester Algebra I tests (Richmond, 2009).

Algebra I is foundational to the study of subsequent high school and college

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

National Mathematics Advisory Panel as major topics of school algebra

(National Mathematics Advisory Panel, 2008). As noted by Kalchman and

Koedinger (2005), understanding the x-intercept, y-intercept and how to graph

linear functions, offers to students the possibility of comparison, differentiation,

9
and generalization to other functions like quadratic, exponential, and

trigonometric functions, and enhances their learning of those non-linear

functions. Not only are these concepts important for preparation for higher level

mathematics, they also are central to the solution of problems involving

measurement conversion and similarity in the pre-secondary school curriculum,

linear models and approximations in calculus and statistics, and linear

programming and vectorial spaces in secondary and post-secondary courses (De

Bock, Van Dooren, Janssens, & Verschaffel, 2002). As noted by Thompson

(1994a, 1994b), a mature conceptual understanding of rate of change as

“reflective abstract ratio” has the potential to be foundational for the

understanding of calculus concepts.

Many educators believe that student difficulties with Algebra stem from the

instructional methods and curriculum itself. To better prepare Algebra students

the curriculum “must simultaneously develop conceptual understanding,

computational fluency, and problem-solving skills" recommends the National

Mathematics Advisory Panel (2008, p. xix ). Despite long-term efforts to

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

in Grades 8-10. Several researchers studied student difficulties with key

concepts of Algebra I, particularly slope or rate of change (Lobato & Siebert,

2002), the x- and y-intercepts (Moschkovich, 1992, 1999; Moschkovich et al.,

1993; Schoenfeld et al., 1993; Smith, Arcavi, & Schoenfeld, 1989), and linear

relationships (Ellis, 2007, 2009).

Lobato and Siebert (2002) examined student understanding of slope in

physical situations as a measure of steepness of a ramp, and slope in functional

situations as a measure of speed or rate of change. Nine students from Grades 8-

10 with average mathematical performance participated in this study. Analyses

revealed student difficulties in making connections between the physical aspect

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

and slope as a measure of steepness may be expressed by the same formula.

Lobato and Siebert’s study pointed to the importance of context in the

conceptualization of slope, and student difficulties transferring knowledge

between contexts, often overlooked by curriculum developers and instructors.

Researchers from the Functions Group at Berkeley used a graphing software

tool specifically designed as an environment for the learning of functions, to

study Grades 8-10 students’ abilities to make connections among various

representations of linear functions. Participants were observed and interviewed

while interacting with the software during various tasks (Schoenfeld et al.,

11
1993), and observed during their classroom work and participation in peer

discussions (Moschkovich 1992, 1999; Moschkovich et al., 1993). Findings

highlighted the crucial importance of the Cartesian connection (Moschkovich et

al., 1993) and revealed student difficulties with making connections among

representations. The studies mentioned above (Lobato & Siebert, 2002;

Moschkovich, 1992, 1999; Moschkovich et al., 1993; Schoenfeld et al., 1993;

Smith et al., 1989) were conducted by researchers in university laboratories. It

also should be noted that classroom teachers, indispensible to the natural

teaching-learning environments, did not participate in any of these studies.

Davis (2007) analyzed eight sophomore and junior students’ understanding

of the y-intercept in real-world contexts and across multiple representations over

a period of 10 weeks of instruction while they were enrolled in Algebra II.

During this time, and with an approach that delayed the introduction of formal

mathematical terminology, and allowed students to use their own terminology,

Davis observed and interviewed the students. Analyses of student performance

revealed a weak connection between various representations of the y-intercept

and the use of student-invented terminology. Davis explained these findings as

being influenced by students’ prior knowledge of the y-intercept in real contexts,

and the teacher’s conception of the y-intercept and his assumptions that students

would have no difficulty understanding the relationship between the formal

mathematical language and the informal language suggested by the real-world

contexts. The use of informal phrase “starting point” for the y-intercept

suggested by the real-world contexts, was not be translated to symbolic, tabular,

12
and graphic representations, and seemed to have a negative impact on student

performance on abstract tasks, i.e., tasks without real-world contexts. Although

the sample was small, the study showed that student difficulties with connecting

among various representations persisted even with instruction. The relationship

between student difficulties and teacher understanding of these difficulties was

analyzed by Davis who pointed to the teacher’s insensitivity to students̓

difficulties. Davis concluded that the teacher’s unquestioned assumptions about

student abilities to make connections between representations and transfer

knowledge of the y-intercept between contexts, as well as lack of appropriate

instructional materials, impacted negatively on student learning.

In summary, studies on our target population (Grades 8-10) suggest that

linearity and linear functions are complex topics, and student difficulties occur

most often with making connections among various representations of linear

relationships and functions (Davis, 2007; Moschkovich, 1992, 1999;

Moschkovich et al., 1993; Schoenfeld et al., 1993; Smith et al., 1989).

Furthermore, misconceptions are robust (Moschkovich, 1992, 1999;

Moschkovich et al., 1993; Schoenfeld et al., 1993; Smith et al., 1989). With the

exception of one study referring to teacher understanding of student difficulties

with the y-intercept (Davis, 2007), all other studies addressed only student

difficulties with aspects of linearity and linear functions, as identified by

researchers. None of the studies asked students to explain the nature of their

difficulties, or compared student difficulties with teachers’ insights into 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

school students, college students, and pre-service teachers, focused on

difficulties with the concepts of slope and rate of change, connections among

various representations of functions, and the modeling of phenomena or real

world situations using functions. Research on proportional reasoning, considered

to be equivalent to linear reasoning (De Bock, Verschaffel, & Janssens, 1998;

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

difficulty with reasoning proportionally, and difficulties persist throughout high

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

in order to identify difficulties and errors exhibited as the students completed

tasks focusing on the concepts of variable, linear equation, and linear function.

After analyzing students’ systematic errors, faulty algorithms, and

misconceptions, a comparison between high-achieving students and low-

achieving students pointed to the possible causes of students’ misconceptions.

The high-achieving students had an object or structural view of the

mathematical concepts while the low-achieving students had a process view.

For example, when asked to solve problems like, “Look at the equation

y 2 x  5 ; if x increases by 1, what happens to y?” (Li, 2006, p. 169), students

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

linear function would make a prompt judgment by looking at slope. Li suggested

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.

At the Precalculus level, two studies are of particular relevance. Stump

(2001a) interviewed 22 Precalculus students to gain insight into their

understanding of slope as a measure of steepness in physical situations, and as a

measure of rate of change in functional situations. Findings confirmed that

student difficulties with the concept of slope are robust. In physical situations,

students used angles instead of ratios as a measure of steepness. They performed

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

informal language in describing steepness (e.g., “angle”, “incline”, “height”),

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.

Hauger (1997) conducted interviews of four high school students enrolled in

a Precalculus course to verify student difficulties with the concept of rate of

change previously reported in the literature, and to confirm the importance of

the linguistic and graphic connection when modeling change with functions.

Hauger’s students had difficulty making connections between linguistic (e.g.,

“steady pace”) and graphic representations of functions. Often, the language

15
used by teachers in instruction or the textbooks themselves contains words and

expressions loaded with meanings considered so transparent that they do not

need explanation. Yet these terms or phrases remain opaque to students. For

example, the constant rate of change of a linear function represented graphically

by a static line might suggest to students that no change occurs.

Zaslavsky et al. (2002) hypothesized that the connection between the

symbolic (algebraic/ analytic) and the graphic (geometric/ visual) aspects of

slope becomes problematic when non-homogenous systems of coordinates are

used. The 124 participants in their study included Grade 11 students, prospective

and in-service secondary mathematics teachers, and university mathematics

educators and mathematicians. Participants were interviewed about two tasks

requiring them to analyze slope appearing as an analytic attribute of a linear

function, and as a geometric attribute of the line representing the graph of the

function. Task 1 required participants to identify the slope of a line graphed in a

homogenous system of coordinates. Task 2 required participants to identify the

slope of a line graphed in a non-homogeneous system of coordinates. Results

revealed that participants with strong mathematical backgrounds could reconcile

the analytic and the geometric perspectives, while others experienced difficulties

in connecting the two perspectives or favored only one of them. This study

draws attention to the unquestioned assumptions behind the graphic

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.

A longitudinal study of several years duration, focusing on participants’

understanding of the use of graphs to represent physical concepts or real–world

situations, was conducted with college students, high school students and pre-

service teachers enrolled in physics courses (McDermott, Rosenquist, & Van

Zee, 1987). The study revealed that the nature of the participants’ difficulties

with graphing were common across the various populations studied. In

particular, participants had difficulty connecting graphs to physical concepts and

to real world situations.

The National Survey of Algebra Teachers (Hoffer T. B., Venkataraman,

Hedberg, & Shagle, 2007) points out that solving word problems is considered

by Algebra teachers to be their students’ poorest area of preparation, but does

not reveal more on the nature of student difficulties. As indicated by

McDermott et al.’s (1987) study, the source of the trouble may be the difficulty

of connecting between situations, concepts, and their representations.

Studies by Stump (1997, 2001a, 2001b), Hauger (1997), Zaslavsky et al.

(2002), McDermott et al. (1987) described the difficulties experienced with

linear functions by various populations (high school students, pre-service

teachers, and in-service teachers). The prevalence of the same difficulties in

these populations as well as the potential deleterious effect on the study of

physics, further reinforces the need for the proposed study. Moreover,

difficulties with slope and connecting among representations seem to be

17
common to both students and teachers, and raise the question of teachers’

understanding of these concepts as well as their students’ difficulties with these

concepts.

1.3.6 Studies of Teacher Understanding. Research on student and teacher

understanding of proportionality (Tourniaire & Pulos, 1985) and the concept of

function, in general, is rich and reveals students’ and teachers’ difficulties

(Breidenbach, Dubinky, Hawks, & Nichols, 1992; Dubinsky & Harel, 1992;

Hansson, 2006; Norman, 1992; Sfard, 1992; Sierpinska, 1992; Slavit, 1997;

Vinner, 1992). However, the research on teacher understanding of student

difficulties, specifically with linearity and linear functions, is scarce. Several

studies confirm the strong correlation between teachers’ subject matter

knowledge and pedagogical knowledge (Stump, 1997; You, 2006) and the

importance of teacher knowledge on addressing student difficulties successfully

(Ding, 2007).

Using interviews and surveys, Stump (1997) examined the mathematical and

pedagogical knowledge of slope of 39 pre-service and in-service secondary

mathematics teachers. Results showed that most of the teachers had a geometric

view of the concept of slope. The algebraic, physical, functional, trigonometric

or ratio aspects of slope were less mentioned in their descriptions of the concept.

Teachers’ pedagogical knowledge mirrored their subject matter knowledge.

They often mentioned student difficulties related to the geometric representation

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.

Learning cannot be separated from teaching, and many of the studies

mentioned in this chapter focused solely on students’ understandings and

difficulties with linearity and linear functions, irrespective of teachers’

understanding of the nature or extent of these difficulties. The present study was

conducted to reveal specific patterns in students’ understandings and difficulties

with linearity and linear functions and to examine teachers’ understanding of

their students’ difficulties. The results of this study have the potential of better

informing researchers and mathematics educators in the design of instructional

materials as well as interventions for addressing the problematic aspects of

students’ difficulties with linearity and linear functions.

Moreover, results of this study have the potential of contributing to teachers’

understanding of students’ difficulties with fundamental algebraic ideas in order

to enable them to better respond to the mathematical talents and needs of their

students.

Chapter Summary. This chapter presented an overview of the study of

student difficulties with linearity and linear functions and teachers’

understanding of student difficulties, followed by the statement of the problem

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

of linearity and linear functions, and compare those to teachers’ assessments of

their students’ difficulties.

In this chapter, an historical perspective of algebra as a body of knowledge

(Section 2.2) and algebra as a course or strand in the school curriculum are

presented first (Section 2.3). This is followed by a review of the research

literature on student difficulties and teacher understanding of rate of change, the

fundamental concept of linearity and linear functions, together with research

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).

2.2 Algebra as a Body of Knowledge: Historical Perspective

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”

(Euler, 1984, p. 270). Cooke (1997) elaborated on Euler’s definition to consider

parts of algebra to be problems that require finding unknown quantities from

given quantities without indicating what operations should be performed on the

given quantities, even in the absence of symbolic representations.

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,

problems involving linear and quadratic relationships were solved not by

formulas, but by examples. Babylonian clay tablets, dated from the twentieth

century B.C. to the seventieth century B.C., described solutions to problems

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,

2007b). This stage in the development of algebra, known as the non-symbolic

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

finding quantities to satisfy certain conditions.

The development of algebra as a body of knowledge was marked throughout

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

Greek idea of intersecting conics. In the sixteenth century, Cardano (1501-1576)

published Ars Magna (1545) which contained algorithms for finding solutions to

third and fourth degree polynomial equations that were based on geometric

proofs. Throughout this non-symbolic stage of algebra, since the Akkadian

22
Empire until the European Renaissance, the justifications for the algorithms

proposed for solving equations appeared to be mainly geometric in nature.

The rise of algebra during the European Renaissance, between the fourteenth

and seventeenth centuries, marked the entrance of the symbolic stage with

François Viète’s (1540-1603) work on equations, characterized by the use of

symbols to represent quantities (e.g., vowels for unknown quantities and

consonants for known quantities). Over time all quantities and all operations and

relationships involving quantities were symbolized, culminating with the

development of rules for symbolic manipulation.

The shift from the static approach for solving equations to the dynamic

function approach developed during the seventeenth century with the work of

Galileo (1564-1642), Kepler (1571-1630), and Newton (1643-1727) during their

attempts to describe the movement of projectiles and the solar system. The focus

shifted from searching for numbers that could be solutions of equations to

searching for equations of curves to represent paths of moving objects.

Justifications for descriptions of the paths of mobiles were difficult to

communicate and remained geometric in nature until the work of Fermat and

Descartes (1637) in analytical geometry facilitated representation of motion using

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

(1736-1813), Abel (1801-1829), and Galois (1811-1832) to solve polynomial

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

solution to a polynomial equation of the fifth degree. Galois invented group

theory to determine conditions needed to be fulfilled in order for a polynomial

equation to have a solution.

2.3 Algebra in the School Curriculum

To understand how school algebra, and in particular the Algebra I curriculum

was shaped in the United States, an historical perspective is presented first

(Section 2.3.1), followed by a curriculum perspective of school algebra (Section

2.3.2).

2.3.1 Algebra in the School Curriculum: Historical Perspective. The rise of

algebra during the European Renaissance marked the beginning of algebra as a

course of study at universities, and the creation of algebra textbooks. The first

algebra textbook written in vernacular, Luca Pacioli’s Summa de Arithmetica,

Geometria, Proportioni et Proportionalità, was published in Venice in 1494

(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

equations with numerical coefficients, and how to simplify simple algebraic

expressions (Kilpatrick & Izsák, 2008). The Benjamin Franklin Academy (later

the University of Pennsylvania) had an algebra course as part of its curriculum in

1751. By 1820, Harvard implemented an algebra entrance examination, and in

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 the secondary curriculum (Kilpatrick & Izsák, 2008).

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

initially shaped by practical, vocational purposes. Slowly, the college algebra

course content became a secondary school curriculum. The practical (vocational,

technical) approach to algebra disappeared by the end of the nineteenth century,

being replaced by the academic approach – to meet college entrance requirements.

The content of the high school algebra course was dominated by exercises in

simplifying algebraic expressions and solving equations of increased complexity

(Kilpatrick & Izsák, 2008).

During the first half of the twentieth century, the utility of algebra for the

average student was seriously questioned. Consequently, the algebra course

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

attention was given to skills like solving quadratic equations or symbolically

manipulating complex algebraic expressions.

For a short period during the New Math era, from the mid 1950s to the mid

1970s, the algebra curriculum was radically changed. As a consequence of the

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

technology. The emphasis on structural properties of the fields of real numbers

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

the nature and role of deductive reasoning in mathematics.

Of interest is the influence of Dieudonné and Piaget during the New Math era.

Dieudonné viewed mathematics as a unified whole, a “totality” kept together by

Bourbaki’s “mother-structures”– topological structure, order structure, and

algebraic structure (Corry, 2004). Dieudonné advocated for a unified structural

perspective in the mathematics curriculum and the elimination of the Euclidean

geometry, as something that escaped this unification. Piaget’s (1971) “natural

order” of intellectual development, topological to Euclidian intuitions, sustained

this point of view.

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

Back-to-Basic movement in the 1970s returned the algebra curriculum to a focus

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,

to make the introductory course, Algebra I, more accessible to students

(Kilpatrick & Izsák, 2008).

The Algebra for All movement (Pennsylvania Council of Teachers of

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

1989, the National Council of Teachers of Mathematics (NCTM) published

Curriculum and Evaluation Standards for School Mathematics. NCTM revised

the earlier document and produced the Principles and Standards for School

Mathematics in 2000. In both Standards documents, algebra was established as a

major strand of the elementary through high school curriculum. Studies of student

achievement documented benefits of students’ universal and early access to

algebra for their future academic achievement (Smith, 1996; Gamoran &

Hannigan, 2000; Speilhagen, 2006). In the 1990s, Algebra I became the

mandatory “gateway” course to the study of mathematics and higher level

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

(Kilpatrick, Mesa, & Sloane, 2007).

The shift from the static equation solving to the dynamic functional approach

was the most notable conceptual change in the algebra secondary curriculum

brought on by the standards-based era. This new algebra curriculum is

characterized by an emphasis on the functional approach to algebra, and increased

attention to integration and connectedness across mathematical disciplines, ideas,

concepts, and representations. Solving equations using a functional perspective

coincided with students’ broad access to graphing technology (e.g., graphing

calculators, computers with graphing software). The algebra curriculum began to

include functions and their properties for purposes other than solving equations.

Functions started to be used to model change, although on a small scale

27
(Kilpatrick et al., 2007). The standards-based era culminated with the publication

of the Common Core State Standards for Mathematics (CCSSO & NGA, 2010)

highlighting modeling as a fundamental high school mathematics objective.

2.3.2 School Algebra from a Curriculum Perspective. School algebra

begins when students use letters to represent numbers. As noted by Usiskin

(1988), school algebra may be regarded as the study of generalized arithmetic,

algorithms, functions, and structures. Each of these is described below, and is

accompanied with MDT problems that are illustrative of each of these four topics.

1. Generalized arithmetic, where variables are used as pattern generalizers.

For example the pattern 12, 18, 24, 30 can be generalized to 6(n  1) or

6n  6 (see Appendix A, MDT problem 5).

2. The study of procedures or algorithms for solving certain kinds of

problems in which values of variables are often unknowns to be found.

For example:

a) To compute the slope of a line graphed in a Cartesian system of

coordinates, two points on the line ( x1 , y1 ) and ( x2 , y 2 ) must be

y 2  y1
identified, and the ratio m must be calculated (see
x 2  x1

Appendix A, MDT problem 4).

b) To complete a table of values for a given function (e.g., y mx  b ),

x-values from the domain of the function must be identified, and

28
 corresponding values for y must be calculated (see Appendix A, MDT

problem 2).

c) To check if a point ( , ) is situated on the line y mx  b , the

coordinates of the point must be substituted into the equation of the

line, and a determination must be made whether the statement

y0 mx0  b is true (see Appendix A, MDT problem 1).

3. The study of relationships among quantities, where variables are

arguments or parameters of functions. For example, to recognize the

f ( x2 )  f ( x1 )
relationship between the rate of change m of the linear
x2  x1

function f : ƒ o ƒ, f ( x) mx  b, m z 0 (or y mx  b) and the slope

y 2  y1
m of the line with the equation y mx  b , and make the
x 2  x1

connection between the symbolic representation and the geometric

representation (see Appendix A, MDT problem 7).

4. The study of structures, where variables are arbitrary marks on the page or

objects that should be manipulated according to the rules of the algebraic

symbolic game. For example, apply the rules for simplifying algebraic

expressions and obtain equivalent equations in order to rewrite the linear

equation from standard form ax  by c in slope-intercept form

y mx  b (see Appendix A, MDT problem 6).

29
2.4 Research on Student Understanding of Linearity and Linear Functions

Various research studies have focused on student understanding of aspects of

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

representations of linear functions, particularly the Cartesian connection. Studies

conducted with students in Grades 8-10 and other age groups are presented

(Section 2.4.1), followed by teaching experiments (Section 2.4.2), and studies in

other content areas (e.g., Physics) (Section 2.4.3).

2.4.1 Studies on Student Understanding of Linearity and Linear

Functions. Greenes, Chang, and Ben-Chaim (2007) conducted a study of 1732

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

students, 43.5% of Korean students, and 58.6% of Israeli students successfully

solved that problem. MDT problems 7b and 7c require calculation of the slope of

a line segment graphed in a non-homogenous system of coordinates (independent

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

segment representing the movement of a car. Percentages of correct answers on

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%

of U.S. students provided an explanation. MDT problem 6 requires the

identification of the y-intercept from five choices (0, 2, 4, 6, and 12), given the

equation of the line 2 x  3 y 12 . About one third (35.2%) of U.S. students

selected the correct answer for the y-intercept.

Interviews of those students revealed difficulties connecting between

representations, particularly with the Cartesian connection. In MDT problem 1,

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

Moschkovich et al. (1993), Smith et al.(1989), and Schoenfeld et al. (1993). In

MDT problem 4, students did not connect x2  x1 (change in x or “run”) with its

graphic representation in a Cartesian systems of coordinates, an oriented segment

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,

in MDT problem 4, students calculated a negative slope, although the linear

function is increasing. This finding is in agreement with research by Moschkovich

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

point (3, 120) in their calculation of slope.

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

inter-rater agreement were not addressed. A new direction of research was

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.

Knuth (2000) studied high school students’ understanding of the Cartesian

connection, that is, the relationship between symbolic and graphic representations

of linear functions. Participants were 178 students from the same high school,

with a traditional curriculum that focused on algebraic/symbolic representations

and their manipulations. Students were enrolled in Algebra I (N=44), Geometry

(N=19), Algebra II (N=65), Precalculus (N=32), and Calculus (N=18). Six

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

the incomplete algebraic representation ? x  3 y 6 where the coefficient of x is

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

required fewer solution steps. In their algebraic solutions, students manipulated

the equation ax  by c , and re-wrote it in slope-intercept form, y mx  b .

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

points ( x0 , y 0 ) as solutions to the linear equation ax  by c . Fewer than one

third of students chose a graphic approach, either as their primary or secondary

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

that students’ reliance on symbolic representations is due to the curriculum and

instructional approach that emphasizes symbolic manipulations, but also to the

nature of the problems themselves, that required solutions to equations. As Arcavi

(2003) points out, the graphic approach may have been rejected by students (and

their teachers) because graphic representations are not precise, and could

introduce estimates and inaccuracies.

In Knuth’s (2000) study, student difficulties with the Cartesian connection

were inferred by the researcher from student answers, without any interview input

from students or their teachers. Also, all students experienced the same

curriculum, raising the question of whether students experiencing a different

curriculum might encounter the same difficulties. In the present study, students

and teachers were asked to comment on difficulties. Furthermore, students’

instructional programs varied by district.

Cheng (2010) studied the relationship between students’ understanding of

steepness and their abilities to reason proportionally. Participants were 413

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

(primarily Grade 6), or Pre-Algebra, Algebra I, or Geometry, and in one of three

34
tracks, regular, honors, or accelerated. Two tests were administered to all

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

homogenous systems of coordinates. Interviews were conducted with 16 of the

middle school students who were enrolled in two private schools.

Cheng (2010) found a positive correlation between students’ abilities to reason

proportionally and their abilities to solve steepness problems. About a quarter of

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

performed better in mathematical situations when asked to compare the slopes of

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 &

Maxwell-West, 1994; Lamon, 1993; Lobato & Siebert, 2002).

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

in Cheng’s (2010) study. The relationship between students’ reasoning abilities

about steepness, and their reasoning abilities about ratio and proportionality was

determined using a two-factor ANOVA and linear regression. However, the

assumption of independence necessary in traditional statistical analyses 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

groups were not independent. The violation of the assumption of independence

casts doubts upon any statistical significance reported in the study (Raudenbush &

Bryk, 2002). Also, the researcher did not take into consideration natural clusters,

like clustering by track or performance level (regular, accelerated or honors) or by

current mathematics course (e.g., Regular Mathematics for 6th, 7th, or 8th grade,

Pre-Algebra, Algebra I, Geometry). In short, the relationship between steepness

and proportionality may be different for various groups of students (e.g., low-

achieving vs. high-achieving, Pre-Algebra students vs. Geometry students). In the

present study, analyses of student performance on the MDT are made by grade,

current mathematics course, and performance/track level.

36
 Stump (2001a) studied student understanding of slope or rate of change.

Twenty-two high school students enrolled in Precalculus used bicycles to

investigate the linear relationship between the number of pedal revolutions

(independent variable) and the number of wheel revolutions (dependent variable)

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

task, the graph of a linear function in a non-homogeneous system of coordinates

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.,

finding the slope of a wheelchair ramp).

The small number of participants makes it difficult to generalize the findings

of Stump’s (2001a) study to other populations of students.

Teuscher and Reys (2010) studied student difficulties with identifying the rate

of change from a graph. Participants were 191 high school students enrolled in

Advanced Placement (AP) Calculus. The graph of a piece-wise function was

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

change, positive (  2 ) or negative (  2 ), and took into consideration the absolute

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

problem or students’ lack of familiarity with the situation (a graphic

representation of a piece-wise function), that could have played a role in students’

failure to correctly solve the tasks. Student difficulties were inferred by

researchers from student solutions, with no input from students or their teachers in

the form of interviews.

Orton (1983) studied student understanding of various concepts of Calculus,

including rate of change. Reported here is only the part of the study focusing on

constant rate of change. Orton conducted interviews with 60 students in high

school, ages 16-18 from four high schools, and 50 students, ages 18-22, from two

colleges. The college students were preparing to become mathematics teachers.

One of the tasks concerned rate of change of a linear function in a real-life

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

values 0, 1, 2, 3, 4, 5 for the independent variable x, and the values 0, 2, 4, 6, 8,

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

symbolic representation of the linear function y 3x  1 were presented, and

1
participants had to figure out the rate of change at x 2 and at x T . Both
2

graphs used a non-homogeneous system of coordinates. Both tasks were scored as

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

or table of values, 26% of students attempted or indentified the rate of change at

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

were due to incorrect understanding of rate of change, and to weak understanding

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

the line would result.

40
 Orton’s (1983) study demonstrated that high school and future mathematics

teachers alike do not have a comprehensive understanding of the concept of rate

of change. Difficulties with rate of change occurred across both groups, regardless

of age and mathematical experience.

Zaslavsky et al. (2002) investigated 124 participants’ understanding of slope in

non-homogenous systems of coordinates, that is, whether slope is an analytical

attribute of the linear function represented by a line or a geometric attribute of the

line. Participants included 28 high school students in Grade 11 enrolled in

Calculus, 28 pre-service secondary mathematics teachers, 50 experienced in-

service mathematics teachers, 14 mathematics educators involved in research, and

four research mathematicians. Two tasks were administered to all participants,

and interviews were conducted with 12 of the participants. In Task 1, the function

f ( x) x was graphically represented in a homogenous system of coordinates,

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

function f ( x) x was graphically represented in a non-homogenous system of

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-

axis. Participants’ answers revealed three perspectives on slope: analytical,

geometric, and mixed. Those perspectives were common across all groups, high

school students, pre-service teachers, experienced teachers, mathematics

educators, and mathematicians. Twenty-four participants held an analytical

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

of the function f ( x) x is m 1 , but 51 participants could not find a value for

tan E or dismissed the question, while 40 participants calculated tan E z 1

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

participants’ (three mathematicians and two doctoral students in mathematics

education) responses were reported in detail. Participants holding either an

analytical or a geometric perspective of slope did not encounter cognitive

conflicts, whereas participants with a mixed perspective had difficulties

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-

homogeneous system of coordinates was used.

Although the researchers reported in detail on the slope perspectives held by

only the Ph.D. candidates in mathematics education and holders of Ph.D. degrees

in mathematics, those perspectives were common across all groups of

participants.

2.4.2 Teaching Experiments on Student Learning of Linearity and Linear

Functions. In teaching experiments, the focus is primarily on student learning

trajectories. The review of teaching experiments on student learning of linearity

42
and linear functions emphasizes student difficulties and researchers’

recommendations for curriculum and instruction.

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

with a constant ratio.

In another situation in the Lobato and Siebert’s (2002) study, involving a

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

matched that of 10 cm in 4 seconds. In summary, the student who successfully

created a ratio as measure for the steepness of a wheelchair ramp, also

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

to student difficulty transferring between different situations or contexts.

The complexity of the situations should be considered in the interpretations of

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

involves proportionality resulting from similarities of triangles. By contrast, the

linear movement context involves direct variation of distance and time.

Hines (2002) conducted a teaching experiment with one Grade 8 student to

gain insight into student’s understanding of linear functions and connections

among representations of linear functions. Hines used a dynamic physical model

as a concrete way of representing functions as covariational processes involving

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

variable varying simultaneously. Moreover, the student was able to make

connections among various representations of linear functions (tables, graphs,

equations). Since Hines created special learning experiences for the student,

generalizing the findings to other populations of students is difficult.

A teaching experiment was performed by Ellis (2009) with two groups of

students. Group 1, consisting of seven Grade 8 Algebra I students studied linearity

and linear functions using a number-pattern approach over a period of 12 lessons,

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

linearity and linear functions with a focus on quantitative relationships, over a

period of 15 one-hour lessons. During instruction, students were presented with

two-real world situations: 1) rotations of physical gears, and 2) linear movement

with constant speed simulated by computer software and involving two

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

student understanding of slope or rate of change, in agreement with Lobato and

Siebert (2002).

Davis (2007) studied student understanding of the y-intercept of linear

functions in real-world contexts and across multiple-representations, and the

terminology students used to describe relationships. Participants were eight high

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

pre-test was given to assess students’ abilities to make connections between

different representations. During instruction, students were allowed to invent their

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

“starting point.” Interviews were conducted with all participants. Analyses of

observations and interviews revealed that students’ successful use of the “starting

point,” suggested by the real-world contexts, did not translate into successful use

of the y-intercept when symbolic, tabular, and graphical representations were

involved. Moreover, the use of the informal terminology seemed to have a

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

instructional materials upon his students’ meanings as well as students’

difficulties to translate their knowledge of the “starting point” between contexts

and representations.

Although conducted with a small number of participants, the observations

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

students have difficulty transferring knowledge of the y-intercept between

situations (Schoenfeld et al., 1993; Smith et al., 1989).

The Functions Group at Berkeley conducted a teaching experiment of seven

hours duration over a period of seven weeks, with a 16-year-old student to explore

the student’s change in knowledge of linear functions, their representations, and

connections between representations, particularly the Cartesian connection

(Schoenfeld et al., 1993). The student was concurrently enrolled in a Calculus

summer course at The University of California-Berkeley. In Grade 8, the student

completed Algebra I on her own, with little help from a teacher.

Schoenfeld et al. (1993) conducted the teaching experiment using the software

environment Grapher, designed by the Functions Group at Berkeley, and

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

of m and b as parameters. The student assigned various values to m in y mx  b

and the software accordingly rotated the graph of the function through the fixed y-

intercept. The student also assigned various values to b in y mx  b and the

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

graphed in a Cartesian system of coordinates to its symbolic representation. A

researcher observed the student interacting with the software and asked questions

about her understanding. Analysis of student answers showed that the student

changed her conceptual understanding about the necessity of three descriptors to

graph the line y mx  b (slope m, y-intercept, and x-intercept) to two descriptors

(slope m, and y-intercept). However, there was limited and unstable change in the

student’s knowledge of the connection between the symbolic representation of the

function y mx  b , its coefficients m and b, and their graphic representation.

At the beginning of the teaching experiment conducted by Schoenfeld et al.

(1993), the student’s knowledge of slope was limited to the algorithm of

y 2  y1
calculating the slope using the formula m , with no connection to the
x 2  x1

graphic representation. She did not know that y 2  y1 and x2  x1 have

magnitude and sense (positive or negative) specified by mathematical

48
conventions, and could not coordinate the negative sign of y 2  y1 with a

decrease in the y-value, or the positive sign of x2  x1 with an increase in the x-

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

the point (0, b) had to be situated on the line.

Although only one student’s difficulties were reported, Schoenfeld et al.

(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

change in y and the change in x represented as line segments with oriented

magnitudes is seen by Schoenfeld et al. as an example of missing the Cartesian

connection.

In another teaching experiment by the Functions Group at Berkeley, with two

students in Grade 8, and with the same software, Grapher, Moschovich et al.

(1993) focused on student flexibility to move between the process-view of

functions and the object view of linear functions. During the teaching experiment

students had to complete various tasks. For example:

1) Given a table of values, students had to write the symbolic representation

of the linear function (process-view).

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).

Other more complex tasks required both a process-view and an object-view of

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

through a given point.

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

students’ difficulties with connecting between various function representations by

the fact that the conceptual development of functions as processes-objects might

not be a linear one - from process to object with the object view reserved for the

mathematicians - but rather a simultaneous one. Both perspectives, and the

flexibility of moving between perspectives, are necessary to solve complex

problems.

2.4.3 Studies on Student Understanding of Slope or Rate of Change in

Other Content Areas. Research on student understanding of slope or rate of

change in other content areas focuses on students’ abilities to connect various

representations of real-world phenomena, and their flexibility to transfer

knowledge between mathematical and real-life situations.

50
 The Physics Education Group at the University of Washington investigated

student difficulties connecting graphs with speed in real-world situations

(McDermott et al., 1987). This multi-year study was based on observations of

several hundred participants, including college students enrolled in a preparatory

course for Physics, pre-service and in-service teachers enrolled in a university

Physics course, college students enrolled in Honors Calculus-based Physics at the

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

a non-homogeneous system of coordinates. Participants were asked to compare

the speeds of the two objects at the moment t 2 . Many of the participants who

incorrectly compared the two speeds focused on the change in position at t 2,

and decided that the higher vertical-coordinate (greater height of position)

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.

Newburgh (2001) conducted a teaching experiment with twenty-six Grade 12

high school students enrolled in Physics to investigate their understanding of

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

(independent variable - height, dependent variable - area), and constructed scatter

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

rectangle. The majority of students omitted the units of measurement when

referring to slope. The same task was repeated in different situations (e.g., the

movement of a cart with constant acceleration in which students measured

velocity and time). The students who participated in this teaching experiment

were more likely than contrast-group students to transfer their knowledge of slope

to other situations. Newburgh’s findings, i.e., the flexibility of transfer between

situations, are not in concert with the findings from other studies, as for example

Lobato and Siebert (2002).

2.5 Research on Teacher Knowledge of Student Understanding of Linearity

and Linear Functions

The studies conducted by Orton (1983), McDermott et al. (1987), and

Zaslavsky et al. (2002) cited previously, showed that pre-service and in-service

teachers have difficulty understanding linearity and linear functions. Some

teaching experiments (Davis, 2007; Thompson & Thompson, 1994) revealed that

teachers are not always successful in attending to their students’ thinking. For

example, in Davis’s teaching experiment on student understanding of the y-

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

students could move flexibly between various representations, contexts and

situations in which y-intercept occurred.

Several additional studies on teacher mathematical and pedagogical knowledge

of slope or rate of change, and connections between representations are described

below. The pedagogical knowledge is focused on understanding of student

difficulties.

In a study conducted with in-service (N=21) and pre-service (N=18)

mathematics teachers, Stump (1997) investigated their knowledge of various

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

mathematics or mathematics education, and eight of the 15 teachers with Master

degrees had a Master in mathematics or mathematics education. All pre-service

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

problems and questions about teachers’ mathematical and pedagogical knowledge

of slope.

Analyses of teachers’ surveys and interviews showed that, on average, teachers

mentioned two representations of slope. More than 90% of the pre-service

teachers and all of the in-service teachers mentioned slope in geometric-analytical

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,

for example, a measure of steepness of hills, ski slopes or wheelchair ramps.

Slope as the parameter m in y mx  b was mentioned by about one third of the

pre-service teachers, and by only one in-service teacher. None of the pre-service

teachers, and only one of the in-service teachers mentioned slope in a

trigonometric context, as the tangent of the angle of inclination. Slope as a ratio

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

situations requiring connections between graphic representations and linear

movement with constant speed. For example, in one of the problems taken from

McDermott et al. (1987), a graph (position versus time) representing linear

motion with constant speed of two objects ( m1 z m2 ) is presented in a non-

54
homogeneous system of coordinates. Teachers were asked to compare the speeds

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

vertical-coordinate (greater height of position) indicated the object with the

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

McDermott et al. (1997).

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

changes of temperature is given, l  l0 al0 (t  t 0 ) , where l is the length of the

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

a linear function of temperature, l (t ) mt  b , and to identify the slope and the y-

intercept. About the same percentage of pre-service teachers (83%) as in-service

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

did not respond.

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

in-service and pre-service teachers had difficulty recognizing slope as a

parameter, as constant rate of change of two covarying quantities, and interpreting

graphs that required connecting various representations of slope. The in-service

teachers identified more connections for slope representations than did the pre-

service teachers. For example, the in-service teachers could connect between the

geometric-analytical context of slope and the trigonometric context of slope.

Teachers’ weak content knowledge of slope was mirrored by their weak

pedagogical knowledge. When asked about their mathematical knowledge, the

most often mentioned context for slope was the geometric-analytical one.

However, analyses of teachers’ pedagogical knowledge revealed that they most

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

well as connections between various contexts/representations of slope. Stump

(1997) pointed out that teachers did not question their students’ abilities to make

connections between various representations or transfer among them. When asked

to identify student difficulties with slope, the in-service teachers mentioned

difficulties with graphing lines and computing slope, while the pre-service

teachers mentioned difficulties with understanding what slope represents. Stump

recommended more research to study the impact of the mathematics curriculum

on teachers’ weak understanding of slope, and the effect of teachers’ weak

mathematical and pedagogical knowledge of slope on their students’

understanding of slope.

Stump’s (1997) study of teachers’ mathematical knowledge of slope was

comprehensive, and the categories used for slope contexts/representations were

used for the categories of conceptual knowledge of slope in the present study (see

Chapter 3).

Stump (2001b) conducted another study on the development of future

mathematics teachers’ knowledge of representations for teaching slope, and

knowledge of students’ difficulties with slope. Participants were six future

secondary mathematics teachers enrolled in a one-semester methods course. Prior

to the start of the methods course, a survey was administered to gauge future

mathematics teachers’ knowledge of slope and knowledge of student difficulties

with slope. The methods course included activities and class discussions on

57
various contexts and representations for slope (e.g., geometric, analytical), and

emphasized slope as rate of change in functional situations. Assignments included

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

future teachers’ knowledge of representations for teaching slope and knowledge

of students’ difficulties with slope, future teachers’ prior knowledge was

compared with their knowledge developed during the methods course, and with

the knowledge demonstrated during the actual teaching.

At the beginning of Stump’s (2001b) study, the future mathematics teachers

referred to slope as a measure of steepness in graphic representations (e.g., the

slope of the line y x ) and physical situations (e.g., inclined plane, roof). At the

end of the study, their representations of slope expanded to functional situations

(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

remained focused on physical situations (slope as a measure of steepness), another

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

situations might be due to the influence of the traditional curriculum of the

algebra textbook used in the basic Algebra I course.

With respect to understanding of student difficulties with the concept of slope,

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

students, he discovered that students had conceptual difficulties. Consequently, he

developed lessons to emphasize the conceptual understanding of slope. Another

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

slope as a measure of steepness. Interviews with students confirmed the third

teacher’s initial hypothesis that students have a weak understanding of slope as a

measure of steepness. Accordingly, she developed lessons focused on slope as a

measure of steepness (“slant”). In conclusion, the future teachers’ initial

assumptions of student difficulties with slope were changed or confirmed by

interviews, and then used in the development of their lessons.

Stump’s (2001b) study demonstrates the importance of teachers’ content and

pedagogical knowledge of various representations of slope and of student

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

measure of steepness. Two of the future teachers’ assumptions about students’

understanding of slope were contradicted and had to be changed.

Even and Tirosh (1995) studied 162 prospective secondary mathematics

teachers’ subject-matter knowledge and knowledge about students’ understanding

of functions and undefined mathematical operations. All participants were in the

final stage of their pre-service preparation. An open-ended questionnaire was

completed by 152 of the prospective teachers, and interviews were conducted

with the other 10 prospective teachers. The questionnaires and the interviews

included problems and student solutions to problems about functions and

undefined mathematical operations. Reported here are only the aspects of

teachers’ pedagogical knowledge referring to student understanding of linear

functions. Teachers were asked to respond to a student error about linear

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

sense of the X-axis.

In one problem, a point A is shown in a rectangular system of coordinates, but

neither units of measurements are given, nor are coordinates marked. Should the

units of measurement be marked (assuming a homogeneous system of

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

coordinates. A student’s answer accompanied the problem:

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

knowledge and knowledge about student understanding is a complex one, and

both types of knowledge are necessary for teaching. The researchers found that

only half of the prospective secondary mathematics teachers could attend to

student reasoning about linear functions in the problem described above, and
61
some of those teachers could not solve the problem themselves. The researchers’

findings confirm Stump’s (1997, 2001b) findings about teachers’ weak

mathematical knowledge of slope, and weak knowledge of student understanding

and difficulties with the concept of slope.

Cunningham (2005) administered a survey to 28 algebra teachers to determine

how often problems requiring connections between function representations are

used during instruction and in teacher-made assessments. Since the teachers

taught students in Grades 8-10, the survey contained only problems focusing on

linear functions, appropriate for Algebra I curriculum. The 12 survey problems

required making connections between algebraic/symbolic (A), numerical/tabular

(N), and graphic (G) representations. Teachers were asked to identify the number

of class periods and teacher-made assignments that included problems similar to

those in the survey. For the problem requiring connecting graphic and numerical

representation, G to N, a linear function graphed in a homogeneous system of

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

algebraic/symbolic and numerical/tabular representations, A to N, a linear

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

problem requiring connecting between the graphic and algebraic/symbolic

representations of a linear function, G to A, a linear function was graphed in a

homogeneous system of coordinates and the task was to find its symbolic

representation y mx  b . Teachers mentioned that they spent 2-3 class periods

teaching the connection G to A (mean 2.21) and used 2-3 problems requiring the

connection G to A in their assessments (mean 2.67).

Cunningham (2005) noted that teachers gave so little attention to those

connections because they assumed that their students would not encounter

difficulties making the connections. However, studies cited previously identify

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

connecting between various representations of functions, especially from graphic

to numerical representations. As a result, they allocate the least amount of

instructional time to connections among representations, and rarely assess student

understanding of these connections.

You (2006) studied teachers’ mathematical and pedagogical knowledge of

linear functions with a focus on various representations of those functions.

Reported here is only the part of the study focusing on teachers’ mathematical

knowledge of the Cartesian connection and the relationship to their abilities to

recognize student difficulties with connections among representations.

Participants were 104 elementary and middle school pre-service teachers in the

63
last stage of their preparation and enrolled in methods of teaching

elementary/middle mathematics courses. One of the problems presented a table of

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)

and (2, 4) , and an incomplete algebraic representation, ? x  3 y 6 , where the

coefficient of x is missing. You’s findings about teachers’ knowledge of the

Cartesian connection confirmed Knuth’s (2000) findings about student difficulties

with the Cartesian connections, and their overreliance on algebraic solutions.

Moreover, You (2006) found that the teachers’ knowledge of representations

of linear functions is related to their pedagogical knowledge, particularly their

ability to attend to their students’ understanding of representations of linear

functions. The pre-service teachers did not successfully recognize student errors,

misconceptions, or difficulties with linear functions and their representations.

You’s (2006) study results also confirmed Stump’s (1997, 2001b), Even and

Tirosh’s (1995), and Cunningham’s (2005) findings. Teachers’ weak

mathematical knowledge is positively associated with weak pedagogical

64
knowledge, particularly teachers’ abilities to attend to their students’

understanding of linearity and linear functions.

2.6 Student and Teacher Identification of Problem Difficulty

Recognizing and identifying the relative degrees of problem difficulty with

accuracy is critical to the construction of assessment instruments and to the design

of instructional approaches that capitalize on what students know. Identifying

problem difficulty is equally important for students, enabling them to identify

their strengths and weaknesses and subsequently take greater responsibility for

their own learning. The following research studies refer to measures of problem

difficulty. In the discussion that follows, studies of student assessment of problem

difficulty are presented first followed by studies comparing student and teacher

assessment of problem difficulty, and then studies on teacher assessment of

problem difficulty.

Lee and Heyworth (1996) conducted a study to develop a measure of problem

difficulty. Participants were Grade 9 students (N=125) from two secondary

schools in Hong Kong, and mathematics teachers (N=28) enrolled in mathematics

education courses at university in Hong Kong. Two mathematics tests, with a total

of 32 problems requiring the simplification of algebraic expressions with

logarithms and the solving of logarithmic equations, were administered to all

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

of problem difficulty. Student perceptions of problem difficulty correlated highly

with their performance on the two tests. Teacher assessments of problem

difficulty for students correlated moderately low with student performance.

Using a regression framework, Lee and Heyworth identified the factors predicting

the three measures of problem difficulty given by: 1) student perceived difficulty,

2) student performance measured as a ratio between the number of correct

answers and the total number of answers, and 3) teacher assessment of problem

difficulty. Student perceived level of problem difficulty could be predicted by the

following four factors: 1) perceived number of difficult steps required to solve the

problem, 2) number of algorithmic steps necessary to solve the problem, 3)

number of operations necessary to solve the problem, and 4) student familiarity

with the type of problem (e.g., logarithmic equations that can be brought in the

general form log a f ( x) log a g ( x) ). Results showed that student performance

could be predicted only by the number of algorithmic steps necessary to solve a

problem. Teacher assessment of problem difficulty could be predicted by student

familiarity with the problem, the number of “log” occurrences in the problem, and

the numerical complexity of the problem. The difference between the factors

predicting student perception of problem difficulty and those factors predicting

teacher assessment of problem difficulty might be explained by the fact that

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

problem (e.g., number of “log” occurrences in the problem).

66
 The researchers neither described the student sample nor accounted for the fact

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

measures of problem difficulty are supported by previous research literature

(Jerman, 1973; Silver & Thompson, 1984; Lane, 1991; Mason, Zollman,

Bramble, & O'Brien, 1992), more research is needed to develop a measure for

problem difficulty that accounts for student perceived difficulty. Nevertheless,

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.

A study by Lingard, Minasian-Batmanian, Vella, Cathers, and Gonzalez

(2009) was conducted to compare student and faculty perception of problem

difficulty in first-year Biochemistry and Physics courses. Participants were

Australian undergraduate students in their first year of college and enrolled in

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

with 30 multiple-choice items. Faculty assigned a level of difficulty to each of the

items (“Pass” - for simple, one-step problems, requiring recall or manipulation of

information; “Credit” - for items requiring what a “Pass” question requires as well

as the application of concepts to practical situations; “Distinction/High

Distinction” - for questions requiring what a “Credit” question requires, as well as

a higher level of analysis and critical thinking). Students answered the questions,

67
and assigned each question a level of difficulty (Pass, Credit, and

Distinction/High Distinction). Results showed disagreement between student

assessment of problem difficulty and faculty judgment of problem difficulty, as

well as between student performance and student assessment of problem

difficulty. Overall, the more difficult questions were associated with lower student

and faculty agreement about difficulty levels, and longer response. For

Biochemistry items, student and faculty agreement on problem difficulty was

about the same for all questions. For Physics students, student and faculty

agreement on difficulty was lower as student performance decreased (i.e., the

agreement was 53% for “Pass” questions that were correctly answered by 71% of

students, and the agreement decreased to 33% for “Distinction/High Distinction”

questions that were correctly answered by only 47% of students). The association

between student performance and student perception of problem difficulty was

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

difficulty assigned by faculty, students who gave an incorrect answer to a

question considered it to be more difficult, while students who correctly answered

a question considered it to be less difficult. Furthermore, students who correctly

identified a “Pass“ question were more likely to answer it correctly. Students who

correctly identified a “Distinction/High Distinction” question were more likely to

answer it incorrectly.

68
 Lingard et al.’s (2009) study is important because it draws attention to the low

agreement on problem difficulty between students and their instructors.

Furthermore, it highlights the complex relationships among a problem’s specific

domain of knowledge (e.g., Physics, Biochemistry), students’ prior knowledge,

student performance, student and faculty assessment of problem difficulty, and

student and faculty agreement on problem difficulty. In other words, assessing

the difficulty of a hard Physics problem by students without prior knowledge of

Physics, or by their instructors, is a different task than assessing the difficulty of

any Biochemistry problem by students or their instructors.

Gire and Rebello (2010) conducted two studies to investigate how factors,

such as problem complexity, and familiarity with the type of problem, are

correlated with student and teacher perceived problem difficulty. Participants in

both studies were college students enrolled in introductory Physics courses and

their instructors (professors, teaching assistants) at a Midwestern state university.

Fifteen students, enrolled in a Calculus-based Physics course, and 14 instructors

participated in the first study. Participants were asked to rate, on a scale from 1 to

10 (1-easiest, 10-most difficult), the level of difficulty of 16 problems on the topic

of Work-Energy. Instructors rated the difficulty level of problems from the

perspective of their students.

In the second study by Gire and Rebello (2010), students (N=21) were enrolled

in an algebra-based Physics course. Students were asked to solve 10 Kinematics

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

level of difficulty for each problem on a scale from 1 to 5 (1-easiest, 5-most

difficult). Instructors (N=16) of the students were asked to rate the level of

problem difficulty, as perceived by their students. Student solutions to the

Kinematics test were scored.

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’

estimation of difficulty. Also, in the second study, student performance as well as

student ratings of familiarity with problems correlated highly with student and

instructor ratings of problem difficulty. High performance scores were associated

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

first study, problem complexity correlated highly with instructor rating of

problem difficulty, but not with student rating of problem difficulty. In the

second study, problem complexity correlated highly with both student and

instructor ratings of problem difficulty. Results concerning problem complexity

suggest the necessity of reconsidering problem complexity scoring criteria, and

that the problem domain (Work-Energy and Kinematics) and the place where the

student is positioned in the mathematical curriculum course sequence (e.g.,

70
Calculus-based Physics, algebra-based Physics) may play important roles in

student perception of problem difficulty.

Zhou (2009) used teacher assessment of relative difficulty of problems and the

factors associated with those assessments, and compared them to the relative

problem difficulty determined by student performance. The relative difficulty of a

problem is given by its rank order in a set of problems. Participants were 10

teachers recruited from professional development workshops held at a university

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

taught all elementary-school subjects, typical of elementary-certified teachers.

Teachers were presented with 11 problems involving linear measurement (e.g.,

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

middle school students (N=360) in several school districts by teachers

participating in other professional development activities. Teacher ranking of

problem difficulty and problem difficulty based on student performance were

used to calculate the Spearman rank-correlations for each teacher. Only three

teachers’ rankings were significantly correlated with student performance. Two of

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

correlations between teacher ranking and student performance. One of the

teachers with a high Spearman rank-correlation incorrectly solved one the

problems. One of the two teachers with mathematics degrees overestimated easier

problems, and the other underestimated the difficulty levels of problems. Neither

the number of years of teaching experience nor the amount of professional

development training was associated with higher correlations between teacher

ranking and student performance.

The small number of teachers participating in this study (N=10) makes it

difficult to generalize findings to other populations. The empirical ranking of

problem difficulty was obtained using a sample of students from a population

different form the population of students taught by the teachers participating in

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

comparison intended by researchers. However, the study draws attention to the

fact that the relationship between teachers’ mathematical knowledge and

pedagogical knowledge is a complex one, and more research is needed to

investigate the factors associated with teacher understanding of their students’

difficulties.

Nathan and Koedinger conducted studies (Koedinger & Nathan, 2004; Nathan

& Koedinger, 2000) to compare teacher and researcher prediction of relative

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

schools enrolled in Algebra I (N=229) and Geometry (N=18), experienced

secondary mathematics teachers (N=67), and researchers in mathematics

education (N=35). Teachers were recruited from professional development

workshops held at a university in the southeastern part of the U.S. Researchers

were recruited from an Internet discussion group. Student performance was

assessed using an 8-item test of “early” algebra problems requiring quantitative

reasoning. The comparison between teacher predictions and student performance

focused on six types of problems determined by problem characteristics like the

position of the unknown quantity (“result-unknown” or “start–unknown”) and the

format (“symbolic equation,” “story problem,” or “word-equation”). For

example, one “start-unknown, word-equation” problem was:

“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).

An example of a “result-unknown, symbolic equation” problem was:

“Solve for x: (81.90 - 66) / 6 = x” (Nathan & Koedinger, 2000, p. 170).

A “start-unknown, story” problem was:

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 &

Koedinger, 2000, p. 170).

73
 Based on student performance, “start-unknown” problems were more difficult

than “result-unknown” problems, and “symbolic-equation” problems were more

difficult than “word-equation” and “story” problems. Teachers and researchers

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

predicted by teachers and researchers, “start-unknown” problems were more

difficult for students than “result-unknown” problems. Contrary to teachers' and

researchers’ expectations, who predicted that “story” problems and “word-

equation” problems would be more difficult than “symbol-equation” problems,

students found “symbolically-presented” problems most difficult. Nathan and

Koedinger (2000) suggest that teachers' and researchers’ expectations of problem

difficulty for their students reflect Algebra textbook instructional sequences in

which “story” problems and “word-equation” problems are presented after

“symbolic equation” problems.

Statistical significances reported by Koedinger & Nathan (2004) might not

reflect the true relationships between problem characteristics and problem

difficulty. The analysis of variance used by Koedinger & Nathan requires

independence of observations, but this assumption was violated since

participating students were grouped within the same classes and schools.

Moreover, students were not asked about their perceptions of problem difficulty,

and the comparison between student performance and teacher prediction of

problem difficulty was made by a group of teachers different from those teaching

the participating students.

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

problems. Fourteen experienced mathematics teachers (about 10 years of teaching

experience), pursuing advanced degrees in mathematics education participated in

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

each of the problems, given by the proportion of students in a specific population

that correctly solved the problem, was determined by administering the test

problems to samples of a minimum of 370 students in Grades 8 and 9. Teachers

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

there was a slight improvement in teacher prediction of percentage of students

with correct answers when information on student performance on 10 of the

questions was given, teachers underestimated the percentage of students with

correct answers. However, teachers were better at predicting the rank-order of

problem difficulty in the two tests, as compared to the rank-order by student

performance. The researchers recommended teacher access to information about

student performance on subsets of problems from tests, in order to improve their

ability to predict student scores with problems of those types.

75
2.7 Concluding Remarks

Solving equations was the purpose of algebra until the nineteenth century.

Justifications for finding solutions to equations were mainly geometric in nature.

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

static equation perspective began to be replaced by the dynamic functional

perspective, and connections between graphic and symbolic representations. The

function approach to modeling change emphasized the importance of the concept

of rate of change, and brought to attention representations of real-world

phenomena.

The review of research literature on student understanding linearity and linear

functions revealed that students had difficulty with slope or rate of change, and

making connections among representations, particularly the Cartesian connection.

The geometric perspective of slope, i.e., slope as measure of the steepness of a

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

have difficulty simultaneously attending to changes in y and x, and considering

the change in y and change in x as oriented segments in the Cartesian system of

coordinates. Students have difficulty connecting between various representations

of functions, most notably, the Cartesian connection, i.e., connecting between the

symbolic and the graphic representation of a linear function. Also, students have

difficulty transferring knowledge between contexts (e.g., mathematical and real-

life situations for linear functions, geometric and analytical/functional situations

76
for slope). Although with various degrees of intensity, research studies reveal that

the same difficulties are experienced across various populations of students,

regardless of age and previous mathematical experience. Moreover, researchers

point out that the same difficulties also are experienced by pre-service and in-

service mathematics teachers.

Research on teacher knowledge of student understanding of linearity and linear

functions is scarce. The review of research literature on teacher understanding of

student difficulties with linearity and linear functions shows that teachers’ weak

understanding of slope or rate of change and the Cartesian connection matches

their weak ability to understand their students’ thinking and attend to their

students’ difficulties with linearity and linear functions.

There is a paucity of research studies on student perception of problem

difficulty as compared to student performance. Although the research on teacher

assessment of problem difficulty as compared with student performance is

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

between student and teacher perception of problem difficulty, and student

performance varies with student and teacher cultural backgrounds, curricular

experiences, type of learners, subject matter of the problem, and content and type

of problem.

Chapter Summary. In this chapter, an historical perspective of algebra as a

body of knowledge and algebra in the curriculum were presented, followed by a

review of the research literature on student and teacher understanding of linearity

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

collected and methods of analyses are presented.

78
 Chapter 3

DATA COLLECTED AND METHODS OF ANALYSES

3. 1 Introduction

The purpose of this study was to identify student difficulties with aspects of

concepts of linearity and linear functions, and teachers’ understanding of the

nature of the difficulties students experience. To gain insight into students’ and

teachers’ knowledge, instruments were either adapted from existing tools or

developed particularly for this study, and then implemented to answer the

following questions:

1. Does student performance on the Mini-Diagnostic Test (MDT) vary by:

a. Grade level?

b. Current mathematics course?

c. Problem?

2. Does student ranking of problems by difficulty, after completing the

MDT, correspond to their performance on the MDT?

3. Does teacher ranking of problems by difficulty, after completing the

MDT, correspond to student performance on the MDT?

4. What is the nature of students’ difficulties?

5. Do teachers’ speculations about the nature of students’ difficulties

correspond to actual student difficulties as identified from student

performance on the MDT and interviews?

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

Categories are presented in Section 3.4. Characteristics of participants, both

students and their mathematics teachers, are described in Section 3.5. The

timeline for recruiting participants, administering tests and interviews, scoring

tests and coding interviews, and analyzing the data, is found in Section 3.6.

Methods of analyses used to answer the research questions are addressed in

Section 3.7.

3.2 Study Instruments

The following instruments were used to collect data for this study: the Mini-

Diagnostic Test on Linearity and Linear Functions (MDT), the Ranking

Questionnaire for Students (RQS) and the Ranking Questionnaire for Teachers

(RQT), the Information Questionnaire for Students (IQS), the Information

Questionnaire for Teachers (IQT), and the Interview questions for students and

teachers.

The primary instrument for measuring students’ understanding of linearity and

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

of the Massachusetts Comprehensive Assessment System (MCAS).

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

construction of a table of five pairs of values ( x0 , y 0 ) that satisfy a linear function

presented in the form y mx  b . MDT problem 3 provides a table of values for a

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

corresponding function. MDT problem 6 involves figuring out the y-intercept of a

line whose equation is given in the form ax  by c . In MDT problem 7, a graph

representing the movement of a car at three different speeds is presented along

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

problem. The MDT requires about 30 minutes to complete. The MDT is

reproduced in Appendix A.

After completing the MDT, all participants were asked to rank the MDT

problems from most to least difficult. The Ranking Questionnaires were

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

students, and comment on the nature of the difficulty, as perceived by their

students, for the two most difficult problems.

Ranking Questionnaires were designed to be administered along with the

MDT. Necessary time for completion is 5 to 10 minutes.

Information Questionnaires were developed to identify several characteristics

of the participants. The Information Questionnaire for Students (IQS) requests

information about gender, grade level in school, current mathematics course, and

current mathematics teacher. The Information Questionnaire for Teachers (IQT)

requests highest academic degree, college major, current and previously taught

mathematics courses, and number of years of teaching experience.

Information Questionnaires were administered along with the other assessment

instruments, but before administration of the Mini-Diagnostic Test and the

Ranking Questionnaires. The Information Questionnaires require 3–5 minutes for

82
completion. The total time required for completion of the Information

Questionnaire, Mini-Diagnostic Test, and Ranking Questionnaire is 38-45

minutes.

To gain greater insight into their conceptual knowledge and nature of

difficulties with MDT linearity concepts, interviews were developed for use with

both student and teacher participants. These semi-structured, think-aloud, task

based interviews required students to describe their solutions to all MDT

problems to the interviewer. During students’ descriptions of their solution

processes, only comments of a prompting nature were used. Prompting comments

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

more information or clarification about the students’ conceptual understanding of

their difficulties. Probing questions included:

1. What is your meaning of slope?

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?

4. What was difficult about this problem? Why?

5. Why did you rank problem (name problem that was ranked most difficult

by the student) as 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

knowledge students needed to solve those problems, and offer instructional

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

greatest difficulty for students.

a) What do students need to know in order to solve this problem?

b) What might you do to help students overcome this difficulty?

2. Problem (name the problem that caused difficulty) has been difficult for

many students.

a) Why do you think they considered it to be difficult?

b) What do students need to know in order to solve this problem?

c) What might you do to help students overcome this difficulty?

3. How would you explain the meaning of slope to your students?

Probing questions were used to gain additional information or clarification for

the answers to the questions above.

3.3 Theoretical Framework

The theoretical framework for analyzing student knowledge and difficulties

with linearity and linear functions was inspired by Vergnaud’s theory of

conceptual fields (Vergnaud, 1994, 2009). The conceptual field is a set of

84
situations, schemes for dealing with situations, sets of concepts and theorems

necessary to analyze the operations of thinking, and sets of representations,

formulations and symbolizations (Vergnaud, 1994).

The Linear Conceptual Field (LCF) consists of:

1. Situations that can be modeled mathematically using linear

relationships or linear functions (e.g., rectilinear movement with constant

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

problems 1, 2, 3, 4, and 6).

2. Schemes of actions (e.g., identifying the coordinates of a point situated

on a line graphed in a system of coordinates in MDT problem 4), or theorems

in action (e.g., the Cartesian connection - a point ( x0 , y 0 ) is on the graph of the

line l if and only if its coordinates satisfy the equation of l, y mx  b ) needed

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

problem. For example, in MDT problem 1, to check if a given point ( x0 , y 0 )

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

the statement y0 mx0  b is true.

3. Concepts of linearity and linear function (e.g., concept of slope in

geometric-analytical context in MDT problems 4 and7b, concept of speed in

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

formal mathematical language; symbols, tables, graphs, diagrams). The MDT

problems contain various representations of linear functions (e.g., symbolic

rule in MDT 1, table of values and graphic representation in MDT problem 3).

3.4 Pilot Study

The primary purpose of the pilot study was to gain insight into the nature of

students’ difficulties with MDT problems, and teachers’ understanding of their

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

their four mathematics teachers. Ninety-nine students were in Grade 8, one-

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

to be interviewed. All four teachers were interviewed, together with 10 students

(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

described in Section 3.4.1. The Preliminary Coding Categories for Ranking

Questionnaires (RQ) are presented in Section 3.4.2. The Preliminary Coding

Categories for Interviews are presented in Section 3.4.3.

3.4.1 Scoring Guide (SG) for the MDT. The Scoring Guide (SG) (see

Appendix A) is a modification of the Preliminary Scoring Guide (PSG) developed

by Greenes et al. (2007b) in Massachusetts. Initially each of the seven problems

(1, 2, 3, 4, 5, 6, and 7) of the MDT was scored 20 points. The PSG modifications

involved reconsidering the number of problems for statistical purposes, and

scaling to facilitate interpretation of the dichotomous scores. The analysis of the

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

changed to 0 or 1. One of the advantages of using dichotomous scores (0 and 1) is

the interpretation of the means for MDT problems 3, 4, 5, 6, 7a, 7b as percentages

of participating students with correct answers. Another advantage is the quick

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),

creating the ceiling effect (McBee, 2010).

87
 Also, the Preliminary Scoring Guide was modified for use in this study to

represent participants’ types of responses to Problems 1 and 4. The adaptation

was minor. A description of the adaptation follows.

For MDT problem 1:

a. The modification consisted of including the type of answer different from

“Yes” or “No” for Problem 1a; and

b. Reformulating the descriptions of types of answers for Problems 1b.

For MDT problem 4:

a. Separating the types of answers “Irrelevant information,” “No answer,”

and “I don’t know” into two disjoint categories, “Irrelevant information or

no answer,” and “I don’t know;”

'y
b. Eliminating the type of answer “Computing with incorrect
'x

coordinates;” and

c. Reformulating the descriptions of types of answers.

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

MDT Scores by Problem

Problem Problem Problem Problem Problem Problem Problem 7

1 2 3 4 5 6 a b c

0, .5 or 0, .2, .4, 0 or 1 0 or 1 0 or 1 0 or 1 0 or 0 or 0, .5,

1 .6, .8 or 1 1 1 or 1

As can be seen in Table 1, MDT problems 3, 4, 5, 6, 7a and 7b are scored 0 for

an incorrect response or no response, and 1 for a correct response. MDT problem

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,

2, 3, 4, 5, 6, 7a, 7b, 7c, and ranges from 0 through 9.

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

MDT Problem 1 Types of Responses

PART 1. Correct answer “Yes”

(a) 2. Answer “No”

3. No answer or an answer different from “Yes” or “No”

PART 1. Correct answer using the Cartesian connection

(b) Identifying (2,8) as ( x 0 , y 0 ) and substituting in y 3x  14 ; and

Checking if the statement 8 3(2)  14 (or equivalent statement) is true.

2. Incorrect or incomplete answer while attempting to use the Cartesian

connection

3. Correct answer using a graph

Graphing the line y 3x  14 (using a quick graph or a table of values);

Plotting the point (2,8) ; and
Inspecting to see if the point is on the line.

4. Incomplete or incorrect answer from use of a graph

5. Incomplete or incorrect responses different from those above from Part

(b), #1-4

6. No answer or wrote “I guessed”

90
 As can be seen from Table 2, the type of response for MDT problem 1a is

separated into three non-overlapping categories: “Yes,” “No,” and no answer or

an answer different from “Yes” or “No.” Based on analyses of the performances

of the participants in this present study two problem solving strategies used by

participants who solved MDT problem 1b correctly were identified. These

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

connection strategy implies the identification of the Cartesian coordinates of the

point (2,8) as ( x0 , y 0 ) in the equation of the line y 3x  14 , and checking

whether the equality  8 3(2)  14 is a true statement. The strategy of using a

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

answer or “I guessed” answer.

Table 3 presents types of responses for MDT problem 2 that requires

participants to construct five rows of a table of values, given the linear function

y 2 x  3 .

91
 Table 3

MDT Problem 2 Types of Responses

1. 5 rows correctly completed

2. 4 rows correctly completed

3. 3 rows correctly completed

4. 2 rows correctly completed

5. 1 row correctly completed

6. No row correctly completed, but attempted to answer

7. No answer

The types of responses for MDT problem 2 presented in Table 3 were

determined by the number of correctly completed rows in the table of values.

Response types 6 and 7 differentiated between participants who attempted to

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

errors or lack of understanding of how to complete a table of values for a function

from those who do not try to complete the table.

Table 4 presents types of responses for MDT problem 4 that requires

identification of the slope of a line graphed in a homogeneous system of

coordinates.

92
 Table 4

MDT Problem 4 Types of Responses

1. Correct answer “3”

2. Computational errors in calculation of slope, but correct formula

3. Reversed ratio

4. Incorrect answer different from answers to 2 and 3

5. Irrelevant information

6. No answer or answer, “I don’t know.”

Types of incorrect answers for MDT problem 4 as shown in Table 4 are based

on most common errors observed in previous studies conducted by Greenes et al.

(2007a, 2007b) (e.g., errors in calculation of slope due to incorrect computations

with signed numbers, or presentation of slope in “upside down” or reversed

order, that is, change in x over change in y). The Type 5 code (“Irrelevant

information”) is applied when, for example, qualitative comments (e.g.,

“increasing,” or “positive”) are made without any attempt to slope calculation.

Type 6 code is used when no answer is given, or a statement is made to the effect

that the solver does not know what to do.

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

MDT Problem 7a Types of Responses

1. Answer “R”

2. Answer “S”

3. Correct answer “T”

4. Answer different “R”, “S”, or “T “

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

those mentioned before, and no answer).

Table 6 presents the types of responses for MDT problem 7b, that requires

determination of the slope of a segment of a line graphed in a non-homogeneous

system of coordinates.

Table 6

MDT Problem 7b Types of Responses

1. Correct answer “40”

2. Incorrect answer

3. No answer or irrelevant information

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

show no work are grouped together as a Type 3 response.

Table 7 presents the types of responses for MDT problem 7c that requires

determination of the speed of the car.

Table 7

MDT Problem 7c Types of Responses

1. Correct answer “40mph”

2. Correct numeric answer, but incorrect or no unit of measurement given

3. Incorrect answer

4. No answer or irrelevant information

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

incorrect, the response is coded as Type 3. As with other problems, Type 4

comprises responses when no indication of the correct computation is evident or

when no response is given.

3.4.2 Preliminary Coding Categories for Ranking Questionnaires. For the

Ranking Questionnaires described in Section 3.2 students and teachers

commented on the nature of student difficulties with MDT problems. Students’

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

same Preliminary Coding Categories (PCC) for Ranking Questionnaires (RQ),

presented in Table 8.

Table 8

Preliminary Coding Categories for RQ Written Responses

Category Description Frequency

(Percentage
of Total
N=213)
H (Hard Problem) A statement about the difficulty of a problem or 16 (7.5%)
concept.

M (Mathematical The nature of difficulty refers to lack of 19 (8.9%)

Difficulty) understanding of mathematical concepts, their
representations, or algorithms of solving a
problem.

R (Not A statement about forgetting or not 5 (2.3%)

Remembering) remembering the mathematical concepts or how
to solve the problem.

U (Not A general statement about not understanding a 97 (45.5%)

Understanding) mathematical concept, how to solve a problem,
without offering any further details about the
nature of the difficulty.

W (Whole Domain) A statement about the difficulty of a whole 2 (0.9%)

domain like algebra or mathematics, in general,
with no specifics.

O (Other Other difficulties, different from the ones in the 4 (1.9%)

Difficulties) categories mentioned above.

No Comment 70 (32.9%)

As can be seen in Table 8, there were six preliminary coding categories: H, M,

R, U, W and O. Category H (“Hard Problem”) was used with students’ comments

about general problem difficulty using expressions like, “It’s a hard problem.”
96
Category M (“Mathematical difficulty”) was used when specific difficulties with

mathematical concepts were identified. Category R (“Not Remembering”) was

used when students made comments about forgetting (“I don’t remember”).

Category U (“Not Understanding”) was used when students commented “I don’t

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

at Algebra,” or “Math is hard for me”). Category O (“Other Difficulties”) was

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

responses in categories W and O were small, 2 and 4, respectively.

3.4.3 Preliminary Coding Categories for Interviews. The most difficult

problems for pilot study students were MDT problems 7b and 4, both requiring

determination of slopes of lines, and MDT problem 1, considering whether a point

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

problem 4 correctly, and 53.5 % solved MDT problem 1 correctly. Consequently,

the interviews focused on student knowledge and difficulties with the Cartesian

connection and slope.

Pilot study data, along with results of research studies (Lobato & Siebert,

2002; Moschkovich 1992, Moschkovich et al., 1993; Schoenfeld et al., 1993;

97
Stump 1997, 2001a, 2001b; Zaslavsky et al., 2002), were used to develop

Preliminary Coding Categories for analyzing data from Interviews. The

Preliminary Coding Categories were developed by the researcher using open

coding at the level of paragraph (Strauss & Corbin, 1998). Answers to the

researcher’s questions were considered to be the paragraphs. Responses that did

not address the question were excluded from analysis. Tables 9 and 10 present the

Preliminary Coding Categories referring to conceptual knowledge of Cartesian

connection and slope, respectively.

Table 9

Preliminary Coding Categories for Conceptual Knowledge of Cartesian

Connection

CKC Categories Description

CKC-D A point ( x 0 , y 0 ) is on the graph of the line l if its

Cartesian Connection from
coordinates satisfy the equation of l, y mx  b .
Point to Line

Of note, in Table 9 the Cartesian connection is stated as a mathematical

theorem. However, when solving MDT problem 1, none of the interviewed

participants referred to the mathematical theorem, but rather, to the algorithm for

solving the problem. Three of the interviewed students missed the Cartesian

connection and failed to solve MDT problem 1.

Table 10 presents the Preliminary Coding Categories for conceptual

knowledge of slope.

98
 Table 10

Preliminary Coding Categories for Conceptual Knowledge of Slope

CKS Categories Description

CKS-P Physical Slope describes the steepness of physical objects

Context such as mountain roads, ski slopes, and
wheelchair ramps. A statement about an entire
class of objects sharing the property of having a
slope or being steep, or an instantiation/particular
example is given.

CKS-G Geometric-Analytical Given two points ( x1 , y1 ) and ( x 2 , y 2 ) , the slope

Context of the line connecting the two points is
y 2  y1
m (or “rise over run”).
x 2  x1
Homogeneous Systems of Coordinates: Slope is a
geometric property of the line y mx  b , as well
as an analytical property of the function
y mx  b .
Non-homogeneous Systems of Coordinates: Slope
is an analytical property of the function
y mx  b , and not a geometric property of the
line graphed.

CKS-A Parametric Slope is the parameter m in the function

Context y mx  b .

CKS-T Trigonometric Slope is the tangent of the angle of inclination T ,

Context m tan T .

Slope represents the constant rate of change

CKS-C Covariational 'y
Context m when linear functions in the form
'x
y mx  b are understood from a covariational
point of view: two quantities varying
simultaneously with a constant rate of change.

As can be seen in Table 10, there are five contexts (physical, geometric-

analytical, parametric, trigonometric, and covariational) for slope. The

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

teacher referred to slope as “tangent of an angle,” and another teacher referred to

slope as “rate of change.”

Tables 11 and 12 show the Preliminary Coding Categories representing student

difficulties with the Cartesian connection and slope in MDT problems 1, 4 and 7b.

Table 11 describes student difficulties with the Cartesian connection in MDT

problem 1.

Table 11

Preliminary Coding Categories for Student Difficulties with the Cartesian

Connection

Categories Subcategories Description of

Difficulty

SDC-D SDC-P Identifying (2,8) as

Student Difficulties with Cartesian Identifying the Point ( x 0 , y 0 ) and replacing
Connection from Point to Line
variables with their
values in y 3x  14 .
A point ( x 0 , y 0 ) is on the graph of
the line l if its coordinates satisfy the
equation of l, y mx  b , i.e., if the SDC-T Checking if the
Checking if True statement
statement y 0 mx 0  b statement 8 3(2)  14
is a true statement.
(or other equivalent
statement) is true.

Table 11 shows conceptual knowledge of the Cartesian Connection from point

to line. Initially stated as a mathematical theorem (a point ( x0 , y0 ) is on the graph

of the line l if its coordinates satisfy the equation of l, y mx  b ), the Cartesian

connection was replaced by a “theorem in action” (Vergnaud, 2009) , i.e., a point

100
( x0 , y 0 ) is on the graph of the line l if the statement y 0 mx0  b is a true

statement. The Cartesian connection from point to line was observed in students’

answers and teachers’ comments only as a “scheme-in-action,” a two-step

algorithm for solving MDT problem 1 (identifying (2,8) as ( x0 , y 0 ) , replacing

the variables x, y with their corresponding values x 0 , y 0 in y 3x  14 , and

checking whether the statement  8 3(2)  14 is true). Students’ difficulties

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

because of the need to explain their answers.

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.,

slope is a geometric attribute of a line, Table 12 details only those geometric

difficulties. Students attempted, unsuccessfully, to calculate the slope of the

oblique line using “slope triangles,” i.e., triangles with two vertices being points

on the line, ( x1 , y1 ) and ( x2 , y 2 ) , and the third vertex being either ( x1 , y 2 ) or

( 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

of the line segment to be measured implies having a clearly established starting

101
point, or origin of measurement, and then imagining or concretely laying down as

many non-overlapping, adjacent line segments of one unit length as necessary to

cover the entire line segment to be measured (“exact” measurement, when the line

segment to be measured has a magnitude expressed by an integer). The transport

of the unit of measurement is rigid during the measurement process, i.e., the unit

of measurement is invariant. As mentioned before, in homogenous systems of

coordinates the unit of measurement is the same for both the X-axis and the Y-

axis, while in non-homogeneous systems of coordinates the unit of measurement

for the X-axis and the Y-axis are different. All interviewed students who

encountered difficulties measuring rise or run from “slope triangles” failed to

coordinate the origin, unit, or sense of measurement.

All other students, with analytical perspectives of slope, i.e., slope as an

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

algorithm: 1) identify the Cartesian coordinates ( x1 , y1 ) and ( x2 , y 2 ) of two points

y 2  y1
on the line, and 2) calculate m . The importance of holding an analytical
x2  x1

perspective of slope is crucial in MDT problem 7b where non-homogeneous

systems of coordinates are used (Zaslavsky et al., 2002).

102
 Table 12

Preliminary Coding Categories for Student Difficulties with Slope

Categories Subcategories Description

SDS-R SDS-MH Identifying the rise and the run, and

Making and Measurement in a their magnitudes by measurement in
Calculating a Ratio Cartesian a “slope triangle” (a triangle with
Making the rise Homogeneous System two vertices being points on the line,
(change in y), run of Coordinates ( x1 , y1 ) and ( x 2 , y 2 ) , and the third
(change in x), their vertex being either ( x1 , y 2 ) or
ratio (slope), and
calculating the ratio ( x 2 , y1 ) ). Measurement refers to
“rise over run” coordinating the origin point and the
endpoint of measurement, the unit of
measurement, and the sense of the
measurement.

SDS-MN Identifying the unit of measurement

Measurement in a by taking in consideration the
Cartesian Non- geometric representation of the line
Homogeneous System and counting the units on the Y-axis,
of Coordinates instead of using values of the
function represented by the line.

SDS-CR Calculating the ratio “rise over run.”

Calculating a Ratio

“Making a Ratio” in Table 12 refers to the process of conceiving a function of

an ordered pair of magnitudes, while calculating a ratio refers to the numerical

operation of evaluating the value of the function for a particular pair of

magnitudes (Freudenthal, 1983). All three interviewed students who had

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

that the numbers are “too big,” making calculations difficult.

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

in action” (Vergnaud, 2009), that is, as difficulties with measurement in a “slope

triangle” (identifying the origin, the unit of measurement in homogeneous and

non-homogeneous systems of coordinates, and the sense, positive or negative, of

the axes of coordinates). None of the pilot-study teachers identified the nature of

their students’ difficulties with slope in geometric context.

3.5 Participants

Participants included 1,565 students in Grades 8, 9, and 10, and their 27

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.

Demographics of the two districts are shown in Table 13.

Table 13

Demographics by School District

Hispanic White African- Native- Asian Students

American American Eligible
for Free or
Reduced-
Price
Lunch
District 1 29% 57% 6% 1% 7% 29%

District 2 18% 71% 5% 1% 4% 2%

As can be seen in Table 13, the majority of the population in both districts is

White, 57% and 71%, respectively. There is a large difference in socio-economic

status. The percentage of students eligible for free or reduced lunch is 29% in

District 1 and only 2% in District 2.

3.5.1 Student Demographics. Numbers of students participating in this study

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

Algebra I is often taught in Grade 8.

105
 Table 14

Student Distribution by District and School

District School Number of Students

District 1 Middle School A 259

High School B 263
Total 522

District 2 Middle School C 494

High School D 549
Total 1043

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

almost equal split: 53.3% girls, and 46.7% boys.

Student distribution by grade and current mathematics course is presented in

Table 15.

106
Table 15

Student Distribution by Grade and Current Mathematics Course

Grade 8 9 10

Course Total per Course

Pre-Algebra 457 0 457

Algebra I 2861 2810 80 575

Geometry 141 1351 3060 455

Algebra II 781 78

Total per Grade 757 416 392 Grand Total 1565

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:

i) Group 0 students, representing 67% of the study participants, are enrolled

in mathematics courses at grade level or below grade level, including Grade 8

students enrolled in Pre-Algebra (N=457), Grade 9 students enrolled in

Algebra I (N=281), and Grade 10 students enrolled in Algebra I (N=8) or

Geometry (N=306) .

ii) Group 1 students, representing 33% of the study participants, are

enrolled in more advanced mathematics courses than their grade level

indicates, including Grade 8 students enrolled in Algebra I (N=286 ) or

Geometry (N=14), Grade 9 students enrolled in Geometry (N=135), and

Grade 10 students enrolled in Algebra II (N=78).

Number of participating students by mathematics course, school, and

teacher are presented in Table 16.

107
Table 16

Participating Students by Current Mathematics Course, School, and Teacher

Current Mathematics School Teacher Number of Total

Course Students
Pre-Algebra Middle School C Teacher 1 118
Teacher 2 72
Teacher 3 73
Teacher 4 98
Middle School A Teacher 26 96
457
Algebra I Middle School C Teacher 5 47
Teacher 6 58
Teacher 7 28
High School D Teacher 8 60
Teacher 9 20
Teacher 10 19
Teacher 11 28
Teacher 12 7
Teacher 15 18
High School B Teacher 21 65
Teacher 22 72
Middle School A Teacher 25 126
Teacher 27 27
575
Geometry High School D Teacher 8 20
Teacher 11 39
Teacher 12 18
Teacher 13 3
Teacher 14 60
Teacher 15 37
Teacher 16 23
Teacher 17 74
Teacher 18 45
High School B Teacher 23 38
Teacher 24 84
Middle School A Teacher 27 14
455
Algebra II High School D Teacher 13 1
Teacher 19 53
Teacher 20 24
78
Grand Total
1565

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

School D) teachers (Teachers 1-20) participated in this study. Grade 8 student

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.

In all participating schools, students were placed in mathematics courses based

on their prior performance in mathematics as reflected by course grades and

teacher recommendations, not by age. Therefore, students of various ages and

abilities could be in the same class with the same teacher.

3.5.2 Teacher Demographics. Table 17 shows number of years of teaching

experience and highest college degree attained by teachers participating in this

study.

109
 Table 17

Participating Teachers’ Experience and Degree

Participating Experience Experienc Experienc Experienc Bachelo Master

Teachers < 4 years e e e r
4 to 6 7 to 9 > 9 years
years years
District 1 Middle 0% 0% 33% 67% 33% 67%
School
(N=3)
High 0% 50% 25% 25% 25% 75%
School
(N=4)
District 2 Middle 29% 14% 14% 43% 71% 29%
School
(N=7)
High 46% 8% 8% 38% 54% 46%
School
(N=13)
Overall (N=27) 33% 19% 11% 37% 56% 44%

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

of teaching experience (63%). Comparing districts, District 2 had fewer teachers

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

school teachers participating in this study had degrees in Mathematics or

Mathematics Education. Advanced mathematics courses (e.g., Algebra I,

Geometry in middle schools; Algebra II, Precalculus, Calculus in high schools)

were taught primarily by more experienced teachers with higher academic degrees

(e.g., Master’s in Mathematics Education).

110
3.6 Timeline for Instrument Development, Recruitment of Participants, Data

Collection and Analyses

The pilot study, development of the instruments and interview protocols, the

recruitment of participants, the data collection and the data analyses were

performed according to the schedule shown in Table 18.

Participant recruitment included meetings with district leaders, school

principals, department chairs, teachers, students, and their parents. There were

multiple site visits, discussions, phone calls, and e-mails. The recruitment

materials and the IRB approvals can be found in Appendix C.

Table 18

Timeline for Instrument Development, Recruitment of Participants, Data

Collection and Analyses

Time Period Action

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

December 2008 –February 2009 Recruitment of participants

January 2009 –February 2009 Administration of the MDT, RQS, RQT, IQS,
IQT, Scoring MDTs

February 2009–March 2009 Administration of interviews

April 2009–January 2010 Data Analyses performed

Refinement of the Preliminary Coding Categories,
development of the Final Coding Categories
Inter-Rater Reliability conducted

111
 As can be seen in Table 18, the conduct of this study began in June 2008 and

concluded in January 2010.

The teachers who agreed to participate in this study invited their students to

participate, as well. Student participation was meritorious; overall, more than 70%

of those invited agreed to participate. The Mini-Diagnostic Test (MDT) on

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

February 2009. The MDT was administered to participating students by their

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

administered to participating students during regular class time. Participating

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

coding were performed.

Interviews with participants were conducted within six weeks from the time of

the administration of the MDT. Twenty teachers from District 2, seven middle

school teachers and 13 high school teachers, agreed to be interviewed. Based on

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.

3.7 Methods of Analyses

In this section, the methods of analyses used to answer Research Questions 1-5

are presented, by research question.

Research Question 1:

Does student performance on the Mini-Diagnostic Test (MDT) vary by:

a. Grade level?

b. Current mathematics course?

c. Problem?

Data were analyzed using descriptive statistics, cluster analysis and factor

analysis. To analyze how student performance on MDT varies by grade and

current mathematics course, descriptive statistics (calculations of means and

standard deviations) were used. The analysis of MDT scores using descriptive

statistics led to the consideration of three possible natural clusters: by grade level,

current mathematics course, and performance level. A comparison of the quality

of clusters, using the largest overall average silhouette (Rousseeuw, 1987) as a

113
criterion for determining the best cluster, was performed. To analyze how student

performance on the MDT varies by problem, an Exploratory Factor Analysis

(EFA), R-type, using the Principal Component Analysis method (Hair, Anderson,

Tatham, & Black, 1998) was employed. The R-type EFA “analyzes relationships

among variables to identify groups of variables forming latent dimensions

(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

fewer number of factors, that would be representative in subsequent analyses for

all MDT problems from the old set.

Research Questions 2 and 3:

2. Does student ranking of problems by difficulty, after completing the

MDT, correspond to their performance on the MDT?

3. Does teacher ranking of problems by difficulty, after completing the

MDT, correspond to student performance on the MDT?

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

perceived difficulty, the coefficient of correlation Kendall-tau was used. Kendall-

tau is a non-parametric correlation coefficient that “may be defined as the

114
probability of concordance minus the probability of discordance” (Daniel, 1990,

p. 365).

Research Questions 4 and 5:

4. What is the nature of students’ difficulties?

5. Do teachers’ speculations about the nature of students’ difficulties

correspond to actual student difficulties as identified from student

performance on the MDT and interviews?

For Research Questions 4 and 5, students’ and teachers’ written comments

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

for probing questions to be posed at the interviewer’s discretion; and

2) the opportunity to gain access to students’ and teachers’ thought processes as

they occur. Open coding techniques and procedures described by Strauss and

Corbin (1998) were used to code students’ and teachers’ answers from Research

Questionnaires and Interviews, and develop Preliminary Coding Categories. Of

note is the convergent approach of the type of analyses performed. Replicability

of observations was desired, and therefore Preliminary Coding Categories were

refined and developed into Final Coding Categories (see Appendix B), and raters

were trained and retrained until satisfactory inter-rater agreement was achieved

(see Chapter 4).

Chapter Summary. In this chapter, the instruments adapted or developed

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

MDT, Ranking and Information Questionnaires, and Interviews, together with

Preliminary Coding Categories for RQ written comments and Interviews.

Characteristics of the sample and the general population were presented, together

with the timeline for the design and administration of the instruments, and data

analyses. A presentation of the methods of analyses used to answer the research

questions concluded the chapter. In the next chapter, Chapter 4, data analyses and

results of analyses are presented.

116
 Chapter 4

DATA ANALYSES

4.1 Introduction

The goals of this study were to gain insight into Grades 8, 9, and 10 students’

difficulties with concepts of linearity and linear functions and mathematics

teachers’ understanding of the nature of those difficulties. Participants included

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

administered to all participants in this study: Information Questionnaire (IQ),

Mini-Diagnostic Test (MDT), and Ranking Questionnaire (RQ). The Information

Questionnaire for Students (IQS) requests demographic information, including

current mathematics course. The Information Questionnaire for Teachers (IQT)

requests demographic information, including current and recently taught

mathematics courses, undergraduate college major, and most advanced academic

degree, and number of years of teaching. The Mini-Diagnostic Test (MDT)

consists of nine problems (seven settings; one problem has three parts) focusing

on aspects of linearity and linear functions. The Ranking Questionnaire for

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

identified as most difficult. The Research Questionnaire for Teachers (RQT)

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

to answer the following questions:

1. Does student performance on the Mini-Diagnostic Test (MDT) vary

by:

a. Grade level?

b. Current mathematics course?

c. Problem?

2. Does student ranking of problems by difficulty, after completing the

MDT, correspond to their performance on the MDT?

3. Does teacher ranking of problems by difficulty, after completing the

MDT, correspond to student performance on the MDT?

4. What is the nature of students’ difficulties?

5. Do teachers’ speculations about the nature of students’ difficulties

correspond to actual student difficulties as identified from student

MDT performance and interviews?

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

Questions 1, 4 and 5 inter-rater agreement analyses are addressed first. For

Research Question 1, the reliability of the instrument (MDT) is reported as well.

118
4.2 Research Question 1

Research Question 1: Does student performance on the Mini-Diagnostic Test

(MDT) vary by:

a. Grade level?

b. Current mathematics course?

c. Problem?

4.2.1 Inter-Rater Agreement and Instrument Reliability. All MDTs

(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,

59 MDTs administered to Grade 9 students enrolled in Algebra I or Geometry

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

6, with multiple-choice answers, were excluded from the inter-rater agreement

evaluation.

The following assumptions were made when calculating the inter-rater

agreement (Cohen, 1960):

i) The units of analysis representing students’ answers on MDT are

independent.

119
 ii) The categories represented by student difficulties with MDT problems

1a, 1b, 2, 4, 7a, 7b, 7c are of nominal scale, independent, mutually

exclusive and exhaustive.

iii) The raters operate independently, and are equally competent.

The inter-rater reliability of scoring was measured using Cohen’s Kappa (k)

coefficient of agreement representing the proportion of agreement beyond chance

agreement:

p0  pc
k ,
1  pc

where p 0 = the proportion of units in which the raters agreed, and p c = the

proportion of units for which agreement is expected by chance (Cohen,

1960, pp. 19-20).

The inter-rater agreement for the two raters is presented in Table 19.

Table 19

Inter-Rater Agreement on Scoring of Responses

MDT Problem Observed Kappa(k) 95% Confidence Interval

Lower Limit Upper Limit

Problem 1a .82 .48 1

Problem 1b .83 .64 1

Problem 4 .86 .66 1

Problems 2, 7a, 7b, 7c 1 1 1

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-

problem 1a, with k = .82 (p < .05), 95% CI (.48, 1).

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

problems taken together measure the construct referring to Algebra I student

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

instrument development (Henson, 2001).

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

violation of the independence assumption necessary for regression. A hierarchical

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

MDT score distribution by grade, the corresponding number of students and

percentage are included.

Table 20

MDT Performance by Grade Level

Grade Number of Students Mean Standard

(Percentage of Total N, N=1561) Deviation

8 757 (48.5%) 4.78 2.27

9 416 (26.6%) 5.20 2.50

10 388 (24.9%) 5.45 2.44

All Students 1561 (100.0%) 5.06 2.39

As can be seen in Table 20, about one-half (48.5%) of the participating

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

deviation was about the same for all grades.

To facilitate visualization of the distribution of the MDT scores by grade

level, Figures 1a, 1b, and 1c present the data in histograms.

Figure 1a. Histogram of MDT Scores Distribution for Grade 8 Students

123
Figure 1b. Histogram of MDT Scores Distribution for Grade 9 Students

Figure 1c. Histogram of MDT Scores Distribution for Grade 10 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

than 50 students in Grades 9 and 10 had MDT scores of 9.

4.2.3 Student Performance on the MDT by Current Mathematics Course.

Table 21 presents the means and standard deviations of MDT scores by current

mathematics course (Pre-Algebra through Algebra II), accompanied by the

number of students and corresponding percentage of participating students

enrolled in those courses.

Table 21

MDT Student Performance by Current Mathematics Course

Current Mathematics Number of Students Mean Standard

Course (Percentage of Total N, N=1561) Deviation
Pre-Algebra 457 (29.3%) 3.75 1.98

Algebra I 575 (36.8%) 5.13 2.17

Geometry 452 (29%) 5.74 2.38

Algebra II 77 (4.9%) 8.36 0.97

All Students 1561 (100.0%) 5.06 2.39

As can be seen in Table 21, there were about the same number of students

enrolled in Pre-Algebra (N=457) as in Geometry (N=452). Algebra I had the

125
highest enrollment (N=575), and Algebra II, an advanced course usually taken by

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

all groups of students.

Figures 2a, 2b, 2c, and 2d present histograms of the MDT score distributions

for students enrolled in Pre-Algebra, Algebra I, Geometry, and Algebra II,

respectively.

126
Figure 2a. Histogram of MDT Scores Distribution for Pre-Algebra Students

Figure 2b. Histogram of MDT Scores Distribution for Algebra I Students

127
Figure 2c. Histogram of MDT Scores Distribution for Geometry Students

Figure 2d. Histogram of MDT Scores Distribution for Algebra II 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

MDT scores were associated with higher-level mathematics courses.

4.2.3.1 Algebra I Students’ Performance on the MDT by Grade. Algebra I

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

functions (Arizona Department of Education, 2008; CCSSO & NGA, 2010). To

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

and corresponding percentage of students in each grade who were enrolled in

Algebra I.

Table 22

Algebra I Students’ Performance on the MDT by Grade

Grade Number of Students Mean Standard

(Percentage of Total N, N= 575) Deviation
8 286 (49.7%) 6.33 1.72

9 281 (48.9%) 3.97 1.89

10 8 (1.4%) 2.95 1.66

All Algebra I Students 575 (100%) 5.13 2.17

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%).

Means on the MDT decreased as grade level increased. On average, Grade 8

students correctly solved about six MDT problems; Grade 9 students correctly

solved, on average, about four MDT problems; and Grade 10 students correctly

solved, on average, three MDT problems. On average, Algebra I students (all

grades) correctly solved about five MDT problems. Standard deviations were

about the same for each grade level.

4.2.4 MDT Student Scores by Performance Group. Data collected led to

consideration of three possible natural clusters for grouping students’ MDT

scores: by grade level, current mathematics course, and performance group. There

were two performance groups, Group 0 and Group 1, described in detail in

Chapter 3. Group 0 consisted of students enrolled in mathematics courses at or

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

student performance group than by grade or current mathematics course. The

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.

Average Silhouette 0.6

Figure 3a. Silhouette Measures of Cohesion and Separation –

Clustering by Grade

Average Silhouette 0.5

Figure 3b. Silhouette Measures of Cohesion and Separation –

Clustering by Current Mathematics Course

Average Silhouette 0.7

Figure 3c. Silhouette Measures of Cohesion and Separation –

Clustering by Performance Group

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

good average silhouette of .7.

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

corresponding percentage of students for each performance level.

Table 23

MDT Scores by Performance Level

Performance Level Number of Students Mean Standard

(Percentage of Total N, N=1561) Deviation

Group 0 1021 (65.4%) 4.06 2.03

(Grade Level or Below)

Group 1 540 (34.6%) 6.95 1.83

(Above Grade Level)

All Students 1561 (100.0%) 5.06 2.39

As Table 23 shows, 65.4% of all students were from Group 0. On average,

Group 0 students correctly solved about four MDT problems with a standard

deviation of about 2. On average, Group 1 students, representing 34.6% of the

total number of students, correctly solved about seven MDT problems with a

standard deviation of about 2.

132
 Figures 4a and 4b show histograms of the MDT score distribution by

performance level for students in Group 0 and Group 1, respectively.

Figure 4a. Histogram of MDT Score Distribution for Group 0 Students

Figure 4b. Histogram of MDT Score Distribution for Group 1 Students

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

many high MDT scores.

4.2.5 Student Performance on the MDT by Problem. To gain insight into

the variation of student performance on the MDT by problem, the nine MDT

problem scores were used to perform an Exploratory Factor Analysis (EFA), R-

type, using the Principal Component Analysis method (Hair et al., 1998).

4.2.5.1 Exploratory Factor Analysis. The EFA was performed to:

i) Reduce the original set of variables (problem scores) to a new set with a

fewer number of variables (factors) that could explain the maximum of

the variance of the original set; and

ii) Detect the structure of relationships among the variables representing the

student scores on the MDT problems.

After the data were reduced, the new reduced data were used in all subsequent

analyses performed to answer Research Questions 4 and 5.

Data Types and the Choice of the Statistical Software. The variables

representing the scores for the MDT problems can be considered ordinal under the

assumption that a higher score for a problem corresponds to a higher level of

student performance. All variables representing the scores for the MDT problems

can be considered dichotomous or polytomous. The variables representing the

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

dichotomous and polytomous variables is better expressed by tetrachoric or

polychoric correlations (Holgado–Tello, Chacón–Moscoso, Barbero–García, &

Vila–Abad, 2008). “The polychoric correlation estimates the linear relationship

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

continuously” (Uebersax, 2000). The Exploratory Factor Analyses (EFA) were

performed using MicroFACT2.0, appropriate for tetrachoric and polychoric

correlations (Uebersax, 2006; Yaffee, 2003).

About the Sample. The sample size is large (N=1561), suitable for EFA

(Costello & Osborne, 2005). It exceeds the criterion of a minimum of 20

observations per variable (Hair et al., 1988).

Table 24 shows the means and standard deviations for the scores for MDT

problems 1, 2, 3, 4, 5, 6, 7a, 7b, and 7c.

135
 Table 24

Means and Standard Deviations for Scores on MDT Problems

Problem Mean Standard Deviation

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

As can be seen in Table 24, there is no mean value indicative of outliers

(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

easy, thus making it more difficult to detect true differences in student

performance. The most difficult MDT problems, with respect to student

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,

and require making connections among representations. The standard deviations

136
for all nine scores of the MDT problems ranged from .39 to .5, with little

variation.

Assumptions and the Correlation Matrices. The basic assumptions, i.e.,

normality, homoscedasticity, and linearity are “more conceptual than statistical”

(Hair et al., 1998, p. 99) and necessary as long as their violation might lower the

observed correlations. “Only normality is necessary if a statistical test is applied

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

assessed by examining the matrices of correlation among variables.

Table 25 shows the teatrachoric and polychoric correlations among the scores

on the nine MDT problems 1, 2, 3, 4, 5, 6, 7a, 7b, and 7c.

137
 Table 25

Correlations among Scores on the MDT Problems (MicroFACT)

Problem 1 2 3 4 5 6 7a 7b 7c

1 1.00 .56 .46 .46 .31 .26 .37 .45 .41

2 1.00 .57 .49 .44 .30 .48 .45 .51

3 1.00 .37 .42 .13 .41 .36 .45

4 1.00 .32 .43 .46 .55 .44

5 1.00 .09 .35 .32 .40

6 1.00 .37 .40 .29

7a 1.00 .50 .59

7b 1.00 .65

7c 1.00

Table 25 shows that most of the correlations are greater than .3 (rounded),

therefore the appropriateness of EFA is justified. The minimum correlation (.09)

is between MDT multiple-choice problem 5 (finding the linear function given a

table of values) and problem 6 (finding the y-intercept given the equation of a line

in standard form). The maximum correlation (.65) is between MDT problem 7b

(finding the slope given the graph of a line) and problem 7c (finding the speed

given the graph representing the change in distance as a function of change in

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

been theoretically justified. However, those correlations express the empirical

138
relationships in the data, how the students thought about those problems, and not

theoretical relationships. Also high are the correlations .56 and .57 for MDT

problems 1 and 2, and problems 2 and 3, respectively. MDT problems 1, 2, and 3

illustrate the Cartesian connection.

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

calculated. Table 26 shows those correlations.

Table 26

Correlations of Problems with MDT Total Score

Problem 1 2 3 4 5 6 7a 7b 7c

Correlation .62 .78 .78 .66 .73 .59 .79 .65 .70
with MDT
Total Score

In Table 26, all correlations are high, ranging from .6 to .8 (rounded).

Therefore no variable was dropped from analysis.

As mentioned before, the internal consistency of the MDT was determined

using Cronbach’s Alpha as a measure of the correlation between observed and

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

more complicated structures, while fewer factors extracted lead to simpler

structures but account for smaller percentages of the explained variance. The

criteria applied when determining the number of factors to extract (Hair et al.,

1998) were the following:

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

criterion was applied for determining the initial number of factors

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

of the initial eigenvalues from Table 27.

Figure 5. Scree Plot of the Initial Eigenvalues

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

with absolute values between 0 and 1.

According to the Scree Test Criterion, two factors account for most of the

variance, therefore two factors should be extracted.

iii. Percentage of Variance Criterion (Hair et al., 1998, p. 104) takes into

consideration the percentage of total explained variance acceptable for

research (above 95% in natural sciences, 60% and below in social

sciences), and extracts the number of factors accounting for this a priori

established amount of cumulative variance.

According to this criterion, a two-factor solution accounts for 49% of the

explained cumulative variance, a moderate, acceptable value in educational

studies (Hair et al., 1998).

iv. Comprehensibility Criterion aims to ensure the parsimony of the model,

and therefore restricts the number of factors extracted, usually to two or

three. At the same time, a factor that accounts for only one or two

variables is weak. Therefore it is desirable to have factors that account for

at least three variables. Since there are nine variables, the number of

factors is restricted to a maximum of three. Trials with three factors

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.

The VARIMAX rotation method is used for easier interpretation of the

142
 factors. With the VARIMAX method, high loadings are expected in each

column, close to 1 or -1, indicating associations between variable and

factor, and small loadings, close to 0, indicating lack of association

between variable and factor (Hair et al., 1998, p. 110).

Table 28

VARIMAX Rotated Factor Loadings (Three Factors)

Problem Factor 1 Factor 2 Factor 3

1 .56 .16 .33

2 .73 .21 .30

3 .69 .24 .08

4 .39 .25 .56

5 .49 .28 .05

6 .06 .14 .68

7a .37 .46 .37

7b .29 .54 .49

7c .35 .80 .23

Table 28 indicates that MDT problems 1, 2, 3, and 5 loaded significantly on

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.

According to Comprehensibility Criterion, and given the small number of

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

homogenous, and more factors may be purposefully extracted. It is

expected that factors reflecting the variables that are more homogeneous

across the sample will be extracted first, while later factors will be loaded

with variables discriminating across the sample.

Table 29 shows the means of scores of the nine MDT problems for all students

(Overall) and for students from Group 0 and Group 1.

Table 29

Means of Scores of MDT Problems by Group of Students

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

As can be seen in Table 29, Group 1 outperformed Group 0. Indeed, all

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.

vi. Summary of Factor Selection Criteria. The prevailing criterion is to obtain

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,

which limited the analysis to two-factor analysis. Table 30 shows the

factor loading and final communalities estimates when two factors are

extracted.

Table 30

Unrotated Factor Loadings

Problem Factor 1 Factor 2 Final Communality Estimates

1 .63 .10 .41

2 .75 .23 .62

3 .65 .39 .57

4 .68 -.18 .50

5 .51 .25 .32

6 .45 -.38 .34

7a .69 -.10 .48

7b .74 -.28 .62

7c .74 -.06 .55

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

homogeneity of the sample with respect to those variables. Of note, regarding

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

y-intercept of a line when given the linear equation ax  by c . The final

communality estimates (last column) indicates how much variance in a variable is

accounted for by the two-factor solution. Note that a high communality of a

variable indicates the reliability of the factor analysis, i.e., the variable is well

represented by the extracted factors. Low communalities are acceptable if the

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

unaccounted for by the two- factor solution.

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

(close to 0) in each column of the factor matrix. Therefore a rotation was

performed using the VARIMAX method.

Table 31 shows the VARIMAX rotated factor matrix.

Table 31

VARIMAX Rotated Factor Loadings (Two Factors)

Problem Factor 1 Factor 2

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

that can be easily interpreted. MDT problems 1, 2, 3, and 5 loaded significantly

on Factor 1 (factor loadings > .5). Conceptually, all of those problems are situated

across the dimension of connections among representations. Indeed, MDT

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

function. MDT problem 5 requires connecting a description of a real-life situation

to its symbolic representation, a linear function. Given these theoretical

considerations, Factor 1 was labeled Connections among Representations.

MDT problems 4, 6, 7a, 7b, and 7c loaded significantly on Factor 2 (factor

loadings > .5). Conceptually, the corresponding MDT problems require

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

identification and interpretation of the rate of change/slope and the y-intercept of a

linear function in various contexts. Therefore Factor 2 was labeled Linear

Functions.

Table 32 illustrates the structure obtained using the two-factor analysis.

148
 Table 32

Labeled Factor Matrix

Problem Factor 1 Factor 2

Connections among Representations Linear Functions

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

the Connections among Representations dimension, and MDT problems 4, 6, 7a,

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

loadings (< .5) were suppressed.

Validation of Factor Analysis. Exploratory factor analyses were performed

separately for Group 0 and Group 1 students.

Table 33 and Table 34 present the rotated factor loadings, using the

VARIMAX method, for Group 0 and Group 1, respectively.

149
Table 33

Rotated Factor Loadings for Group 0

Problem Factor 1 Factor 2

Connections among Representations Linear Functions
1 .48

2 .72

3 .64

4 .48

5 .50

6 .33

7a .53

7b .72

7c .51

Table 34

Rotated Factor Loadings for Group 1

Problem Factor 2 Factor 1

Linear Functions Connections among Representations

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,

and 7c were grouped along the Linear Functions factor.

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.

Notwithstanding the reduction in the number of variables, the relationships

between variables remained the same for both groups of students, i.e., the

variables common to both groups were associated in the same way.

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

surrogate variables, uncorrelated if possible (Hair et al., 1998, p. 129).

The magnitude of loading is decisive in selecting the surrogate variable. As

such, the following surrogate variables were selected:

a) For Group 0, MDT problem 2, with loading .72 (Table 33), was

selected as surrogate variable for the Connections among

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

the Connections among Representations factor was MDT problem 1

with loading .81 (Table 34). MDT problem 4 with loading .71 (Table

34) was selected as surrogate variable for the Linear Functions factor.

From a theoretical point of view, the choice of surrogate variables is easy to

defend. Both MDT problems 1 and 2 illustrate the Cartesian connection, which is

representative of Connections among Representations. Both MDT problems 7b

and 4 focus on slope, which is representative of Linear Functions.

4.2.5.2 Variation of Student Performance on the MDT Representative

Problems by Grade and Current Mathematics Course. After data reduction, the

analyses about the nature of student difficulties (Research Questions 4 and 5)

considered only student performance on MDT problems 2 and 7b for Group 0

students and MDT problems 1 and 4 for Group 1 of students.

Table 35 presents the means of scores for MDT problems 2 and 7b by grade

for Group 0 students.

152
 Table 35

Mean Scores for MDT Problems by Grade for Group 0

Grade 8 Grade 9 Grade 10

(N=457) (N=263) (N=301)

Problem 2 .47 .49 .57

Problem 7b .11 .14 .24

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

9 students, and 24% of Grade 10 students.

Table 36 shows the means of scores for MDT problems 2 and 7b by current

mathematics course for Group 0 students.

Table 36

Mean Scores for MDT Problems by Current Mathematics Course for Group 0

Pre-Algebra Algebra I Geometry

(N=457) (N=271) (N=293)

Problem 2 .47 .49 .58

Problem 7b .11 .14 .25

It is easily noticeable that the means of scores in Table 36 are almost

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

Algebra I, and 97% of Grade 10 students were enrolled in Geometry. A

comparison of student performance on the MDT by current mathematics course

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

problems 2 and 7b remained small, at most .11.

Table 37 presents mean scores for MDT problems 1 and 4 by grade for

students in Group 1.

Table 37

Mean Scores for MDT Problems by Grade for Group 1

Grade 8 Grade 9 Grade 10

(N=300) (N=153) (N=87)

Problem 1 .68 .80 .94

Problem 4 .46 .70 .77

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

mastered by Grade 10 students with a mean score of almost 1 (.94). Although

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

students did not solve MDT problem 4.

Table 38 shows mean scores for MDT problems 1 and 4 for students in

Group1 by current mathematics course.

Table 38

Mean Scores for MDT Problems by Current Mathematics Course for Group 1

Algebra I Geometry Algebra II

(N=304) (N=159) (N=77)

Problem 1 .67 .83 .94

Problem 4 .45 .72 .79

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

were enrolled in Algebra I, 88% of Grade 9 students were enrolled in Geometry,

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.

Following is a summary of the results for Research Question 1:

1. There was little variation in student performance on the MDT from grade

8 through Grade 10. Mean scores increased from 4.78 to 5.45,

representing, on average, less than one more MDT problem correctly

solved by Grade 10 students as compared with Grade 8 students.

155
2. Student performance on the MDT increased with more advanced courses,

from Pre-Algebra to Algebra II. Mean scores more than doubled from

3.75 (Pre-Algebra students) to 8.36 (Algebra II students). Algebra II

students correctly solved, on average, more than eight of the nine MDT

problems.

3. Group 1 students enrolled in mathematics courses above their grade level

outperformed Group 0 students enrolled in courses at or below their grade

level. The difference on mean scores was 2.88, representing, on average,

about three more MDT problems correctly solved.

4. The most difficult problem was identifying the slope of a line from a

graph (MDT problem 4). Student performance on MDT problems 7b and 4

representative of the Linear Functions factor, remained very low across

grades and current mathematics courses for Group 0 students (mean

scores between .11 and .25) and low to moderately high across grades and

current mathematics courses for Group 1 students (mean scores between

.45 and .79). Student performance on MDT problems 2 and 1,

representative of the Connections among Representations factor,

remained low to moderate across grades and current mathematics courses

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

Research Question 2: Does student ranking of problems by difficulty, after

completing the MDT, correspond to their performance on the MDT?

All participating students (N=1561) completed the MDT and Ranking

Questionnaire for Students (RQS) described in Chapter 3. The MDTs were

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

Questionnaires for Students were used to calculate frequencies of responses by

MDT problem, and obtain the ranking of MDT problems by difficulty, as

perceived by students. The most difficult MDT problem, as perceived by students,

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

ranking of MDT problems by difficulty as perceived by students, the coefficient

of correlation Kendall-tau was used (Daniel, 1990).

Tables 39 and 40 present the comparison between student performance on 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

Group 0 Performance on the MDT vs. Ranking of Problem Difficulty by

Students

Ranking by Problem Ranking by Problem

MDT (Mean) RQS (Number of Students)

performance

I 4 (.24) I 1 (323)

II 7 (.38) II 2 (158)

III 1 (.39) III 7 (136)

IV 6 (.41) IV 4 (105)

V 2 (.50) V 6 (112)

VI 3 (.69) VI 3 (38)

VII 5 (.76) VII 5 (16)

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

.38). However, students ranked MDT problem 1 (N=323) as most difficult,

followed by MDT problem 2 (N=158). MDT problems 3 and 5 were least

difficult, and were identified by Group 0 students as being least difficult for them.

Of note, when ranking MDT problems by difficulty, students considered MDT

problem 7 as being one problem, therefore for ranking analyses, problem 7 was

considered as one problem.

158
 Table 40

Group 1 Performance on the MDT vs. Ranking of Problem Difficulty by

Students

Ranking by Problem Ranking by Problem

MDT performance (Mean) RQS (Number of Students)

I 4 (.58) I 7 (187)

II 6 (.71) II 2 (96)

III 7 (.74) III 4 (82)

IV 1 (.76) IV 6 (81)

V 2 (.89) V 1 (33)

VI 5 (.91) VI 3 (21)

VII 3 (.95) VII 5 (18)

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,

followed by MDT problem 2 ranked most difficult by 96 students. MDT problems

3 and 5 were the least difficult for Group 1 students, and were identified by them

as being the least difficult.

To measure the correlation between the ranking of MDT problem by mean

scores and the ranking of MDT problems by difficulty as perceived by students,

the non-parametric coefficient of correlation Kendall-tau, W , was calculated. A

value of 0 for W would mean independence for the two ordinal variables

representing rankings. A value of 1 for W would correspond to perfect agreement.

159
By contrast, a value of  1 would correspond to perfect disagreement (Daniel,

1990, pp. 365-367). The hypothesis (one-sided) that the variables representing the

two rankings are independent was tested:

H0: The variables representing the ranking of mean scores for MDT problems,

and the ranking of MDT problems by difficulty as perceived by students, are

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

students at α = .05 (a priori established). Therefore, for Group 0 students, the

hypothesis that the variables representing the two rankings are independent could

not be rejected. Thus, there is no relationship between the variables representing

the ranking of mean scores of MDT problems and the ranking of MDT problems

by difficulty as perceived by students. However, for Group 1 students, the

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

ranking of MDT problems by difficulty as perceived by students.

The following summarizes the results for Research Question 2:

1. There is no agreement for Group 0 students and moderate agreement for

Group 1 students between MDT performance and ranking of MDT

problems by difficulty.

160
 2. For both groups of students, MDT problems 3 and 5 (with the highest

mean scores) were identified as the least difficult by students, whereas

MDT problems 4, 6, and 7 (with the lowest mean scores) were not

identified as most difficult for students.

3. All students could correctly identify the least difficult problems, but only

Group 1 students could identify some of the difficult problems.

4.4 Research Question 3

Research Question 3: Does teacher ranking of problems by difficulty, after

completing the MDT, correspond to student performance on the MDT?

All participating teachers (N=26) completed the Ranking Questionnaire for

Teachers (RQT). The RQT is described in Chapter 3. Mean scores by MDT

problem were calculated for all students, without distinguishing by group of

students (Group 0 or Group 1). The lack of distinction was necessary because the

teachers’ Ranking Questionnaire responses did not discriminate between groups

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,

was assessed using the coefficient of correlation Kendall-tau (Daniel, 1990).

Table 41 presents the comparison of rankings by student performance on the

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)”

column, and with frequencies of the RQT responses in “Problem (Number of

Teachers)” column.

Table 41

Student Performance on the MDT vs. Ranking of Problem Difficulty by

Teachers

Ranking by Problem Ranking by Problem

MDT performance (Mean) RQT (Number of Teachers)

I 4 (.35) I 7 (15)

II 6 (.51) II 6 (6)

III 1 (.52) III 5 (2)

IV 7 (.57) IV* 1 (1)

V 2 (.64) IV* 2 (1)

VI 3 (.78) IV* 4 (1)

VII 5 (.81) VII 3 (0)

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

considered MDT problem 7 as one problem. Therefore, for ranking analyses

purposes this problem was considered as one problem.

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

ranked third in terms of difficulty by teachers.

To assess the correlation between the ranking of MDT problem by mean

scores and teachers’ ranking of MDT problems by difficulty as perceived by

students, the non-parametric coefficient of correlation Kendall-tau, W , was

calculated, and the hypothesis (one-sided) that the variables representing the two

rankings are independent was tested:

H0: The variables representing the ranking of mean scores for MDT problems,

and teachers’ ranking of MDT problems by difficulty as perceived by students are

independent ( W 0 ).

H1: W ! 0 .

The test statistic W was calculated using PASW18, and found to be .10, and

statistically non-significant at α = .05 (a priori established). Therefore, the

163
hypothesis that the variables representing the two rankings are independent could

not be rejected. In conclusion, there is no relationship between the variables

representing the ranking of mean scores of MDT problems and teachers’ ranking

of MDT problems by difficulty as perceived by students.

In summary, the results for Research Question 3 are the following:

1. There is no agreement between student performance on the MDT and teacher

ranking of MDT problems by difficulty as perceived by their students.

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

rate of change. According to student performance, 7b was most difficult for

students. Twice as many students correctly solved 7a and 7c, than 7b.

3. Teachers did not consider MDT problem 4 difficult for their students.

According to student performance, problem 4 was surpassed in difficulty only

by 7b. Both problems 4 and 7b require the calculation of the slope of a line

when the graph is given.

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

6 were about the same.

164
4.5 Research Question 4

Research Question 4: What is the nature of students’ difficulties?

All participating students (N=1561) completed the MDT and Ranking

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

understanding of student difficulties with MDT aspects of linearity and linear

functions, the researcher selected 40 students to be interviewed, two from each

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

B) and analyzed using the Linear Conceptual Field, a theoretical framework

inspired by Vergnaud’s conceptual field theory (1994, 2009).

4.5.1 Coding and Identifying the Nature of Student Difficulties from RQS.

The Ranking Questionnaires for Students (RQS) included students’ descriptions

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

Preliminary Coding Categories (see Table 8, Chapter 3) the researcher and a

second rater coded a sample of 50 Ranking Questionnaires and agreed to Final

Coding Categories by reconsidering the definitions and boundaries of the

165
Preliminary Coding Categories. Table 42 presents descriptions of the Final

Coding Categories for RQ written responses (see also Appendix B).

Table 42

Final Coding Categories for RQ Written Responses

Category Description

H (Hard Problem) A general statement about the difficulty of a problem, topic or

domain. Sometimes a mathematical concept is identified as
“hard” or “difficult.” Other times a reference to unsatisfactory
performance is made, but no specific details about the nature of
the difficulty are given.

L (Never A statement about the novelty of a problem, or that a

Learned/Taught) mathematical concept or how to solve a certain problem was not
learned or taught.

M (Mathematical Nature of difficulty refers to lack of understanding of

Difficulty) mathematical concepts or their representations, or difficulties
with algorithms/procedures, computations, and symbolic
manipulations.

R (Not A statement about forgetting or not remembering the

Remembering) mathematical concepts or how to solve the problem.

T (Thinking Time) A reference to the amount of time, thinking, work, memory or

any other resource necessary to solve a problem, or to a problem
feature, as for example the number of subparts to be completed.

U (Not A general statement about not understanding a mathematical

Understanding) concept, topic, domain, how to solve a problem, or the
vocabulary used, without offering any further details about the
nature of the difficulty.

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

major changes consisted of eliminating the preliminary categories W (“Holistic

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

Time”); and reconsidering the boundaries of the categories H (“Hard Problem”),

M (“Mathematical Difficulty”), and U (“Not Understanding”). Comments in the

category W referring to whole domains as being “hard” were moved into

category H, and comments referring to ”not understanding” whole domains were

moved into category U. Also, comments referring to not understanding the

vocabulary were moved from category O to category U. Both categories H and U

were redefined to include the comments from the old category W. The remaining

comments falling initially in category O, referring to the amount of thinking time

necessary to solve a problem, were moved into a new category, T, whose

boundaries were enlarged by adding comments about the amount of work or

memory needed to solve a problem. Finally, the category M was redefined by

adding difficulties with algorithms/procedures (or steps of those), and with

calculations or symbolic manipulations.

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

conducted by the researcher and the second rater. An inter-rater reliability

analysis using the Kappa statistic (Cohen, 1960) was performed to determine

consistency among raters. The same assumptions were made for calculating the

inter-rater agreement on RQ, as when calculating the inter-rater agreement on

MDT scoring (Cohen, 1960):

i) The units of analysis representing participants’ answers on RQ are

independent.

167
 ii) The categories representing the nature of student difficulties are of

nominal scale, independent, mutually exclusive and exhaustive.

iii) The raters judge independently, and are equally competent.

The inter-rater agreement for the two raters is presented in Table 43.

Table 43

Inter-Rater Agreement on RQ Written Responses

Group of Participants Observed 95% Confidence Interval

Kappa (k)
Lower Limit Upper Limit

Students Group 0 .94 .91 .98

Students Group 1 .90 .81 .99

As shown in Table 43, the inter-rater agreement on coding of RQ student

responses was calculated for each of the following groups of participating

students: Group 0 students who were enrolled in mathematics courses at their

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

mathematics curriculum sequence. The inter-agreement value for coding

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.

Analyses of the nature of student difficulties with MDT problems were

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

difficulties with MDT problems 1 and 4, respectively. Examples of students’

comments, by category of difficulty, are included in the tables. Comments have

the following format: “Student written comment [problem number].” For

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

of students with comments in a category, followed by the corresponding

percentage in parentheses. Before each table, the text of each problem is

reproduced to aid the reader (see Figure 6).

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?

Figure 6. MDT Problems 2 and 7

170
Table 44

Group 0 Student Difficulties with MDT Problems 2 and 7b

Category Examples Frequencies by Problem

Problem 2 Problem 7b

(N=158) (N=136)

H (Hard “I thought that one was the hardest 3(1.9%) 5(3.7%)

Problem) because those types are just hard to
me.”[2]
“This problem was difficult because
the slope was difficult to find.”[7b]

L (Never “I didn't know from where to start. I 5(3.2%) 1(0.7%)

Learned/Taught) didn't know the first step of the
problem. I also haven't learned that
problem yet.”[2]
“I hadn't seen anything like it. Don't
get me wrong it was easy just
new.”[7]

M (Mathematical “You had to figure out what numbers 19(12%) 33(24.3%)

Difficulty) you had to put in because it doesn't
tell you.” [2]
“There was no x value.”[2]
“I got confused on whether the slope
was supposed to be 4/3 or 120/3.“
[7b]

R (Not “Forgot how to solve.”[2] 15(9.5%) 9(6.6%)

Remembering) “I couldn't remember how to do the
slope or how to find the slope of a
line. Had a mind blank.” [7b]

T (Thinking “At first I didn't know what it was 6(3.8%) 12(8.8%)

Time) asking so it took more thought then
most questions.”[ 2]
“It took a long time for one
problem.”[7]

U (Not “Your wording was weird and I just 70(44.3%) 51(37.5%)

Understanding) didn't understand it!”[2]
“I chose #7 because I do not
understand slope very much at
all.”[7b]
No comment 40(25.3%) 25(18.4%)

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

numbers to pick for x-value”).

More than half of the 136 students from Group 0 who ranked MDT problem 7

as most difficult, either made no comment on the nature of the difficulty or

commented, “I don’t understand.” Students’ comments about the mathematical

difficulty of MDT problem 7b identified the difficulty of figuring out which

numbers to use in the formula to compute the slope (e.g., “In part b I didn't know

whether to take from the values on the side or the units”).

Note that about 16% of the 1040 students from Group 0 correctly solved MDT

problem 4 requiring the calculation of the slope of a line graphed in a

homogeneous system of coordinates, but failed to solve MDT problem 7b

requiring calculation of the slope of a line graphed in a non-homogeneous system

of coordinates. Also, it may be that problem context played a crucial role in

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

the line (a mathematical context).

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

and 4 are presented (see Figure 7) before Table 45.

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.

Figure 7. MDT Problems 1 and 4

173
 Table 45

Group 1 Student Difficulties with MDT Problems 1 and 4

Category Examples Frequencies by Problem

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]

M (Mathematical “I have never heard the question 10(10.4%) 32(39%)

Difficulty) asked that way. ‘On the graph’
could be anywhere.”[1]
“It wasn't a grid to help.”[4]
“I was not sure whether or not to
put in decimal or keep it in fraction
form.”[4]

R (Not “I forgot how to do this type.” [1] 22(22.4%) 18(22%)

Remembering) “I couldn't remember the formula
for slope.”[4]

T (Thinking “It took me a long time to 15(15.6%) 9(11%)

Time) remember how to solve it.”[1]
“It was the problem I had to think
. on. All were easy just that problem
made me think.”[4]

U (Not “I didn't know it.”[1] 38(39.6%) 13(15.9%)

Understanding) “No understandings of slope.”[4]

No comment 10(10.4%) 7(8.5%)

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

their difficulty or made a comment about “Not Understanding.” Only 10 students

commented about the problem’s “Mathematical Difficulty.” Those students

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

situated on the line y 3x  14 (e.g., “I didn't understand at first what x and y

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,

referring to the Cartesian connection from line to points.

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

the difficulty or made a comment about “Not Understanding.” Two-fifths of the

students who identified MDT problem 4 as most difficult commented on their

mathematical or computational difficulty. Students identified the following

obstacles concerning measurement: identifying the point where the measurement

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

number goes on top or bottom”), assigning decimal numbers as magnitudes for

rise or run, and then computing with those (e.g., “With the decimal it made it

harder to turn into a slope”).

Analyses of students’ written comments on the nature of their difficulties

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

student written comments:

175
 1. More than one-half of students did not describe the nature of their

difficulties.

2. About 28% of students described the nature of their difficulty by referring

to the amount of time necessary to solve the problem, or difficulty

remembering how to solve the problem.

3. Only 20% of students described their specific mathematical or

computational difficulty.

4. For MDT problem 1 students had difficulty identifying (2,8) as

coordinates of a point ( x0 , y 0 ) , and checking whether the point (2,8) is

situated on the line y 3x  14 .

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

change in y, since they had to decide between counting “tick marks” or

taking the y values from the Y-axis into consideration.

4.5.2 Coding and Identifying Students’ Mathematical Difficulties from

Interviews. To gain greater understanding about students’ conceptual knowledge

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

Knowledge of Cartesian Connection (CKC).

Table 46

Final Coding Categories for Conceptual Knowledge of Cartesian Connection

CKC Categories Description

CKC-D A point ( x 0 , y 0 ) is on the graph of the line l if its

From Point to Line coordinates satisfy the equation of l, y mx  b , i.e.,
if the statement y 0 mx 0  b is a true statement.

CKC-C If a point ( x 0 , y 0 ) is on the graph of the line l, then its

From Line to Points coordinates satisfy the equation of l, y mx  b .
There is an infinite number of points on a line, i.e.,
there is an infinite number of points ( x 0 , y 0 ) on the line
l satisfying the equation of the line l, y mx  b .

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

connection from line to points in the situation from MDT problem 2.

Table 47 describes the Final Coding Categories for Conceptual Knowledge of

Slope (CKS).

177
 Table 47

Final Coding Categories for Conceptual Knowledge of Slope

CKS Categories Description

CKS-P Physical Slope describes the steepness of physical objects such as

Context mountain roads, ski slopes, and wheelchair ramps. A
statement about an entire class of objects sharing the
property of having a slope or being steep, or an
instantiation/particular example is given.

CKS-G Geometric- Given two points ( x1 , y1 ) and ( x 2 , y 2 ) , the slope of the

Analytical y 2  y1
Context line connecting the two points is m ( or “rise
x 2  x1
over run”).
Homogeneous Systems of Coordinates: Slope is a
geometric property of the line y mx  b , as well as an
analytical property of the function y mx  b .
Non-homogeneous Systems of Coordinates: Slope is an
analytical property of the function y mx  b , and not a
geometric property of the line graphed.

In Table 47, there are only two final categories for conceptual knowledge of

slope: 1) slope in a physical context and 2) slope in a geometric-analytical

context. This is not surprising since Algebra I textbooks, used by participants in

this study, introduce the concept of slope in physical and in geometric-analytical

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

(CKS-G). In the Preliminary Coding Categories (see Chapter 3) the conceptual

knowledge of slope referred to slope in other contexts, parametric, trigonometric,

and covariational (CKS-A, CKS-T, and CKS-C). However, the inter-agreement

analyses revealed that participants’ answers referring to slope in other contexts

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

to two contexts, physical and geometric-analytical.

Table 48 presents the Final Coding Categories for Student Difficulties with the

Cartesian Connection (SDC) in the situations from MDT problems 1 and 2.

Table 48

Final Coding Categories for Student Difficulties with the Cartesian

Connection

Categories Subcategories Description of Difficulty

SDC-D SDC-DP Identifying (2,8) as

From Point to Line Identifying the ( x 0 , y 0 ) and replacing
A point ( x 0 , y 0 ) is on the graph Point
variables with their values
of the line l if its coordinates in y 3x  14 .
satisfy the equation of l,
y mx  b , i.e., if the statement
SDC-DT Checking if the statement
y 0 mx 0  b Checking if 8 3(2)  14
is a true statement. True statement (or other equivalent
statement) is true.

SDS-C SDC-CX Identifying values for x.

From Line to Point(s) Identifying X-
If a point ( x 0 , y 0 ) is on the values
graph of the line l, then its
coordinates satisfy the equation SDC-CY Using x-values to
of l, y mx  b . Calculating Y- calculate the
There is an infinite number of values corresponding y-values ,
points on a line, i.e., there is an so that ( x 0 , y 0 ) satisfies
infinite number of points ( x 0 , y 0 ) y 2 x  3 .
on the line l satisfying the
equation of the line l, y mx  b .

As shown in Table 48, students’ difficulties with the Cartesian connection

occurred at each of the steps of the algorithm. Therefore, each of coding

179
subcategory represents one of these steps. In MDT problem 1, students had

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.

In MDT problem 2, students had difficulty identifying the x-values necessary to

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

Slope (SDS) by category and subcategory.

Table 49

Final Coding Categories for Student Difficulties with Slope

Categories Subcategories Description

SDS-R SDS-MH Identifying the rise and the run, and

Making and Measurement in a their magnitudes by measurement in
Calculating a Ratio Cartesian a “slope triangle” (a triangle with
Making the rise Homogeneous System two vertices being points on the line,
(change in y), run of Coordinates ( x1 , y1 ) and ( x 2 , y 2 ) and the third
(change in x), their vertex being either ( x1 , y 2 ) or
ratio (slope), and
calculating the ratio ( x 2 , y1 ) . Measurement refers to
“rise over run” coordinating the origin point and the
endpoint of measurement, the unit of
measurement, and the sense of
measurement.

SDS-MN Identifying the unit of measurement

Measurement in a by taking in consideration the values
Cartesian Non- of the function represented by the
Homogeneous System line, and not the geometric
of Coordinates representation of the line.

SDS-CR Calculating the ratio rise/run

Calculating a Ratio

180
 The categories identifying student difficulties with slope presented in Table 49,

refer only to the geometric-analytical context proposed by MDT problems 4 and

7b.

An inter-rater agreement analysis using the Kappa (k) statistic (Cohen, 1960)

was performed to determine the consistency among raters.

Inter-rater agreement analyses for Conceptual Knowledge (CKC, CKS) and

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.

4.5.2.1 Student Difficulties with the Cartesian Connection. Mathematical

knowledge of the Cartesian connection can be stated as a mathematical theorem

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.

4.5.2.1.1 Student Difficulties with the Cartesian Connection in MDT

Problem 1. Nine of the interviewed students from Group 1 solved MDT problem

1 using the following algorithm:

Step 1. Substitute the ordered pair (2,8) for ( x0 , y 0 ) in y 3x  14 ; and

Step 2. Check whether the statement  8 3(2)  14 is true.

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”).

The following is a description of a student’s successful solution of MDT

problem 1 following the steps of the algorithm described above:

I will substitute in the x value for 2 and the y value as  8 . So my equation

would now be  8 3(2)  14 . First I multiply 3 by 2 and that equals 6 and

I subtract 14;  8 6  14 ; and that is  8 , and  8 8 . That’s a check so I

would say, yes, the point (2,8) is on the line y 3x  14 .

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

questioning the accuracy of her graphing method.

182
 Figure 8. Example 1 of Student Answer to MDT Problem 1

Another student manipulated the equation y 3x  14 and correctly solved for x

(see Figure 9). From his representation of the equation, he could not decide

whether the point (2,8) is situated on the line.

Figure 9. Example 2 of Student Answer to MDT Problem 1

183
 None of the schemes tried by the third student were successful (see Figure 10).

Figure 10. Example 3 of Student Answer to MDT Problem 1

As can be seen in Figure 10, the student identified (2,8) as coordinates

( x0 , y 0 ) and replaced the variables x, y with those values in y 3x  14 , but failed

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

point (2,8) . Her description follows:

184
 I draw a graph first. I put in the numbers y 3x  14 ;  8 3(2)  14 ;

 8 6  14 …but I don’t know what I’m doing. So I set up a table. Then,

you plug all the numbers in and you draw a line, this is going up. So

(2,8) it’s not there.

4.5.2.1.2 Student Difficulties with the Cartesian Connection in MDT

Problem 2. All interviewed students from Group 0 (N=28), with the exception of

two, successfully solved MDT problem 2 using the following algorithm:

Step1. Identify values for x; and

Step 2. Substitute the x-values in y 2 x  3 , and calculate the y-values.

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

the equation and got the y values.”

Figure 11. Example 1 of Student Answer to MDT Problem 2

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.

The absence of given x-values created an insurmountable obstacle for one

student who failed to solve MDT problem 2:

“Because it didn't have any numbers, patterns to start off with, then I

wasn’t able to really solve it.”

Overcoming the obstacle would imply, albeit at the operational level,

mathematical knowledge about the existence of an infinite number of points

( x0 , y 0 ) with x 0 and y 0 real numbers, satisfying the equation of the line

y 2 x  3 .

Students who successfully identified x-values, could still encounter difficulty

calculating corresponding y-values. This would suggest knowledge that for a

given x 0 there exists a unique y-value that satisfies the equation of the line

y 2 x  3 . When such knowledge exists, the observed scheme of action would

be substituting each of the x-values into y 2 x  3 and calculating the

corresponding y-values. However, in the absence of such knowledge, one student

used the scheme of calculating the y-intercept of a line given the point (x, 0) , as

shown in Figure 12. He gave the values -1, 0, 1, 2, 3 to x, considered y 0 , and

calculated the values of the y-intercept b from y 2 x  b :

I just put  1, 0, 1, 2, 3, and to find y just put it in, put the x value in the

equation up here so y equals negative 2 times negative 1 plus b, and that’s

positive 2 so the y is negative 2 and then y equals negative two times 0

186
 plus b and 0 and … then do the same thing for 1…Negative 2 times 1 plus

b, negative 2, I add 2 here, would be positive 2 for y-intercept.

Figure 12. Example 2 of Student Answer to MDT Problem 2

Following is a summary of the results from interviews about difficulties with

the Cartesian connection:

1. All students described their schemes of action without referring to the

Cartesian connection as a formal mathematical theorem.

2. Students who missed the Cartesian connection from point to line in MDT

problem 1, failed to identify the ordered pair (2,8) as ( x0 , y 0 ) in

187
 y 3x  14 or to check whether the statement  8 3(2)  14 is true. They

tried (unsuccessfully) schemes of action involving the construction of

graphs or symbolic manipulations of the equation of the line.

3. Students who missed the Cartesian connection from line to points,

illustrated in MDT problem 2, failed to identify x-values, or after replacing

x with numerical values in y 2 x  3 , had difficulty calculating

corresponding y-values.

4.5.2.2 Student Difficulties with the Concept of Slope in a Geometric-

Analytical Context. Because both MDT problems 4 and 7b place slope in

geometric–analytical contexts, the analysis of student difficulties focused on

students’ concepts of slope in a geometric-analytical context: given two points

( 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

geometric perspective of slope (slope is an attribute of the line) used “slope

triangles,” counted the rise and the run, and calculated slope as “rise over run.”

Students with an analytical perspective of slope (slope is an attribute of the linear

y 2  y1
function) calculated the slope using the formula m .
x2  x1

4.5.2.2.1 Student Difficulties with Slope in MDT Problem 4. The algorithm

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

slope for MDT problem 4:

Student 1 (geometric perspective):

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,

5, 6, 7, 8, 9 and the run is 1, 2, 3, I believe. So it would be 9 over 3

because rise over run, or reduced, would be a slope of positive 3.”

Student 2 (analytical perspective):

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

in x by calculating the differences y 2  y1 and x2  x1 , and their ratio.

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

triangles, sometimes referred to as “slope triangles” or just line segments, and

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

run, but he failed to make the ratio “rise over run”:

“I can’t remember if I’m supposed to add 9 plus 3 or multiply or divide.”

The student identified the slope of the line as the ordered pair (3, 9) since he could

not make the ratio “rise over run.”

Figure 13. Example 1 of Student Answer to MDT Problem 4

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

(x-intercept, 0). The student associated the x-coordinate  3 of the endpoint

(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:

“I made a line up to the negative 3 and then counted over 3 from 0

(indicating the vertical line segment he drew as slope).”

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 .

Figure 14. Example 2 of Student Answer to MDT Problem 4

191
 In conclusion, the geometric perspective of slope prevailed in both schemes of

action of the interviewed students who encountered difficulties with slope.

4.5.2.2.2 Student Difficulties with Slope in MDT Problem 7b. The importance

of holding an analytical perspective of slope versus a geometric perspective of

slope is further illustrated by students’ difficulties with identifying the slope of a

line graphed in a non-homogeneous system of coordinates as in MDT problem 7b.

Nine of the 28 interviewed students from Group 0 held a geometric perspective of

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):

“I go up 4 and over 3… slope is four thirds.” [7b]

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

car for R in MDT problem 7c, indicating no difficulty in reading or interpreting

the graph. He no doubt holds a geometric perspective of slope, appropriate for

193
homogenous systems of coordinates, but not for non-homogeneous systems of

coordinates where an analytical perspective would have been useful.

The following are results from interviews about student difficulties with slope:

1. The geometric perspective of slope prevailed in all schemes of action for

the interviewed students who encountered difficulties with slope in MDT

problems 4 and 7b.

2. For MDT problem 4, students attempted to construct “slope triangles,” and

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

of the function for rise ( y 2 120 , y1 0 ) when calculating y 2  y1 , and

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

minor numerical errors.

In summary, results from students’ written comments and interviews showed

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.

Students had difficulty measuring in a homogeneous system of coordinates and

194
forming a ratio for slope. Also, students had difficulty identifying the unit of

measurement in a non-homogeneous system of coordinates. From the interviews,

only those students with a geometric perspective of slope had difficulty figuring

out the slope of a graphed line.

4.6 Research Question 5

Research Question 5: Do teachers’ speculations about the nature of students’

difficulties correspond to actual student difficulties as identified from student

MDT performance and interviews?

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-

rater agreement on coding of RQ written responses was computed for teacher

response, and was high, k = .89 (p <.05), 95% CI (.74, 1).

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

compared with students’ assessments of problem difficulty. Teachers’

speculations about student difficulties with the Cartesian connection (MDT

problems 1 and 2) and slope (MDT problems 4 and 7b), along with student

difficulties identified from their performance on MDT (N=1561) and in

interviews (N=40), are presented.

4.6.1 Teachers’ Assessment of the Nature of Problem Difficulty. Table 50

provides examples of teachers’ written comments on the Ranking Questionnaire

for Teachers. The number in brackets identifies the MDT problem.

Table 50

Teacher Assessment of Problem Difficulty

Category Examples Number of Teachers

H (Hard Problem) 0

L (Never “Haven’t seen the material.” [7] 10

Learned/Taught)
“If the students plug the #s into the 11
M (Mathematical equation, they might know what to do
Difficulty) 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.” [1]

R (Not Remembering) “Students forget how to find slope.” [7] 1

T (Thinking Time) “Lots of reading.”[ 7] 3

U (Not Understanding) “Most difficult: Mostly because of the 1

wording. My students can solve it but
the wording is not clear.” [2]

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

familiarity with the problem’s content or context (never seen, never

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

problem (e.g., MDT problem 1 is an extended–response type of problem requiring

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

used in a problem (MDT problem 2 requiring completion of a table of values).

Although 11 teachers (42%) referred to the mathematical nature of problem’s

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.

4.6.2 Teachers’ Comments on Student Difficulties with the Cartesian

Connection.

4.6.2.1 MDT Problem 1: Cartesian Connection from Point to Line. On the

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

mathematical difficulty students might encounter:

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.

In other words, if a student attempted to use the Cartesian connection and

replace ( x0 , y 0 ) in the equation y 3x  14 with (2,8) then difficulty may arise at

checking whether  8 equals 3(2)  14 . If a student used the graphing strategy to

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

(2,8) as ( x0 , y 0 ) in y 3x  14 , checking whether the statement  8 3(2)  14 is

true, as well as deciding whether the point is on the graph after they plotted the

point and graphed the line.

All interviewed teachers (N=20), except for the teacher mentioned above, were

prompted by the researcher to comment on students’ knowledge of the Cartesian

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

“theorem in action” (Vergnaud, 2009) and a solution algorithm:

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

equal on both sides.

Fourteen (70%) teachers described the Cartesian connection from point to line

only procedurally, that is, as an algorithm for solving the problem:

We can test it to see if it’s a solution, by replacing the x and the y

coordinates and if we simplify and it makes a true sentence it is a solution

to the equation of two variables because it makes the sentence true, right?

Thirteen (65%) teachers identified the Cartesian connection as an obstacle for

their students who could not solve MDT problem1, or tried to solve it using a

graphing approach:

Two is x and  8 is y; putting those coordinates in for x and y and making

sure that it’s a true statement. If they don’t have, can make that connection

with the point, they tried to graph it.

Three (15%) teachers identified student difficulties with MDT problem 1 as

switching the x-value with the y-value in (2,8) when substituting those values

into the equation of the line, y 3x  14 :

They have a problem of substituting, basically, and knowing an ordered

pair. Knowing that x comes first, y comes second and knowing that they

are substituting in these values into this equation.

Although most interviewed teachers, when prompted by the researcher,

identified lack of understanding of the Cartesian connection as the major obstacle

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

student difficulties could arise at each of the steps of the algorithm.

4.6.2.2 MDT Problem 2: Cartesian Connection from Line to Points. On the

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

wording, it is not clear.”

Interviews revealed that students had difficulty identifying x-values to enter

into the table, as well as substituting x values into the equation y 2 x  3 in

order to calculate corresponding values of y.

When prompted by the researcher, six (30%) of the interviewed teachers

identified the mathematical difficulty as the absence of any given x-values to help

students get started with the completion of the table:

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

less difficult for them.

Two (10%) of the interviewed teachers referred to the novelty of the format of

the problem as a potential source of difficulty. According to them, the x-values

were always given to their students, and the format of the table was different from

the one in MDT problem 2.

200
 Teacher 1:

I always tell my students to create a table where we’re going to be using

the values of  2 to  2 . I say that all the time. I let them know what

points I want them to use.

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.

None of the teachers commented on the difficulty calculating corresponding y-

values once x-values are identified, although this type of difficulty arose for two

of the interviewed students. Two (10%) of the interviewed teachers, when

prompted by the researcher, commented about the necessity for students to

conceive the existence of an infinite number of points ( x0 , y 0 ) composing the line

y 2 x  3 , and therefore satisfying its equation:

They have to understand that this equation describes a relationship of all

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

line to points illustrated in problem 2 as an algorithm to solve the problem.

However, they failed to point out difficulties their students may encounter at each

step of the algorithm.

4.6.3 Teachers’ Comments on Student Difficulties with Slope in a Geometric-

Analytical Context

4.6.3.1 MDT Problem 4: Slope in Homogeneous Systems of Coordinates. On

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

points [were] labeled.”

Interviews revealed that only students with a geometric perspective of slope

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.”

When prompted by the researcher, 18 (90%) of the interviewed teachers

described their students’ knowledge of slope in a geometric-analytical context:

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

hopefully is a little bit easier for them.

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

interviewed teachers commented on this type of difficulty:

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

(0,1) because it’s close enough to  1 .

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

difficulty. Only one (5%) of the interviewed teachers identified choosing or

replacing values in the formula for slope as a mathematical difficulty in MDT

problem 4 for students:

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

their ( x1 , y1 ) and their ( x2 , y 2 ) and they’ll…for their slope, they’ll do

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

top and they do my y1  x1 , and so they’ll get the formula confused,

they’ll… So they mistake which one is x and which one is y.

y 2  y1
All interviewed students who employed the formula to compute the
x 2  x1

slope, used it correctly.

In general, teachers underestimated the difficulty of identifying the slope in

problem 4. When prompted by the researcher to comment on the nature of their

students’ difficulties with slope, teachers did not mention any of the difficulties

associated with the construction of a “slope triangle.”

4.6.3.2 MDT Problem 7b: Slope in Non-Homogeneous Systems of

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

teachers comment on the nature of the difficulty with problem 7b as calculating

the slope of 4/3:

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,

commented on the necessity of explaining how to measure in a non-homogeneous

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

going up 1 or how many am I really going up? So it’s a matter of paying

attention to what the intervals are on the graph.

Although none of the interviewed teachers identified measuring in non-

homogenous systems of coordinates as the main mathematical difficulty for their

students, when prompted by the researcher, they could describe the nature of the

difficulty that led students to answer 4/3 instead of 120/3 to problem7b.

Analyses of teachers’ written comments and interviews about the nature of

their students’ difficulties, showed that:

1. Speculations about the nature of student difficulties did not correspond to

actual student difficulties, as identified from student performance on the

MDT and student Interviews.

2. Only 11 (42%) of the 26 teachers who completed RQT described a

mathematical difficulty. Only six of those 11 teachers described a

mathematical difficulty that was actually encountered by students, namely

205
 missing the Cartesian connection (MDT problem 1) and considering slope

as rate of change (MDT problem 7b).

3. During the interviews, only four (20%) teachers described a mathematical

difficulty that corresponded to actual student mathematical difficulties.

When prompted by the researcher to comment on a specific type of

difficulty, the percentage of interviewed teachers that described an actual

mathematical difficulty encountered by students improved. With

prompting, 13 (65%) teachers described missing the Cartesian connection

in MDT problem 1, six (30%) teachers described the difficulty of finding

x-values for the table in MDT problem 2, four (20%) teachers referred to

measurement difficulties when estimating the x-intercept in MDT problem

4, and four (20%) teachers described the difficulty measuring the run or

the change in y in MDT problem 7b.

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

8 through Grade 10. Student performance on the MDT increased with

more advanced courses, mainly due to Group 1 performance. Group 1

students outperformed Group 0 students.

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

difficult for their students.

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

geometric aspect of slope prevailed in all schemes of action for the

interviewed students who encountered difficulties with those problems.

For MDT problem 4, students attempted to construct “slope triangles,” and

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,

except for minor numerical errors.

iv. Teachers’ comments about the nature of their students’ difficulties did not

agree with student difficulties as identified from student performance on

the MDT and Interviews. The agreement remained low, even when

teachers were prompted by the researcher to comment on specific student

difficulties (e.g., students’ difficulty with slope in MDT problem 7b).

In Chapter 5, results of this study are discussed in relation to other research on

students’ understanding of aspects of linearity and teachers’ knowledge of student

difficulties. Conclusions that can be drawn from the results of the data analyses,

207
along with limitations for generalizing results and recommendations for further

research, are presented.

208
 Chapter 5

DISCUSSIONS AND CONCLUSIONS

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

many researchers with various populations of students (Ben-Chaim et al., 1998;

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

educators in implementing curricular and instructional changes, student

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

investigated their assessments of their own difficulties and their mathematics

teachers’ understanding of the nature of their students’ difficulties.

In this chapter, results of data analyses are summarized, and discussions of

those results with respect to previous research findings are presented by research

question (Section 5.2). Limitations of this study (Section 5.3) and

recommendations for future research (Section 5.4) conclude the chapter.

5.2 Results and Discussion

The following questions were addressed in this study:

1. Does student performance on the Mini-Diagnostic Test (MDT) vary by:

a. Grade level?

b. Current mathematics course?

c. Problem?

2. Does student ranking of problems by difficulty, after completing the MDT,

correspond to their performance on the MDT?

3. Does teacher ranking of problems by difficulty, after completing the MDT,

correspond to student performance on the MDT?

4. What is the nature of students’ difficulties?

210
5. Do teachers’ speculations about the nature of students’ difficulties correspond

to actual student difficulties as identified from student performance on the MDT

and interviews?

5.2.1 Research Question 1 Results and Discussion.

Research Question 1: Does student performance on the Mini-Diagnostic Test

(MDT) vary by:

a. Grade level?

b. Current mathematics course?

c. Problem?

To answer Research Question 1, the Information Questionnaire for Students

(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.

Results for Question 1 revealed little variation in student performance on the

MDT across grade levels. Regardless of grade level, on average, students

correctly solved five of the nine MDT problems. However, analyses of the MDT

score distributions revealed greater numbers of higher scores with increasing

grade level. The presence of higher scores at higher grade levels may be due to

increased exposure to the concepts by virtue of completion of more advanced

mathematics courses.

With respect to variations in performance as related to current mathematics

course, results showed that Group 1 students, enrolled in mathematics courses

211
above their grade level in the mathematical instructional sequence, outperformed

Group 0 students, enrolled in courses at or below their grade level in the

sequence. Performance gains on the MDT across courses were due mainly to

Group 1 students’ score contributions. These results might be explained by the

relationship between the type of learners participating in this study and their

current mathematics courses. Group 0 students, identified as low-achieving or

average-achieving and on remediation or median tracks, were enrolled in lower

division mathematics courses (e.g., Pre-Algebra, Algebra I) . By contrast, Group 1

students, identified as high-achieving and on the college-track, were enrolled in

more advanced courses, such as Geometry or Algebra II. These performance

results, as well as their justifications, are supported by Gamoran and Hannigan’s

research (2000) and by the American Diploma Project (2010). During a

longitudinal study, Gamoran and Hannigan (2000) adminstered achievement tests

to students in Grades 8 and 10, and found that those who completed Algebra I

earlier in the instructional sequence outperformed students who took Algebra I

later or did not take Algebra I . The American Diploma Project (ADP) reported

that Algebra I student performance on end-of-course examinations decreased by

grade level. That is, students who completed Algebra I in earlier grades

outperformed students who completed Algebra I later in their education. Although

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

studies, assessments were more comprehensive (i.e., included topics in addition to

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

performance on only aspects of linearity and linear functions, was administered

during the instructional period and not at the end of a course, and was not

restricted to Algebra I students.

With regard to variations in performance based on problem type, students had

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

Group 1). Student performance on MDT problems 1, 2, 4, and 7b increased with

enrollment in more advanced mathematics courses, like Geometry and Algebra II.

One possible explanation for higher MDT scores for Geometry and Algebra II

students, may be due to their completion of more mathematics courses at more

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

curricular and instructional practices influenced student performance, and 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

increased with more advanced mathematics courses, mainly due to Group 1

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,

confirmed the persisting difficulty of mastering the concept of slope in geometric-

analytical contexts (Greenes et al., 2007a; Knuth, 2000; Orton, 1983; Stump,

2001a; Teuscher & Reys, 2010; Zaslavsky et al., 2002).

5.2.2 Research Question 2 Results and Discussion.

Research Question 2: Does student ranking of problems by difficulty, after

completing the MDT, correspond to their performance on the MDT?

To answer Research Question 2, the Mini-Diagnostic Test (MDT) and

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

ranking of relative problem difficulty. Frequencies of student rankings of

problem difficulty from the RQS were used to obtain student subjective ranking

of relative problem difficulty. The empirical ranking based on student

214
performance on the MDT and the subjective ranking based on student perception

of difficulty were compared.

Results for Question 2 showed no statistically significant agreement for

Group 0 students and moderate agreement for Group 1 students between their

empirical rankings of difficulty based on MDT performance and their rankings of

MDT problems by difficulty. A relative comparison of rankings showed that all

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

not identified as most difficult.

There is a paucity of studies that examined students’ perceptions of problem

difficulties as compared to their actual performance. A study conducted with

Australian undergraduate students enrolled in Biochemistry and Physics courses

(Lingard et al., 2009) found a similar result of moderate agreement between the

ranking based on student performance and the ranking based on student

perception of difficulty. Another study in Hong Kong found that Grade 9

students’ perceptions of problem difficulty correlated highly with their

performance on an algebra test on logarithms (Lee & Heyworth, 1996). These

findings suggest that student cultural background, curricular experiences, type of

215
learner, and the problem content and context may play important roles not only in

student performance, but also in students’ perceptions of problem difficulty.

5.2.3 Research Question 3 Results and Discussion.

Research Question 3: Does teacher ranking of problems by difficulty, after

completing the MDT, correspond to student performance on the MDT?

To answer Research Question 3, the Mini-Diagnostic Test (MDT) was

administered to all participating students and teachers, and the Ranking

Questionnaire for Teachers (RQT) was administered to all participating teachers.

Student performance, given by the MDT mean score for each of the MDT

problems, was used to obtain the ranking of problem difficulty. Teacher

assessment of problem difficulty was given by the frequencies of teacher rankings

of problem difficulty from the RQT. Of note, teachers’ rankings of problem

difficulty referred to problem difficulty as perceived by their students. The

ranking based on student performance and the ranking based on teacher

assessment of problem difficulty were compared.

The comparison between teachers’ ranking of problem difficulty and their

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

being the most difficult for their students.

Results showed no significant agreement between student performance on the

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,

and methods of administration differed, comparisons are difficult to make.

5.2.4 Research Question 4 Results and Discussion.

Research Question 4: What is the nature of students’ difficulties?

To answer Research Question 4, the Ranking Questionnaire for Students

(RQS) was administered to all participating students, and interviews were

conducted with 40 students (see Chapter 3). Students’ written comments from the

RQS about the nature of their difficulties, and students’ interviews were analyzed.

The focus of analysis was on the mathematical nature of difficulties encountered

by the students (N=472) who identified problems referring to slope or the

Cartesian connection as being difficult for them.

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

a problem (e.g., amount of time, memory). The positive correlation between

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

number of computational operations necessary to solve a problem.

Only 20% of the 472 students in the present study made comments about their

specific mathematical difficulty. Analyses of the nature of students’ written

comments and interviews revealed that students had difficulty with the Cartesian

connection and slope. Difficulties with the Cartesian connection included

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).

Difficulty recognizing points ( x0 , y 0 ) as solutions of the linear equation

y mx  b or ax  by c is confirmed by studies conducted by Greenes et al.

(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,

change in y); and 3) identifying the unit of measure in a non-homogeneous system

of coordinates (e.g., counting the “tick marks” on the Y-axis and disregarding the

scale). Difficulty connecting x2  x1 (change in x or run) or y 2  y1 (change in y

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

identifying the slope of a line graphed in a system of Cartesian coordinates is a

more difficult task than identifying the rate of change of a linear function when

two instances of the varying quantities, ( x1 , y1 ) and ( x2 , y 2 ) , are given. One

explanation may be that the graphing context for slope implies knowledge of

mathematical conventions (e.g., if the oriented segment representing the rise is

toward the positive sense of the Y-axis, its magnitude has a positive sign),

considered by Schoenfeld et al. (1993) as part of the Cartesian connection. Thus,

difficulty in identifying the slope of a graphed line lies with the connection with

the graphic representation. This explanation is supported in the present study by

findings from students’ Interviews; only those students with a geometric

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

those students with a geometric perspective of slope (Kaput & Maxwell-West,

1994; Lamon, 1993; Lobato & Siebert, 2002). Indeed, calculating the slope of a

graphed line as “rise over run,” implies knowledge of proportionality in the

geometric context of similar triangles, or at minimum, coordination while

identifying the rise and the run from a “slope triangle.” By contrast, students with

an analytical perspective of slope, who calculated slope as “change in y over

219
 y 2  y1
change in x,” or used the formula m , had less difficulty since the
x2  x1

connection with the graphic representation was reduced to identifying the

coordinates of the points, ( x1 , y1 ) and ( x2 , y 2 ) . Nevertheless, studies conducted

by Lobato and Siebert (2002), Orton (1983), and Stump (2001a) suggest that the

concept of slope or rate of change remains a concept difficult to master for

students, regardless of the context of the problem (e.g., graphic representation,

functional situation) or the perspective held (e.g., geometric or analytical). For

example, in those studies and in the present study, students demonstrated

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

graphed in a non-homogeneous system of coordinates (MDT problem 7b). This

difficulty is confirmed by studies conducted by Zaslavsky et al. (2002). Also of

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

understanding the “wording.” Notwithstanding that the content of the problems

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

graphic representations, and with connections between symbolic and graphic

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

presented and developed in U.S. mathematics curricula. Student difficulties may

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

school districts participating in this study, slope is introduced to Pre-Algebra

students in a physical context followed by a geometric-analytical context:

In skiing, the term slope refers to a slanted mountainside.[…] In math,

slope defines the “slant” of a line.[…] For a line on the coordinate plane,

vertical change change in y

slope is the following ratio:
horizontal change change in x

(Bennett et al., 2004, p. 545).

In an Algebra I textbook used by the study’s participants the physical context

is followed by the geometric context of slope:

Some roofs are steeper than others. In mathematics, a number called slope

is a measure of the steepness of a line. The slope of a line is the ratio of

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

perspective of slope as rate of change, the latter being introduced in advanced

mathematics classes, like Precalculus and Calculus. The results of this study

221
showed that only students with a geometric perspective of slope had difficulties

determining the slope of a graphed line.

One possible explanation for student difficulties with respect to mathematical

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

explicitly in another Algebra I textbooks used by this study’s participants. In

Larson et al. (2004), students are reminded in Chapter 1 about the procedure to

graph a function, and an example is given:

When you are given the rule for a function, you can prepare to graph the

function by making a table showing numbers in the domain and their

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).

In this presentation, there is no reference to the units of measurement for the

axes (e.g., scale for the Y-axis). Difficulty with identifying the scale of the Y-axis

in non-homogeneous system of coordinates was evident in MDT problem 7b.

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

tick marks was 30 units.

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.

5.2.5 Research Question 5 Results and Discussion.

Research Question 5: Do teachers’ speculations about the nature of students’

difficulties correspond to actual student difficulties as identified from student

MDT performance and interviews?

To answer Research Question 5, the Mini-Diagnostic Test (MDT) was

administered to all participating students, the Ranking Questionnaire for Teachers

(RQT) were administered to all participating teachers, and interviews were

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

Heyworth’s (1996, 2000) findings that teachers attend to problem’s surface

features, rather than to problem’s solution requirements to determine problem

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.

About 42% of the 26 teachers referred to their students’ difficulty as being

mathematical in nature. Student performance on the MDT and student interviews

showed that students had difficulty with MDT problems 1 and 2 referring to the

Cartesian connection, and MDT problems 4 and 7b referring to slope. Teachers’

speculations about the nature of their students’ difficulties did not correspond to

actual student difficulties, as identified from student performance on the MDT

and student interviews. Only six teachers (23%) mentioned an actual student

difficulty in their written comments, missing the Cartesian connection in MDT

problem 1, and connecting slope in MDT problem 7b with the rate of change in

MDT problem 7c. Teachers’ descriptions of student difficulties were mainly at

the procedural level (e.g., difficulty substituting (2,8) for ( x0 , y 0 ) in y 3x  14 ).

Only four (20%) of the interviewed teachers mentioned actual student difficulties,

without being prompted by the researcher. When specifically prompted by the

researcher, the percentage of teachers who could describe an actual mathematical

difficulty increased overall by about 14%. Mathematical difficulties described by

the interviewed teachers were the following: a) difficulty checking whether

(2,8) is a solution to the linear equation y 3x  14 (MDT problem 1);

224
b) difficulty finding x-values to complete a table of values for the linear function

y 2 x  3 (MDT problem 2); c) difficulty estimating the x-intercept of a line

graphed in a homogeneous system of coordinates (MDT problem 4); and

d) difficulty measuring the run or the change in y in a non-homogenous system of

coordinates (MDT problem 7b).

Previous research studies reported on teachers’ weak pedagogical knowledge

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

(1995) studies mentioned difficulty with the meaning of slope, understanding

slope as a measure of steepness, difficulty graphing lines, remembering the slope

formula and applying it, and difficulty performing numerical calculations. When

prompted by researchers, teachers did not mention student difficulties connecting

among various representations (Cunningham, 2005; Davis, 2007; Even & Tirosh,

1995; Stump, 1997, 2001b).

Teachers’ weak pedagogical knowledge was mirrored by their weak

mathematical knowledge of slope and connections among representations (Stump,

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

graphed in a non-homogeneous system of coordinates, recognizing slope as a

parameter, and as a constant rate of change of two covarying quantities, and

connecting between various representations. In the present study, all participating

teachers correctly solved the MDT problems. In others studies (Even & Tirosh,

225
1995; Zhou, 2009) although involving small numbers of teachers, teachers with

weak mathematical knowledge, incorrectly solved problems, but could identify

some student difficulties and predict student performance. Those studies (Even &

Tirosh, 1995; Zhou, 2009) together with the present study, point out that the

relationship between mathematical and pedagogical knowledge is complex. In

other words, being able to correctly solve a mathematical problem is not

associated with being able to predict student difficulties with the problem.

In summary, with respect to students' and teachers’ assessments of the nature

of problem difficulty, students from Group 1 were better able to describe the

mathematical nature of their difficulties than were their teachers. Teachers’

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

format of the problem (e.g., extended response, requiring explanations). Without

prompting, teachers did not attend to the mathematical nature of actual student

difficulties. When prompted by the researcher during interviews, teachers

commented on student difficulties regarding the correct application of an

algorithm, a formula, or the reading of a graph, and not on conceptual difficulties

like conceiving of slope as a measure of steepness in a geometric context.

Teachers’ weak understanding of student difficulties with important topics like

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

after being taught for a long period of time (8-10 weeks).

226
5.3 Limitations of the Study

Limitations of the study focus on the instruments, the study sample, the

administration of instruments, and the methods of analysis.

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,

according to previous studies, appeared to be difficult for students in Grades 8-10;

and 2) Students must be able to complete the MDT, IQS, and RQS in one class

session (i.e., 50 minutes or less).

As a consequence, while providing information on aspects of linearity and

linear functions, the MDT is not comprehensive. For example, only two problems

focus on slope in geometric–analytical contexts (MDT problems 4 and 7b), two

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

students enrolled in Algebra II (N=77) was 8.4, on a scale of 0 to 9, indicating

that the problems were too easy for those students, thus creating a ceiling effect

(McBee, 2010) and, as a consequence, making it difficult to detect true

differences in student knowledge.

The Population. Secondary-level mathematics teachers and their students

from two predominately middle-class school districts in the Southwest of U.S.

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

Education Statistics, 2010).

With respect to current mathematics course, about two-thirds of the

participating students were enrolled in courses at their grade level in the

mathematics course sequence (Grade 8 - Pre-Algebra, Grade 9 - Algebra I, Grade

10 - Geometry), and one-third of the participating students were enrolled in

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

have to be considered in light of average-achieving and high-achieving students.

For the reasons stated above, the findings of this study cannot be generalized

beyond the particular population, and the educational context (cultural settings

and school curricula).

Administration of Instruments. The MDT and interviews were administered

in the middle of the academic year. As a consequence, the timing of

administration was not synchronized with the instructional sequences. For

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

agreed to observe test-like conditions, no supervision of teachers and the test

administration took place. Whether or not teachers provided some assistance (e.g.,

defining a term or clarifying a relationship) is not known.

Finally, interviews were conducted solely by the researcher. As such, probing

questions were subject to the researcher’s interest. If, instead, the interviews were

conducted by a team of researchers, even under the assumption of agreement

upon the perspective of looking at the nature of student knowledge and

difficulties, and the theoretical framework used for analysis, results might have

varied.

Quantitative Methods of Analyses. Data were analyzed using descriptive

statistics, including cluster analysis and factor analysis. A regression framework

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

independence assumption necessary for regression. The violation of the

assumption of independence may cast doubts upon any statistical significance

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.,

student socio-economic status, previous scores on standardized tests).

229
5.4 Recommendations for Future Research

Based on the limitations described above, recommendations for future research

address the assessment instruments, the population, the administration of the

MDT and interviews, and the research design.

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

as a measure of steepness in a physical context, slope as rate of change in a

covariational context, slope as a parameter in a linear function), would provide a

more comprehensive picture of student understanding. Furthermore, including

problems of increased complexity, like those recommended by Knuth (2000) and

Moschovich et al. (1993), would alleviate the ceiling effects observed in the

present study. To avoid ceiling and flooring effects (McBee, 2010), different tests

could be administered to different groups of students based on their performance

levels, i.e., low-achieving (students enrolled in courses below their grade level),

average-achieving (students enrolled in courses corresponding to their grade

level), and high-achieving (students enrolled in courses above their grade level).

With respect to the Ranking Questionnaire, participants could be asked to not

only rank problems by difficulty, but also, for each problem, identify factors that

contributed to the difficulty (e.g., terminology, number of solution steps).

The number of questions in the interviews could be increased to better assess

student and teacher familiarity with the problems and problem types. For

example, for students, one might ask, “Which of these problems have you seen

before? “and “Which problems are new to you?”

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

before the end of the course?”

The Population. Using the instruments and procedures but with a greater

cross-section of the general population (e.g., greater variation in socio-economic

status) would enhance interpretation and generalization of the findings of the

present study.

Administration of Instruments. To ensure that test-like conditions are

observed by all participants in the study, supervision by a team of researchers is

highly recommended during administration of the MDT and Ranking

Questionnaires. To minimize bias in the nature of the probing questions,

interviews could be conducted by a team of researchers.

Quantitative Methods of Analyses and Research Design. In the present

study, data analyses employed descriptive statistics. Future studies, using

Hierarchical Linear Modeling (HLM) will enable investigations of the effects of

cross-level data (e.g., curriculum, teaching style) on individual outcomes (e.g.,

student performance on the MDT). For example, it would be interesting to

compare the effects on student performance on the MDT of introducing slope as a

rate of change in functional contexts (e.g., constant speed, constant water flow),

versus introducing slope as a measure of steepness in a geometric context.

The present study provided a snapshot of student knowledge, and student and

teacher perception of student difficulties with linearity and linear functions.

Employing a longitudinal study could provide a clear picture of changes in

231
student knowledge, as well as changes in their teachers’ perception of their

students’ difficulties, over time. This would be valuable information for

curriculum developers in the design of mathematics instructional materials. As

well, it would be helpful to mathematics educators in the design and conduct of

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

American Diploma Project. (2004). Ready or Not: Creating a High School

Diploma ThatCounts. Washington, DC: Achieve. Retrieved March 25,
2011, from http://www.achieve.org/files/ADPreport_7.pdf

American Diploma Project. (2010). ADP End-of-Course Exams: 2010 Annual

Report. Washington, DC: Achieve. Retrieved March 25, 2011, from
http://www.achieve.org/files/AchieveADPEnd-of
CourseExams2010AnnualReport.pdf

Arcavi, A. (2003). The Role of Visual Representations in the Learning of

Mathematics. Educational Studies in Mathematics , 52 (3), 215-241.

Arizona Community Foundation . (2008). Educating Arizona: Assessing Our

Education System (Birth–Grade 12). Phoenix, AZ: Arizona Community
Foundation & Ellis Center for Educational Excellence. Retrieved March 25,
2011, from http://www.azftf.gov/WhatWeDo/Documents/Resource%205-
B-ACF%20-%20Full%20Report%20-%20Educating%20Arizona.pdf

Arizona Department of Education. (2008, June 24). 2008 Mathematics Standard

Articulated by Grade Level. Retrieved June 22, 2010, from Arizona
Department of Education:
http://www.ade.state.az.us/standards/math/Articulated08/

Ben-Chaim, D., James, F. T., Fitzgerald, W. M., & Benedetto, C. (1998).

Proportional Reasoning among 7th Grade Students with Different
Curricular Experiences. Educational Studies in Mathematics , 36 (3), 247–
273.

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.

Bozick, R., & Lauff, E. (2008). Education Longitudinal Study of 2002

(ELS:2002): A First Look at the Initial Postsecondary Experiences of the
High School Sophomore Class of 2002 (NCES 2008308). Washington, DC:
National Center for Education Statistics. Retrieved March 25, 2011, from
http://nces.ed.gov/pubs2008/2008308.pdf

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

Business-Higher Education Forum. (2005). Securing America’s Leadership in

Science,Technology, Engineering, & Mathematics. Retrieved March 25,
2011, from
http://californiastem.org/documents/BHEF%20Securing%20America%27s
%20Potential.pdf

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

Chazan, D. (2008). The Shifting Landscape of School Algebra in the United

Greenes, C. & R. Rubenstein (Eds.), Algebra and Algebraic Thinking in
School Mathematics: Seventieth Yearbook (pp. 19-36). Reston, VA:
National Council of Teachers of Mathematics.

Cheng, D. S. (2010). Connecting proportionality and slope: Middle school

students' reasoning about steepness (Doctoral dissertation). Available from
ProQuest Dissertations & Theses database. (UMI No. 3405966)

Clement, J. (2000). Analysis of Clinical Interviews: Foundations and Model

Viability In A. E.Kelly & R. A. Lesh (Eds.), Handbook of Research Design
in Mathematics and Science Education. Mahwah, NJ: Lawrence Erlbaum
Associates.

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.

Committee for Economic Development. (2003). Learning for the Future:

Changing the Culture of Math and Science Education To Ensure a
Competitive Workforce: A Statement by the Research and Policy Committee
of the Committee for Economic Development. Retrieved March 25, 2011,
from
http://www.ced.org/images/library/reports/education/report_scientists.pdf

Cooke, R. (1997). The History of Mathematics: A Brief Course. New York: John
Wiley & Sons.

Corry, L. (2004). Modern Algebra and the Rise of Mathematical Structures

(Second ed.). Birkhäuser Verlag.

Costello, A. B., & Osborne, J. (2005). Best practices in exploratory factor

analysis: four recommendations for getting the most from your analysis.
Practical Assessment Research & Evaluation , 10 (7), Available online:
http://pareonline.net/getvn.asp?v=10&n=7

Council of Chief State School Officers [CCSSO] & National Governors

Association [NGA]Center for Best Practices. (2010). The standards:
Mathematics. Retrieved March 25, 2011, from
http://www.corestandards.org/the-standards/mathematics

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

Cunningham, R. F. (2005). Algebra Teachers' Utilization of Problems Requiring

Transfer between Algebraic, Numeric, and Graphic Representations. School
Science and Mathematics , 105 (2), 73-81.
doi:10.1111/j.1949-8594.2005.tb18039.x

Daniel, W. W. (1990). Applied Nonparametric Statistics (Second ed.). Pacific

Grove, CA: Duxbury.

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

Davis, J. D. (2007). Real-World Contexts, Multiple Representations, Student-

Invented Terminology, and Y-Intercept. Mathematical Thinking and
Learning:An International Journal , 9 (4), 387-418.
doi: 10.1080/10986060701533839

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

Ding, M. (2007). Knowing Mathematics for Teaching: A Case Study of Teacher

Responses to Students' Errors and Difficulties in Teaching Equivalent
Fractions (Doctoral dissertation). Available from ProQuest Dissertations &
Theses database. (UMI No. 3281048)

Doerr, H. (2004). Teachers' Knowledge and The Teaching of Algebra. In K.

Stacey, H. Chick, & M. Kendal (Eds.), The Future of the Teaching and
Learrning of Algebra: The 12th ICMI Study (pp. 267-290). Boston, MA:
Kluwer Academic Publishers.

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

Ellis, A. B. (2007). Connections Between Generalizing and Justifying: Students'

Reasoning with Linear Relationships. Journal for Research in Mathematics
Education , 38 (3), 194-229.

Ellis, A. B. (2009). Patterns, Quantities, & Linear Functions. Mathematics

Teaching in the Middle School , 14 (8), 482-491.

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.

Flora, D. B., & Curran, P. J. (2004). An Empirical Evaluation of Alternative

Methods of Estimation for Confirmatory Factor Analysis With Ordinal
Data. Psychological Methods ,9 (4), 466–491.
doi: 10.1037/1082-989X.9.4.466

Freudenthal, H. (1983). Didactical Phenomenology of Mathematical Structures.

Dordrecht: D. Reidel Publishing.

Gamoran, A., & Hannigan, E. C. (2000). Algebra for Everyone? Benefits of

College-Preparatory Mathematics for Students with Diverse Abilities in
Early Secondary School. Educational Evaluation and Policy Analysis , 22
(3), 241-254. doi: 10.3102/01623737022003241

Gire, E., & Rebello, N. S. (2010). Investigating the Perceived Difficulty of

Introductory Physics Problems. Paper presented at Physics Education
Research Conference 2010, Portland, Oregon. Retrieved December 29,
2010, from
http://www.compadre.org/Repository/document/ServeFile.cfm?ID=10426&
DocID=1892

Goldin, G. (1999). A Scientific Perspective on Structured, Task-Based Interviews

in Mathematics Education Research. In A. E. Kelly & R. A. Lesh (Eds.),
Handbook of Research Design in Mathematics and Science Education (pp.
517-545).

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.

Hansson, Ö. (2006). Studying the Views of Preservice Teachers on the Concept of

Function (Un published doctoral dissertation). Luleå University of
Technology, Luleå, Sweden. Retrieved November 12, 2010 from
http://epubl.luth.se/1402-1544/2006/22/LTU-DT-0622-SE.pdf

Hauger, G. S. (1997). Growth of Knowledge of Rate in Four Precalculus

Students. Paper presented at the Annual Meeting of the American
Educational Research Association.Chicago, IL.

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.

Henson, R. K. (2001). Understanding internal consistency reliability estimates: A

conceptual primer on coefficient alpha. Measurement and Evaluation in
Counseling and Development , 34 (3), 177-189.

Hines, E. (2002). Developing the concept of linear function: one student’s

experiences with dynamic physical models. Journal of Mathematical
Behavior , 20, 337–361. doi:10.1016/S0732-3123(02)00074-3

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

Jerman, M. (1973). Problem Length as a Structural Variable in Verbal Arithmetic

Problems. Educational Studies in Mathematics , 5 (2), 109-123.

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.

Johnson, K. A. (2008). The Association Between Socioeconomic Status and High

School Mathematics Scores and Enrollment Rates in Virginia Public
Schools (Doctoral dissertation). Available from ProQuest Dissertations &
Theses database. (UMI No. 3308025)

Kalchman, M., & Koedinger, K. R. (2005). Teaching and learning functions. In S.

Donovan, & J. Bransford (Eds.), How Students Learn: Mathematics in the
Classroom (pp. 351-393). Washington, DC: National Academy of Sciences.

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.

Kaput, J. J., & Maxwell-West, M. (1994). Missing-value proportional reasoning

problems: Factors affecting informal reasoning patterns . In G. Harel, & J.
Confrey (Eds.), The development of multiplicative reasoning in the learning
of mathematics (pp. 235–287). Albany, NY: State University of New York
Press.

Katz, V. J. (1998). A History of Mathematics: An Introduction (2nd ed.). Reading,

MA: Addison Wesley Longman.

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

Katz, V. J. (2007). Stages in the History of Algebra with Implications for

Teaching. Educational Studies in Mathematics , 66 (2), 185–201.
doi: 10.1007/s10649-006-9023-7

Kieran, C. (1992). The Learning and Teaching of School Algebra. In D. A.

Grouws (Ed.), Handbook of Research on Mathematics Teaching and
Learning (pp. 390-419). Reston, VA: National Council of Teachers of
Mathematics.

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.

Kliebard, H. M. (2004). The Struggle for the American Curriculum, 1893-1958 .

New York: Routledge Falmer.

Kline, M. (1973 ). Why Johnny Can't Add: The Failure of the New Mathematics.
St. Martin's Press.

Knuth, E. J. (2000). Student Understanding of the Cartesian Connection: An

Exploratory Study. Journal for Research in Mathematics Education , 31(4),
500-508.

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.

Lamon, S. J. (1993). Ratio and proportion: connecting content and children’s

thinking. Journal for Research in Mathematics Education , 24 (1), 41–61.

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.

Lanier, S. M. (1999). Students' Understanding of Linear Modeling in A College

Mathematical Modeling Course (Doctoral dissertation). Available from
ProQuest Dissertations & Theses database. (UMI No. 9928955)

Larson, R., Boswell, L., Kanold, T. D., & Stiff, L. (2004). Algebra I: Concepts
and Skills. Evanston, IL: McDougal Littell.

Lee, F. L., & Heyworth, R. M. (1996). Teachers’ Estimation of Problem

Difficulty. Paper presented at the Thirteenth Annual Conference, Hong Kong
Educational Research Association. Retrieved October 25, 2010, from
http://www.fed.cuhk.edu.hk/fllee/Papers/ConferPa/TEAESTIM.html

Lee, F. L., & Heyworth, R. (2000). Problem Complexity: A Measure of Problem

Difficulty in Algebra by Using Computer. Education Journal , 28 (1), 85-
107.

Lewin, A. Y., & Couto, V. (2006). Next Generation Offshoring:The Globalization

of Innovation. 2006 Survey Report. (Booz Allen Hamilton and Offshoring
Research Network, Duke University). Retrieved March 25, 2011, from
https://offshoring.fuqua.duke.edu/orn_report.pdf

Li, X. (2006). Cognitive Analysis of Students’ Errors and Misconceptions in

Variables, Equations, and Functions (Doctoral dissertation). Available from
ProQuest Dissertations & Theses database. (UMI No. 3296453)

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

Lobato, J., & Siebert, D. (2002). Quantitative Reasoning in a Reconceived View

of Transfer. Journal of Mathematical Behavior , 21, 87-116.
doi:10.1016/S0732-3123(02)00105-0

Lorge, I., & Kruglov, L. (1953). The Improvement of Estimates of Test

Difficulty. Educational and Psychological Measurement , 13 (1), 34-46.
doi: 10.1177/001316445301300104

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

Ma, X. (2009). Understanding the Relationship Between Mathematics and

Science Coursework Patterns. Teachers College Record , 111 (9), 2101-
2126. Retrieved March 25, 2011, from
http://www.tcrecord.org.ezproxy1.lib.asu.edu

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.

Mason, J. (2008). Making Use of Children’s Powers To Produce Algebraic

Thinking. In J. J. Kaput, D. W. Carraher, & M. L. Blanton (Eds.), Algebra
in the Early Grades. Mahwah, NJ: Lawrence Erlbaum Associates.

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

McBee, M. (2010). Modeling Outcomes with Floor or Ceiling Effects: An

Introduction to the Tobit Model. The Gifted Child Quarterly , 54 (4), 314-
320. doi:10.1177/0016986210379095

McDermott, L. C., Rosenquist, M. L., & Van Zee, E. H. (1987). Student

Difficulties in Connecting Graphs and Physics: Examples from Kinematics.
American Journal of Physics , 55 (6), 503-513.

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

Moschkovich, J. N. (1992). Making Sense of Linear Equations and Graphs: An

Analysis of Students' Conceptions and Language Use (Doctoral
dissertation). Available from ProQuest Dissertations & Theses database.
(UMI No. 9305013)

Moschkovich, J. N. (1999). Students' Use of The X-Intercept as an Instance of a

Transitional Conception. Educational Studies in Mathematics , 37 (2), 169-
197.

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.

Moses, R. (1995). Algebra, the New Civil Right. In C. B. Lacampagne, W. Blair,

& J. J. Kaput (Eds.), The Algebra Initiative Colloquium (Vol. 2, pp. 53-67).
Washington, DC: U. S. Department of Education, Office of Educational
Research and Improvement.

Nathan, M. J., & Koedinger, K. R. (2000). Teachers' and Researchers' Beliefs

about the Development of Algebraic Reasoning. Journal for Research in
Mathematics Education , 31 (2), 168-190.

National Center for Education Statistics. (1996). Eighth-Grade Algebra Course-

Taking and Mathematics Proficiency (NCES 96-815). Washington, DC:
Author. Retrieved March 25, 2011, from
http://nces.ed.gov/pubs96/web/96815.asp

National Center for Education Statistics. (2010). Contexts of Elementary and

Secondary Education: Characteristics of Public Schools (Table A-24-4).
U.S. Department of Education.Retrieved April 3, 2011 from
http://nces.ed.gov/programs/coe/2010/section4/table-cps-4.asp

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

National Council of Supervisors of Mathematics. (1977). Position Paper on Basic

Mathematical Skills. Retrieved March 25, 2011, from
http://www.mathedleadership.org/docs/resources/positionpapers/NCSMPosi
tionPaper01_1977.pdf

National Council of Teachers of Mathematics. (1980). An Agenda for

Action:Recommendations for school mathematics of the 1980s. Reston,
VA: Author. Retrieved March 25, 2011, from
http://www.nctm.org/standards/content.aspx?id=17278

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 Council of Teachers of Mathematics. (2000). Principles and Standards

for School Mathematics. Reston, VA: Author. Retrieved March 25, 2011,
from http://www.nctm.org/standards/content.aspx?id=16909

National Mathematics Advisory Panel. (2008). Foundations for Success: The

Final Report of the National Mathematics Advisory Panel. Washington,
DC: U.S. Department of Education. Retrieved March 25, 2011, from
http://www2.ed.gov/about/bdscomm/list/mathpanel/report/final-report.pdf

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

National Science Board. (2010). Globalization of science and engineering

research: A companion to the science and engineering indicators 2010.
Arlington, VA: National Science Foundation (NSB-10-3). Retrieved March
25, 2011, from http://www.nsf.gov/statistics/nsb1003/

Newburgh, R. (2001). An experiment to give physical meaning to the

mathematical concept of slope. Physics Education , 36 (4), 316-319.
doi: 10.1088/0031-9120/36/4/306

Norman, A. (1992). Teachers’ Mathematical Knowledge of the Concept of

Function. In G. Harel, & E. Dubinsky (Eds.), The Concept of Function:
Aspects of Epistemology and Pedagogy (pp. 215-232). Washington, DC:
Mathematics Association of America.

Orton, A. (1983). Students' Understanding of Differentiation. Educational Studies

in Mathematics , 14 (3), 235-250.

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

Piaget, J. (1971). Genetic Epistemology. (E. Duckworth, Trans.) New York:

W.W. Norton.

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.

Public Works. (2006). From Education to Work: Is Arizona Prepared? The

Alignment Project Report. Submitted to Arizona P-12 Council . Retrieved
March 25, 2011, from
http://azmemory.lib.az.us/cgibin/showfile.exe?CISOROOT=/statepubs&CI
SOPTR=3074&filename=3359.pdf

Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical Linear Models:

Applications and Data Analysis Methods . Thousand Oaks, CA: Sage.

Rettig, M. D., & Canady, R. L. (1998). High Failure Rates in Required

Mathematics Courses: Can a Modified Block Schedule Be part of the Cure?
NASSP Bulletin , 82 (596), 56-65. doi:10.1177/019263659808259611

Rey, J. D. (2006). Reading Luca Pacioli’s Summa in Catalonia: An early16th-

century Catalan manuscript on algebra and arithmetic. Historia
Mathematica , 33, 43–62. doi:10.1016/j.hm.2005.03.004

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/

Rowntree, R. V. (2009). Students' understandings and misconceptions of

algebraic inequalities. School Science and Mathematics , 109 (6), 311-312.
doi:10.1111/j.1949-8594.2009.tb18100.x

Rousseeuw, P. J. (1987). Silhouettes: a graphical aid to the interpretation and

validation of cluster analysis. Journal of Computational and Applied
Mathematics , 20, 53-65. doi:10.1016/0377-0427(87)90125-7

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.

Sfard, A. (1991). On the Dual Nature of Mathematical Conceptions: Reflections

on Processes and Objects as Different Sides of the Same Coin. Educational
Studies in Mathematics , 22 (1), 1-36.

Sfard, A. (1992). Operational Origins of Mathematical Objects and the Quandary

of Reification- The Case of Function. In G. Harel, & E. Dubinsky (Eds.),
The Concept of Function: Aspects of Epistemology and Pedagogy (pp. 59-
84). Washington, DC: Mathematics Association of America.

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.

Sierpinska, A. (1992). On understanding the notion of function. In E. Dubinsky,

& G. Harel (Eds.), The Concept of Function: Elements of Pedagogy and
Epistemology. (Vol. 25, pp. 25-58). Notes and Reports Series of the
Mathematical Association of America.

Silver, E. A., & Thompson, A. G. (1984). Research Perspectives on Problem

Solving in Elementary School Mathematics. The Elementary School
Journal , 84 (5), 529-545.

Slavit, D. (1997). An Alternate Route to the Reification of Function. Educational

Studies in Mathematics , 33 (3), 259-281.

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.

Speilhagen, F. (2006). Closing the achievement gap in math: The long-term

effects of eighth grade algebra. Journal of Advanced Academics , 18 (1),
34–59. doi: 10.4219/jaa-2006-344

Strauss, A., & Corbin, J. (1998). Basics of Qualitative Research: Techniques and
Procedures for Developing Grounded Theory. Thousand Oaks, CA: Sage.

Stump, S. L. (1997, March). Secondary Mathematics Teachers' Knowledge of the

Concept of Slope. Paper presented at the Annual Meeting of the American
Educational Research Association, Chicago, IL. Retrieved March 25, 2011,
from http://www.eric.ed.gov/PDFS/ED408193.pdf

Stump, S. L. (2001). High School Precalculus Students' Understanding of Slope

as Measure. School Science and Mathematics , 101 (2), 81-89.
doi: 10.1111/j.1949-8594.2001.tb18009.x

Stump, S. L. (2001). Developing preservice teachers' pedagogical content

knowledge of slope.Journal of Mathematical Behavior , 20 (2), 207-227.
doi:10.1016/S0732-3123(01)00071-2

Teuscher, D., & Reys, R. E. (2010). Slope, Rate of Change, and Steepness: Do
Students Understand These Concepts? Mathematics Teacher , 103 (7), 519-
524.

Thompson, A. G., & Thompson, P. W. (1996). Talking about rates conceptually,

Part II: Mathematical knowledge for teaching. Journal for Research in
Mathematics Education , 27 (1), 2-24.

Thompson, P. W. (1994). Images of Rate and Operational Understanding of the

Fundamental Theorem of Calculus. Educational Studies in Mathematics ,
26 (2/3), 229-274.

Thompson, P. W. (1994). The development of the concept of speed and its

relationship to concepts of rate. In G. Harel, & J. Confrey (Eds.), The
development of multiplicative reasoning in the learning of mathematics (pp.
179-234). Albany, NY: SUNY Press.

Thompson, P. W., & Thompson, A. G. (1994). Talking about rates conceptually,

Part I: A teacher's struggle. Journal for Research in Mathematics
Education , 25 (3), 279-303.

247
Tourniaire, F., & Pulos, S. (1985). Proportional Reasoning: A Review of the
Literature. Educational Studies in Mathematics , 16 (2), 181-204.

Uebersax, J. S. (2000). Estimating a Latent Trait Model by Factor Analysis of

Tetrachoric Correlations. Retrieved October 7, 2009, from Statistical
Methods for Rater Agreement: http://john-uebersax.com/stat/irt.htm#tetra

Uebersax, J. S. (2006). The tetrachoric and polychoric correlation coefficients.

Retrieved October 5, 2009, from Statistical Methods for Rater Agreement:
http://john-uebersax.com/stat/tetra.htm

Usiskin, Z. (1988). Conceptions of school algebra and uses of variables. In A. F.

Coxford & A. P. Shulte, (Eds.) The Ideas of Algebra, K-12, 1988 Yearbook
of the National Council of Teachers of Mathematics (pp. 8-19). Reston, VA:
National Council of Teachers of Mathematics.

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

Vergnaud, G. (1994). Multiplicative Conceptul Field: What and Why? In G.

Harel, & J. Confrey (Eds.), The Development of Multiplicative Reasoning in
the Learning of Mathematics (pp. 41-60). SUNY Press.

Vergnaud, G. (2009). The Theory of Conceptual Fields. Human Development , 52

(2), 83–94. doi:10.1159/000202727

Vinner, S. (1992). The Function Concept as a Prototype for Problems in

Mathematics Learning. In G. Harel, & E. Dubinsky (Eds.), The Concept of
Function: Aspects of Epistemology and Pedagogy (pp. 195-213).
Washington, DC: Mathematics Association of America.

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

You, Z. (2006). Preservice Teachers' Knowledge of Linear Functions Within

Multiple Representation Modes (Doctoral dissertation). Available from
ProQuest Dissertations & Theses database. (UMI No. 3280530)

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.

Zhou, W. (2009). Teachers’ Estimation of Item Difficulty: What Contributes to

Their Accuracy? In S. L. Swars, D. W. Stinson, & S. Lemons-Smith (Ed.),
Proceedings of the 31st annual meeting of the North American Chapter of
the International Group for the Psychology of Mathematics Education. 5,
pp. 261-264. Atlanta, GA: Georgia State University.

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?

(b) How did you decide?

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

in the table below.

Which of the following equations represents the relationship between the number

of cars sold (s) and the number of weeks (w)?

(a) s = 6w; (b) s = 12w; (c) s = 6w + 6; (d) s = 6w + 12

Problem 6

Given this linear equation 2x + 3y = 12, which of the following is the y-intercept?

(a) 0; (b) 2; (c) 4; (d) 6; (e) 12

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?

(b) What is the slope of part R of the graph?

(c) What is the speed of the car in part R?

255
 2. Information Questionnaire for Students

Tell us about yourself:

1. Name (print)
_____________________________________________________

2. Grade (8, 9 or 10) ___________

3. Name of your math course (e.g., Pre-Algebra, Algebra I, Geometry)

____________________________________________________

4. Name of your math teacher

____________________________________________________

256
3. Ranking Questionnaire for Students

1. Solve the seven problems.

2. Show all of your work on the pages.

3. When you finish the problems, please complete the table.

Tell us which problem was most difficult.

Then write the other numbers of the problems in order of difficulty.

Difficulty Level Problem Number

Most difficult

Least difficult

4. For the problem you named the most difficult, tell us why you chose it.

257
4. Information Questionnaire for Teachers

Information About Yourself

1. Highest college degree?

2. College major?

3. Titles of courses you are currently teaching.

4. Titles of the courses you have taught in the past 2 years.

5. Number of years you have been teaching?

6. Name (print)

7. School:

8. Phone:

9. Email:

258
5. Ranking Questionnaire for Teachers

1. Solve the seven problems.

2. Which of the problems will be most difficult for students? Least difficult?

Rank order the problems by difficulty.

Use the numbers of the problems.

Complete the table.

Difficulty Level Problem Number

Most difficult for students

Least difficult for students

3. For the two problems that you identified as most difficult for your students,

describe the student difficulty.

a) Most Difficult:

b) Next Most Difficult:

259
 APPENDIX B

SCORING GUIDE AND CODING MANUAL

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

the number of correctly completed rows (0, 1, 2, 3, 4, or 5) in the table of values.

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

Scoring Guide for MDT Scores by Problem

Problem Problem Problem Problem Problem Problem Problem 7

1 2 3 4 5 6
a b c
0, .5 or 0, .2, .4, 0 or 1 0 or 1 0 or 1 0 or 1 0 or 0 or 0, .5,
1 .6, .8 or 1 1 1 or 1

Use Tables 52-60 to record the frequencies (number of students) for the

types of responses for each MDT problem.

261
 Table 52

Scoring Guide for MDT Problem 1 Types of Responses

PART Types of Answer Number of

Students

PART 1. Correct Answer “Yes”

(a)
2. Answer “No”

3. No answer or an answer different from “Yes” or “No”

PART 1. Correct answer using the Cartesian Connection

(b) Identifying (2,8) as ( x 0 , y 0 ) and substituting in
y 3x  14 ; and
Checking if the statement 8 3(2)  14 (or equivalent
statement) is true

2. Incorrect or incomplete answer while attempting to use the

Cartesian connection

3. Correct answer using a graph

Graphing the line y 3x  14 (using a quick graph or a table of
values);
Plotting the point (2, -8); and
Inspecting to see if the point is on the line

4. Incomplete or incorrect answer from use of a graph

5. Incomplete or incorrect responses different from those

above from Part (b), #1-4

6. No answer or wrote “I guessed”

262
 Table 53

Scoring Guide for MDT Problem 2 Types of Responses

Types of Answers Number of

Students

1. 5 rows correctly completed

2. 4 rows correctly completed

3. 3 rows correctly completed

4. 2 rows correctly completed

5. 1 row correctly completed

6. No row correctly completed, but attempted to answer

7. No answer

Table 54

Scoring Guide for MDT Problem 3 Types of Responses

Types of Answers Number of Students

1. a

2. Correct answer b

3. c

4. d

263
Table 55

Scoring Guide for MDT Problem 4 Types of Responses

Types of Answers Number of

Students

1. Correct answer “3”

2. Computational errors in calculation of slope, but correct formula

3. Reversed ratio

4. Incorrect answer different from answers to 2 and 3

5. Irrelevant information

6. No answer or answer, “I don’t know.”

Table 56

Scoring Guide for MDT Problem 5 Types of Responses

Types of Answers Number of Students

1. a

2. b

3. Correct answer c

4. d

264
Table 57

Scoring Guide for MDT Problem 6 Types of Responses

Types of Answers Number of Students

1. a

2. b

3. Correct answer c

4. d

Table 58

Scoring Guide for MDT Problem 7a Types of Responses

Types of Answers Number of Students

1. Answer “R”

2. Answer “S”

3. Correct answer “T”

4. Answer different “R”, “S”, or “T “

5. No answer

265
 Table 59

Scoring Guide for MDT Problem 7b Type of Responses

Types of Answers Number of Students

1. Correct answer “40”

2. Incorrect answer

3. No answer or irrelevant information

Table 60

Scoring Guide for MDT Problem 7c Types of Responses

Types of Answers Number of Students

1. Correct answer “40mph”

2. Correct numeric answer, but incorrect

or no unit of measurement given

3. Incorrect answer

4. No answer or irrelevant information

PART II: Coding the Written Comments from Ranking Questionnaires

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

found the problem difficult.

266
Code student and teacher comments on the Ranking Questionnaires using the

categories H, L, M, R, T, U shown in Tables 61-66. Note: Each table contains a

description of the category and examples of what constitutes evidence for its

presence, for both students and teachers.

Table 61

Category H: Hard Problem

Description: A general statement about the difficulty of a problem, topic or

domain. Sometimes a mathematical concept is identified as “hard” or “difficult.”

Other times a reference to unsatisfactory performance is made, but no specific

details about the nature of the difficulty are given.

Examples from Students’ Written Examples from Teachers’ Written

Comments Comments

“It was a difficult question.” “This problem is hard for my

students.”
“I thought that one was the hardest was
because those types are just hard to me.” “Slope is difficult for student.”

“I just really have a hard time on slopes.”

267
Table 62

Category L: Never Learned /Taught

Description: A statement about the novelty of a problem, or that a mathematical

concept or how to solve a certain problem was not learned or taught.

Examples from Students’ Written Examples from Teachers’ Written

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

Category M: Mathematical Difficulty

Description: Nature of difficulty refers to lack of understanding of mathematical

concepts, or their representations, or difficulties with algorithms/procedures,

computations, and symbolic manipulations.

Examples from Students’ Written Examples from Teachers’ Written

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.“

“Because it was hard to determine the “No points labeled.”

exact numbers the dots were placed.”
“Using numbers so high.”

Table 64

Category R: Not Remembering

Description: A statement about forgetting or not remembering the mathematical

concepts or how to solve a problem.

Examples from Students’ Written Examples from Teachers’ Written

Comments Comments

“I forgot how to do this type.” “Students forget how to find slope.”

“I didn't remember how to do it.” “Students forget what slope means.”

“I couldn't remember the formula for “They don’t know where to start.”
slope.”

269
Table 65

Category T: Thinking Time

Description: A reference to the amount of time, thinking, memory or any other

resource necessary to solve a problem, or to a problem feature, as for example the

number of subparts to be completed.

Examples from Students’ Written Examples from teachers’ Written

Comments Comments

“Had to do the most work.” “However, they do have difficulty

explaining how they decide things.”
“You had to find 5 values which took
time.” “Lots of reading.”

“It was the problem I had to think on.”

“It required the biggest formula.”

“It took a long time for one problem.”

Table 66

Category U: Not Understanding

Description: A general statement about not understanding a mathematical

concept, topic or domain, how to solve a problem, or the vocabulary used, without

offering any further details about the nature of the difficulty.

Examples from Students’ Written Examples from Teachers’ Written

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.”

“No understandings of slope.”

270
Part II: Coding the Interviews

Students and teachers solve the MDT problems aloud, and comment on the

nature of difficulties for only MDT problems 1, 2, 4 and 7b.

Conceptual Knowledge of the Cartesian Connection (CKC) is inferred from

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

related to this problem?”

Table 67 presents CKC subcategories and samples of student and teacher

comments to assist in the coding.

271
 Table 67

Conceptual Knowledge of the Cartesian Connection (CKC) and its

Subcategories

CKC Subcategories Examples from:

A. Students’ Interviews
B. Teachers’ Interviews
CKC-D From Point to Line A. “I will substitute in the x value for 2 and
A point ( x 0 , y 0 ) is on the graph of the the y value as 8 . So my equation would now
line l if its coordinates satisfy the be 8 3(2)  14 . First I multiply 3 by 2 and
equation of l, y mx  b , i.e. if the that equals 6 and I subtract, 6 minus 14 and
statement y 0 mx 0  b that is 8, and 8 equals 8. That’s a check
is a true statement. so I would say ‘yes’.”

B. “They need to recognize that points should

make this equation true and if it is true then
the point is on the line and if it’s not true the
point’s not on the line.”

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 .

The Conceptual Knowledge of Slope (CKS) is inferred from students’ answers

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

difficulties (e.g., with slope)?”

CKS subcategories and samples of student and teacher comments to assist in

the coding are presented in Table 68.

272
Table 68

Conceptual Knowledge of Slope (CKS) and its Subcategories

CKS Subcategories Examples from:

A. Students’ Interviews
B. Teachers’ Interviews
CKS-P Physical Context A.”Slope is like the elevation of the
Slope describes the steepness of line.”
physical objects such as mountain
roads, sky slopes, and wheelchair B. “Like when you have a ladder, if
ramps. A statement about an entire you are leaning it up against the wall
class of objects sharing the property of and when you go down the slide. A
having a slope or being steep, or an roof has a slant that slant is the slope of
instantiation/particular example given. the roof.”

CKS-G Geometric-Analytical Context A. “To find the slope of a line I would

Given two points ( x1 , y1 ) and do the rise over the run. I actually don’t
( x 2 , y 2 ) , the slope of the line remember how to find the slope with
just an equation or formula right now,
connecting the two points is “rise over so I’m just going to go ahead and
y 2  y1 count… the rise is 1, 2, 3, 4, 5, 6, 7, 8,
run,” or m .
x 2  x1 9 and the run is 1, 2, 3, I believe. So it
would be 9 over 3 because rise over
Homogeneous Systems of Coordinates: run, or reduced, would be a slope of
Slope is a geometric property of the positive 3.”
line y mx  b as well as an analytical
B. “The the rise over the run using the
property of the linear function
formula that we’ve given to them,
y mx  b .
y 2  y1
, which is the slope formula.”
x 2  x1
Non-homogeneous Systems of
Coordinates: Slope is an analytical
property of the function y mx  b .

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

steps of the algorithm used to solve MDT problem 1.

Table 69 describes the SDC-D subcategories and presents samples of student

and teacher comments in each subcategory.

Table 69

Student Difficulties with Cartesian Connection from Point to Line SDC-D

(MDT Problem 1)

SDC-D Subcategories Examples from:

A. Students’ Interviews
B. Teachers’ Interviews
SDC-DP A. “ I’m not too sure about this but, I’d say that…
Identifying (2,8) as like that.” (Researcher Note: The student
( x 0 , y 0 ) and replacing wrote y 3(2,8)  14 )
variables with their values
in y 3x  14 . B. “They have a problem of substituting,
basically, and knowing an ordered pair.
Knowing that x comes first, y comes second
and knowing that they are substituting in these
values into this equation.”

SDC-DT A. “You take this equation y 3x  14 and you

Checking if the statement plug in what you got right here so -8 equals
8 3(2)  14 3times 2 minus 14 is going to be right here so
(or other equivalent you get 8 6  14 and then from there wait,
statement) is true . hold on, oh, I forgot how to do this.”

B. “They need to know that 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 equal on
both sides.”

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

the steps of the algorithm used to solve MDT problem 2.

Table 70 describes SDC-C subcategories and presents examples of student and

teacher comments to assist in the coding.

Table 70

Student Difficulties with Cartesian Connection from Line to Point SDC-C

(MDT Problem 2)

SDC-C Subcategories Examples from:

A. Students’ Interviews
B. Teachers’ Interviews
SDC-CX A. “ I think you go and plug or for x you put 0, -1, or you
Identifying x-values put a negative, 0, a positive and then fractions so you can
get all of them, see, if you were going to put and use I
think but I don’t know…”

B.“…It’s sort of open-ended in that they can fill in any

value for x …just because of how open-ended it was. If
the problem included values for x, it might be a little bit
less difficult for them.”

SDC-CY A.”I would then go on and do is find out my y coordinates

Finding corresponding y- by replacing x with 0, and then continuing on from there
values and so after I replace x with 0 I get that y 3
…”(Researcher Note: She gives values to x and solves for
y; then she gives values to y and solves for x. She records
these values in her table, but they do not satisfy
simultaneously the equation y 2 x  3 ). “

B. “Finding y is going to be difficult, and filling out that

table.”

Student Difficulties with Slope (SDS) are inferred from student answers and

teachers’ comments on difficulties with slope. Only difficulties with slope in

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

Homogenous System of Coordinates (SDS-H), accompanied by samples of

student and teacher comments to assist in the coding.

Table 71

Student Difficulties with Slope in Homogeneous Systems of Coordinates

SDS-H (MDT Problem 4)

SDS–H Subcategories Examples from:

A. Students’ Interviews
B. Teachers’ Interviews
SDS- HR A. “I can’t remember if I’m supposed to add 9 plus 3
Calculating slope as a ratio or multiply or divide.”

B. “Students don’t know what to do with the rise and

the run. They don’t know they have to divide because
slope is a ratio.”

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)…”

Table 72 presents the subcategories for Student Difficulties with Slope in a

Non-Homogenous System of Coordinates (SDS-N) and samples from student

and teacher comments to assist in the coding.

276
Table 72

Student Difficulties with Slope in Non-Homogeneous Systems of Coordinates

SDS-N (MDT Problem 7b)

SDS-N Subcategories Examples from:

A. Students’ Interviews
B. Teachers’ Interviews
SDS-NM A. “I don’t know if you have to take these numbers
Considering slope as a into like, consideration, if you count it as 30, as go
geometric property of the line, up 30 or 4 over 3 but it doesn’t really make sense
and not an analytical property because that means you’d have to go up 4, oh no, it
of the function in non- does make sense … ah, I don’t know. “
homogeneous systems of
coordinates from problem 7. B. “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, 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.”

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.
------------------------------------------------------------------------------------------------

I, ______________________________, agree to participate in the algebra study.

(Printed Name)

________________________________ ____________
(Signature ) (Date)

279
2. Parental Consent for Interviewing Minors

Dear Parent or Guardian,

The School of Educational Innovation and Teacher Preparation at Arizona
State University is focusing on ways to help children learn mathematics. One of
the topics that is often difficult for students is algebra. As part of this study, your
child completed a short quiz. You signed a consent form for that activity. To gain
greater insight into students’ reasoning processes, we would like your permission
to have your child talk about how some of the problems were solved. This “think
aloud” will be audio taped, and will take about 20 minutes to complete. The
taping will take place during school time. Your child’s name will not be used
during the taping and will not be used in the transcription. Only the research team
will have access to the audiotape. At the end of our study, the tape will be
destroyed.
Your child's participation in this study is voluntary. If your child chooses not
to participate or to withdraw from the study at any time it will not affect the
course grade.
If you would like additional information, please contact me or Valentina
Postelnicu using the contact information below.
Thank you again for assisting with this important study.
With best wishes,
Carole Greenes, Ed.D.
Dean and Professor Mathematics Education, School of Educational Innovation
and Teacher Preparation, Arizona State University (480) 727-1820,
e-mail: [email protected]

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.
---------------------------------------------------------------------------------------------------

______________________________________ to be audio taped for this study.

Student Name

__________________ _____________________________ ____________

Parent Signature Parent Name Date

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.
----------------------------------------------------------------------------------------------

I, _______________________, agree to be taped while I solve algebra problems.

(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