LIBRARY
Michigan State
University

 

——v———

This is to certify that the
dissertation entitled

TEST ANXIETY ASSOCIATED WITH HIGH-STAKES TESTING
AMONG ELEMENTARY SCHOOL CHILDREN: PREVALENCE,
PREDICTORS, AND RELATIONSHIP TO STUDENT
PERFORMANCE

presented by

NATASHA KATHERINE SEGOOL

has been accepted towards fulfillment
of the requirements for the

PhD. degree in School Psychology

 

 

 

 

(”rig gr Professor’s Signature

(1/31/0 ‘2
Date

MSU is an Affirmative Action/Equal Opportunity Employer

 

4 _._-_-—-—-—-—-a-—

V‘-~<-

WW', 4

PLACE IN RETURN Box to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

DATE DUE DATE DUE DATE DUE

 

TEST ANXIETY ASSOCIATED WITH HIGH-STAKES TESTING AMONG
ELEMENTARY SCHOOL CHILDREN: PREVALENCE, PREDICTORS, AND
RELATIONSHIP TO STUDENT PERFORMANCE
By

Natasha Katherine Segool

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
School Psychology

2009

ABSTRACT
TEST ANXIETY ASSOCIATED WITH HIGH-STAKES TESTING AMONG
ELEMENTARY SCHOOL CHILDREN: PREVALENCE, PREDICTORS, AND
RELATIONSHIP TO STUDENT PERFORMANCE
By
Natasha Katherine Segool

The current study explored differences in test anxiety on high-stakes standardized
achievement testing and classroom testing among elementary school children. This is the
first study to directly examine differences in student test anxiety across two testing
conditions with different stakes among young children. Three hundred and thirty-five
students in grades three through five participated in the current study. Wilcoxon Signed
Rank Tests revealed that as a whole, students reported significantly more overall test
anxiety in relation to high-stakes testing than classroom testing on two difl‘erent measures
of test anxiety (r = -.21 and r = -.10). Students also reported significantly more cognitive
(r = -.20) and physiological (r = -.24) symptoms of test anxiety in relation to high-stakes
testing than classroom testing. Teachers’ perceptions of student anxiety were also
examined in the current study and consistent with student self-reported test anxiety,
results indicated that teachers believed students experienced significantly more
anticipatory anxiety in relation to high-stakes testing versus classroom testing.

Analyses of variance indicated that low test-anxious students performed
significantly better on high-stakes testing while there was no difference in performance
between moderate and high test-anxious students. Further, multiple regression analyses
indicated that test anxiety contributed significantly to English Language Arts test

performance among third and fourth grade students and Math test performance among

third grade students. Additionally, the current study examined the relationship between
test anxiety and student demographic characteristics. Results of multiple regression
analyses indicated that student gender and grade significantly predicted student test
anxiety, while student ethnicity, educational verification, and socioeconomic status did
not.

The current study also examined the relationship between three measures of test
anxiety and differences in how three test anxiety classification systems sorted students
into low, moderate, and high test-anxious groups. Correlation analyses indicated that the
Children’s Test Anxiety Scale for Children (CTAS), Test Anxiety Scale for Children
(TASC), and Behavior Assessment Scale for Children, Second Edition, Test Anxiety
subscale (BASC-Z-TA) were strongly related to one another (rs > .70). However, chi-
square goodness of fit tests revealed that there were significant differences in the
proportions of students classified as low, moderate,'and high test-anxious across
classification systems. Using the CTAS and TASC to classify students, prevalence rates
of high test anxiety in relation to classroom testing ranged from 11 to 21 percent,
respectively. Similar rates of students were classified as highly test-anxious across the

high-stakes and classroom testing conditions using the CTAS.

Copyfightby
NATASHA KATHERINE SEGOOL

2009

ACKNOWLEDGEMENTS

I would like to acknowledge several individuals who provided much appreciated
support to me throughout my graduate schooling. First, my thanks are extended to Dr.
John Carlson, my advisor and dissertation chair, who encouraged and supported me to
pursue my passion of applied psychosocial research throughout my studies at Michigan
State University. Additionally, I want to thank my dissertation committee members, Drs.
Sara Bolt, John Kosciulek, and Ed Roeber, for their time, thoughtful critiques, and
interest in this project. My appreciation is expressed to the school teachers, staff, and
administrators who provided invaluable support in completing this research project.
Additionally, I want to thank all of the school children and families who participated in
this project. My heartfelt gratitude also extends to Anisa Goforth who diligently
coordinated the data collection effort and to Justin Barterian who contributed to data
collection and entry efforts. Finally, I would like to'thank my family for their unwavering
interest, support, and care throughout my graduate studies. This research project was
partially funded by the Michigan Association of School Psychologists Research Grant
and the School Psychology Program Student Research Award. Additional support was
provided by the Leadership Training Grant Fellowship from the US. Department of

Education, Office of Special Education Programs.

TABLE OF CONTENTS

 

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

LIST OF FIGURES ......................................................................... x
CHAPTER 1

INTRODUCTION ............................................................................. 1
CHAPTER 2

LITERATURE REVIEW .................................................................... 5

TestAnxiety” 5

Theoretical Medals ......................................................... 6

Prevalence Rates ........................................................... 9

Demographic Patterns ..................................................... 12

Effects on Performance and Educational Outcomes.....................15

Educational Accountability Movement ........................................... 20

No Child Left Behind Act of 2001 ...................................... 22

Michigan Educational Assessment Program ........................... 24

Test Anxiety and High-Stakes Assessment ...................................... 26

Direct Studies on the Impact of Testing on Children ................ 27

Indirect Studies on the Impact of Testing on Children .. ................ 31

Current Study ........................................................................ 35

Research Questions and Hypotheses .................................... 38

Question One ...................................................... 38

Question Two ..................................................... 39

Question Three .................................................... 39

Question Four ..................................................... 40

Question Five ..................................................... 41
CHAPTER 3

METHOD ..................................................................................... 43

Subjects .............................................................................. 43

Measures ............................................................................... 48

Children’s Test Anxiety Scale ........................................... 48

Behavior Assessment Scale for Children, Second Edition, Test

Anxiety “ L ‘ 49

Test Anxiety Scale for Children ......................................... 50

Demographic Information Form .......................................... 50

Academic Performance Fonn.. .51

Michigan Educational Assessment Program Data ..................... 51

Teacher Perception Questionnaires ..................................... 52

Operational Definition of Test Anxiety .......................................... 53

Definition One ............................................................. 53

Definition Two ............................................................ 53

Definition Three ............................................................ 54

Data Collection Procedures ........................................................ 54

Recruitment, Consent, and Assent ...................................... 54
MEAP Test Anxiety Data Collection..................................... 55
Classroom Test Anxiety Data Collection ................................ 56
Data Analysis ........................................................................ 56
Question One ............................................................... 57
Aim 1 ................................................................ 57
Aim 2 ............................................................... 57
Question Two ............................................................... 58
Question Three .............................................................. 58
Aim 1 ............................................................... 58
Aim 2 ............................................................... 58
Question Four .............................................................. 59
Aim 1 .-. .......................................................... 59
Aim 2 ............................................................... 59
Question Five ............................................................... 60
Aim 1 ............................................................... 6O
Aim 2 ............................................................... 60
CHAPTER 4
RESULTS ..................................................................................... 62
Question One ............................................................................ 62
Aim 1 ........................................................................ 62
Aim 2 ........................................................................ 64
Question Two .................................... 69
Question Three ....................................................................... 71
Aim 1 ........................................................................ 71
Aim 2 ........................................................................ 73
Question Four ........................................................................ 81
Aim 1 ........................................................................ 81
Aim 2 ........................................................................ 88
Question Five ........................................................................ 92
Aim 1 ........................................................................ 92
Aim 2 ........................................................................ 97
CHAPTER 5
DISCUSSION ............................................................................... 99
Student Self-Report of Test Anxiety ............................................. 100
Teacher Perceptions of Testing ................................................... 102
Test Anxiety Classifications and Rates across Measures .................... 103
Test Anxiety Classifications across Conditions ............................... 107
Test Performance Differences across Test Anxiety Levels ................... 111
Association between Test Performance, Test Anxiety, and
Student Characteristics ............................................................ 114
Association between Test Anxiety and Student Characteristics ............. 116

vii

Measures of Test Anxiety .......................................................... 120

Limitations ........................................................................... 122
Clinical and Educational Implications ........................................... 124
Future Research ..................................................................... 126
APPENDICES ................................................................................ 130
Appendix A: Demographic Information Form .................................... 131
Appendix B: Academic Performance Form ...................................... 133
Appendix C: Teacher High-Stakes Testing Questionnaire .................... 135
Appendix D: Teacher Classroom Testing Questionnaire ...................... 137
Appendix E: Parental Information Letter ....................................... 139
Appendix F: Consent Form ........................................................ 141
Appendix G: Survey One; .......................................................... 144
Appendix H: Survey Two .......................................................... 149
Appendix 1: Survey Directions ................................................... 156
REFERENCES ................................................................................. 160

viii

LIST OF TABLES
Table 1: Demographic Characteristics of Student Participants...
Table 2: Demographic Characteristics of Student Participants by School ............

Table 3: Test Anxiety Scores on the MEAP Assessment and
Classroom Testing ..................................................................

Table 4: Differences in the Number of Students Reporting Low, Moderate,
and High Test Anxiety between MEAP and Classroom Testing
Conditions by Definition ...........................................................

Table 5: Teacher Perceptions of Student and Teacher Test Anxiety
by Testing Condition ...............................................................

Table 6: Differences in MEAP Performance Level by Test Anxiety Level ..........
Table 7: Multiple Regression of Predictors of MEAP Scores in Grade 3 ..........
Table 8: Multiple Regression of Predictors of MEAP Scores in Grade 4 ..........
Table 9: Multiple Regression of Predictors of MEAP Scores in Grade 5 ..........

Table 10: Differences in Test Anxiety in Relation to the MEAP Testing
Condition by Demographic Variables ............................................

Table 11: Differences 1n Test Anxiety In Relation to the Classroom Testing
Condition by Demographic Variables.

Table 12: Multiple Regression of Demographic Variables PredictIng
Test Anxiety Scores..

Table 13: Correlations between Measures of Test Anxiety in Relation
to the MEAP

 

Table 14: Correlations between Measures of Test Anxiety in Relation
to Classroom Testing

Table 15: Correlation between Measures Assessing Test Anxiety In Relation
to the MEAP and Classroom Testing. .... ...

45

47

63

65

7o
74
77 .
79

82

83

.. 86

. 90

94

95

98

LIST OF FIGURES

Figure 1: Conceptual Framework of Test Anxiety Adapted

from Zeidner, 1997 and Lowe et al., 2008 ..................................... 10
Figure 2: Methodological Design of the Current Study .................................. 37
Figure 3: Rates of Test Anxiety by Classification Definitions... . . . . . . . . . 68
Figure 4: Shifts in Students’ Test Anxiety Classifications across .................... 109

Testing Conditions

Figure 5: Test Performance Levels by Test Anxiety Classification .................. 112

CHAPTER 1
INTRODUCTION
Test anxiety is associated with significant impairment in performance on
academic achievement assessments, lower courses grades, decreased motivation, and
increased stress (Cizek & Berg, 2006; Hembree, 1988, Zeidner, 1998). Test anxiety is
comprised of psychological, physiological, and behavioral reactions that occur in
association with concern about the negative outcomes resulting from failure or poor
performance in evaluative situations (Zeidner). In comparison to low test-anxious
students, students with high levels of test anxiety score lower on tests of academic
achievement in core academic areas including reading, English, and mathematics
(Hembree). Additionally, students’ grade point averages (GPA) are inversely related to
test anxiety levels, and significantly more test anxious students drop out of school than
non-test anxious students (Cizek & Berg; Hembree; Spielberger, 1966).

The prevalence rate of impairing levels of test anxiety among young children has
been estimated to range from as little as 10% to as much 40% of school-aged children
(King & Ollendick, 1989; Turner, Beidel, Hughes, & Turner, 1993). Female students and
non-Caucasian students consistently report higher prevalence rates of test anxiety
(Hembree, 1988; Putwain, 2007; Turner et al., 1993). Higher prevalence estimates of test
anxiety in contemporary society have been attributed to the increased testing that children
are exposed to in public schools. As Sarason foreshadowed in 1959, “we live in a test-
conscious, test-giving culture in which the lives of people are in part determined by their
test performance” (p.26).

The impact that testing has on the educational and occupational outcomes of

children in the United States has increased as a result of the educational accountability
movement, which has dramatically increased the prevalence of standardized testing for
public school children in elementary, middle, and high school. The No Child Lefi Behind
Act of 2001 (N CLB) requires annual testing and reporting of statewide academic
achievement assessments in the areas of reading and mathematics during grades three
through eight and once in high school. Testing in the content area of science is done once
in elementary, middle, and high school. These testing requirements mean that students
take no fewer than seventeen state-mandated standardized achievement assessments prior
to high school graduation (U .8. Congress, 2002). NCLB testing has high-stakes for
educators and communities because student achievement data are publicly reported and
directly linked to school funding, administration, and employment decisions. However,
students’ psychological reactions to and perceptions of this testing is less clear.

Although there is a clear need to understand the relationship between student test
anxiety and testing required by NCLB, relatively little research has been conducted that
directly examines the relationship between test anxiety and federally mandated annual
achievement assessments. A number of researchers have examined the perceptions that
teachers, parents, and school administrators have about the impact of large-scale testing
programs on students; however, most of these studies are indirect assessments that may
have been influenced by the respondents’ beliefs, concerns, or worries about high-stakes
assessments. Indirect perception studies suggest that state testing programs have resulted
in increased student anxiety, increased stress, lowered motivation (Abrams, Pedulla, &
Madaus, 2003; Barksdale-Ladd & Thomas, 2000; Jones & Egley, 2004; 2006; Jones et

al., 1999), increased focus on test preparation (Abrams et al.; Barksdale-Ladd & Thomas;

Jones et al.), and increased job stress and lowered job satisfaction for teachers
(Barksdale-Ladd & Thomas; Jones et. al; Jones & Egley, 2004; 2006). While these
studies provide important data about how these stakeholders perceive the impact of
statewide testing programs, there is a need to study the impact of testing programs on
children directly in order to understand how students are actually impacted by testing
programs.

The few existing studies that directly examine how students are impacted by
high-stakes testing are equivocal and limited by methodological issues. One projective
study of students’ feeling about testing found that students overwhelmingly felt stress,
anxiety, worry, and isolation as a result of testing (Triplett & Barksdale, 2005).
Alternatively, Mulvenon and colleagues (2001; 2005) found that most students felt
positively about testing and that test anxiety, when present, did not impair student
performance on the state assessment. However, these findings must be interpreted with
caution because the measure of test anxiety used in these studies was not empirically
validated and did not reflect current conceptual models of test anxiety. Most recently,
Putwain (2008) found that although secondary students in the United Kingdom reported
higher levels of test anxiety in relation to low- versus medium- and high-stakes
examinations, there was a small but significant negative relationship between test anxiety
and performance on all examinations. These contradictory findings across studies
highlight the need to systematically examine the relationship between test anxiety and
high-stakes testing among students using empirically validated measures of test anxiety
and methodologically sound data collection techniques.

This dissertation study examined how young children perceive and experience

high-stakes standardized achievement testing. By surveying elementary school students
to assess test anxiety in relation to both classroom testing and state achievement testing
required by NCLB, the levels and prevalence rates of low, moderate, and high test
anxiety associated with high-stakes testing in comparison to classroom testing were
explored. Also, teachers’ perceptions of students’ responses to the different testing
conditions were examined. Next, the relationship between test anxiety and test
performance was examined. Additionally, student demographic characteristics associated
with test anxiety were examined. Finally, the relationship between three measures of
childhood test anxiety was examined. This study significantly adds to and expands upon
the current literature on test anxiety by directly examining the relationship between test
anxiety and high-stakes testing among young students who are required by law to

complete annual standardized achievement assessments.

CHAPTER 2
LITERATURE REVIEW

High levels of test anxiety are associated with detrimental outcomes, including
impaired learning and performance on recall tasks, achievement measures, and
intellectual assessments (Hembree, 1988; Zeidner, 1998). The current political and
educational climate in the United States has resulted in the frequent use of standardized
academic achievement assessments to measure student performance and educational
quality at the school, community,iand state level. This increased emphasis on testing
necessitates a more complete understanding of the impact that test anxiety has on
students’ performance in high-stakes contexts. This literature review will provide a
context for this study by a) defining test anxiety, its presentation, and its effects on
students, b) examining the educational accountability movement and the increasing
emphasis of testing in American public schools, and c) critiquing the existing literature
on test anxiety and assessments administered for educational accountability. Finally, the
rationale for the current dissertation study will be presented along with the current
study’s research questions and hypotheses.

Test Anxiety

Test anxiety is defined as the psychological, physiological, and behavioral
responses associated with concern about experiencing negative outcomes as a result of
failure or poor performance on an assessment or in an evaluative situation (Cizek & Berg,
2006; Zeidner, 1998). In explaining the need to study test anxiety, early test anxiety
researcher Seymour Sarason (1959) argued that “we live in a test-conscious, test-giving

culture in which the lives of people are in part determined by their test performance”

(p.26). Currently, educational testing programs have an even more significant impact on
the lives of children, and a subset of children respond to testing programs with high levels
of test anxiety. In turn, children’s performance on these tests may be negatively affected
by test anxiety (Meijer, 2001). Therefore, there is a pressing need to consider how
students perceive testing programs and how their testing performance is affected by test
anxiety. The following sections will review the construct of test anxiety, the prevalence
rates and developmental patterns of test anxiety, and the relationship between test anxiety
and learning and academic performance.

Theoretical Models

Yerkes and Dodson (1908) published a now classic work on the facilitating and
debilitating effects of arousal on learning. They found that test performance was
enhanced when low levels of arousal were present, whereas high levels of arousal
resulted in performance decrements. The tipping pOint between facilitating and
debilitating levels of arousal shifted depending on the task’s difficulty, with impairment
resulting at lower arousal levels on more difficult tasks. This work anticipated the study
of test anxiety, which focuses on the impact of anxious arousal in testing situations.

In the 19505, Seymour Sarason formalized the study of test anxiety and developed
standardized measures to examine its cognitive and affective components. Sarason and
his colleagues hypothesized that high test-anxious individuals experience increased worry
and autonomic arousal during testing situations. Further, they suggested that this worry
and arousal results in impaired task performance (Zeidner, 1998). Liebert and Morris
(1967) advanced Sarason’s theoretical model of test anxiety by examining how cognitive

and physiological aspects of test anxiety differentially affect test performance. Their

model defined worry as the cognitive concerns related to fear of failure and emotionality
as the affective or autonomic responses to testing situations (St6ber & Pekrun, 2004).
Liebert and Morris found that worry was more strongly related to test performance than
emotionality, suggesting that the impairment associated with test anxiety results from
cognitive interference rather than physiological arousal.

Later, Spielberger’s distinction between state and trait anxiety led to further
investigation about the specific relationship between testing performance and each
anxiety type (Stbber & Pekrun, 2004). State anxiety is transitory and results from a
specific situation whereas trait anxiety is a relatively stable personality trait. Spielberger
and colleagues (1976) found that underlying trait anxiety predisposes individuals to
situation-specific test anxiety when faced with a performance task. Spielberger’s work
suggests that trait anxiety scales that measure underlying anxiousness and worry may be
more closely related to test performance than general state anxiety scales that focus more
on querying physiological arousal (Zeidner, 1998). Later research by Irwin Sarason
confirmed that cognitive aspects of anxiety such as self-critical beliefs, high self-focus,
and a focus on the competitive nature of assessments are more impairing to test
performance than autonomic arousal (Zeidner). The cognitive model of test anxiety
suggests that excessive focus on the self and the task result in impaired attention,
memory, and retrieval during testing, which ultimately results in impaired test
performance (Smith, Ingram, & Brehm, 1983; Wine, 1971). For example, Matters and
Burnett (2003) found that task-irrelevant thinking is associated with a greater rate of
omitting answers to items on a high-stakes achievement test, suggesting that the cognitive

aspects of test anxiety impede students’ testing behaviors and strategies.

More recently, Zeidner (1997) proposed an integrative transactional model of test
anxiety that considers the impact of personal, task, and situational factors on test anxiety,
rather than focusing entirely on person-specific factors, such as worry and emotionality.
The transactional model considers how task-specific factors, such as the test difficulty,
the testing atmosphere, the examiner characteristics, the presence of time constraints, and
the mode of administration, and person-Specific factors, such as trait test anxiety, self-
efficacy, scholastic ability, previous testing experience, and need for achievement, all
influence the responses that individuals have to testing situations (Zeidner). The
transactional model emphasizes the importance of cognition in the development and
expression of test anxiety by suggesting that person- and task-specific factors affect the
cognitive appraisal an individual makes about the test situation (Zeidner). In turn, this
appraisal influences the level of test anxiety experienced by the individual.

Lowe and colleagues (2008) have expanded Zeidner’s transactional model of test
anxiety using a biopsychosocial framework. This model encompasses Zeidner’s
transactional model while also recognizing that societal factors, such as schools,
communities, parents, and families also influence the development and expression of test
anxiety. This biopsychosocial model is particularly relevant given the increasingly
important role that educational testing programs have in evaluating schools and
educators. The model highlights three different processes involved in the expression of
test anxiety, including the individual’s behavior, cognition, and physiology. Behaviors
include both task-relevant and task-irrelevant behavior (e. g., focusing attention on task or
skimming through items). Physiological reactions include emotional arousal (e.g.,

increased heart rate, rapid breathing, or muscle tightness). Cognitions include worry that

interferes with the task (e.g., thoughts about social humiliation or the consequences of
failure). Each of these behavioral, cognitive, and physiological responses affect task
performance, and in turn, affect how the individual responds to future testing experiences
(Lowe et al.). Lowe and colleagues’ biopsychosocial model of test anxiety and Zeidner’s
transactional model of test anxiety have been combined and adapted to serve as the
conceptual framework for the present study (see Figure 1).
Prevalence Rates

Estimates on the prevalenCe rate of test anxiety among school-aged children have
varied widely (Zeidner, 1998). Traditional measures of test anxiety have not included
norms with specific cut-points for diagnosing clinical levels of test anxiety. Therefore,
researchers have used different criteria to define clinical levels of test anxiety, which may
explain the range in test anxiety prevalence estimates. King and Ollendick (1989) report
that the prevalence of test anxiety among school-aged children may range from as little as
10% to as much as 30%. However, they estimate that the rate of children who experience
clinically significant impairment is likely to be on the lower end of this spectrum (King &
Ollendick). Similarly, Hill and Wigfield (1984) suggest that between two and three
children in typical classrooms, or approximately ten percent of children, are highly test-
anxious and experience impairments in test performance as a result. More recently, some
researchers have used traditional scales and developed their own operational definitions
of high test anxiety in order to measure the prevalence rate of test anxiety among
children.

For example, Turner and colleagues (1993) conducted an analysis of test anxiety

among eight to twelve year-old African American school children. Children were

 

 

 

meow ..B 6 250A Ea 32 .coquN 88m 3E3“ xaoioeacm 333800 J oSmE

 

flouoem Eoomonofim

         

523 “mob =So>O

E3285
Eases;

   

:3 mo
nomumonsuom

 
   

8330m—
awe...

 

ooggofiom
o—mfl.

 

 

 

      
   

\

 

Engage E 082

 

sashes 25

 

became v.3.

 

8558820 Hoagxm

 

 

 

engaged“ magnum.

 

 

 

been oasesa.

 

comer—3x0 8cm

 

 

 

lO

 

considered to be test-anxious if their test anxiety scores fell in the upper third of a
distribution based on a normative sample obtained from the Pittsburg area
(Bednarek,1985). According to a study by Beidel and Turner (1988), this system of
classification results in the accurately classifying 92% of children with clinically
significant levels of test anxiety. Using this cutpoint, Turner and colleagues found that
the prevalence of test anxiety among Afi'ican American elementary school children may
be as high as 41%.

Turner and colleagues (1993) found that children’s mean scores on the Test
Anxiety Scale for Children (TASC) were significantly greater than the scores of children
who took the TASC three decades previously (mean=12.6 versus 7.0). The higher levels
of test anxiety identified by Turner and colleagues may be explained by a number of
biopsychosocial factors hypothesized to contribute to the development of test anxiety (see
Figure 1). First, the educational and social environrhent in the United States has shifted
since the construct of test anxiety was first studied in the 19503. Educational
accountability programs and the expectation that children will seek higher educations
have increased the relevance and importance of testing among young children. Second,
this shifiing environment has increased children’s experience with testing. Prior testing
experiences, and specifically prior experiences with failure or unpleasant arousal,
influence the cognitive attributions children make in future testing situations. The
increased pressure associated with heightened expectations for educational attainment
and the greater experience children have with test taking may all be associated with
children obtaining higher scores on the TASC.

Additionally, the racial background of the children in the study may have

 

influenced the rate of test anxiety found in the study. Stereotype threat is the fear that
one’s behavior will confirm an existing stereotype about the identity group that the
person identifies with. Steele and Aronson (1995) found that when race was salient to
intellectual test-takers, the performance of African Americans was impaired in
comparison to Caucasians. This finding suggests that if children of color perceive
educational testing situations as a measure of worth with respect to race, the result may
be increased anxiety and impaired performance among students of minority backgrounds.
However, the extent to which stereotype threat is related to performance-based anxiety
remains unclear (Ryan & Ryan, 2005). Further studies are needed to clarify the
relationship between race, stereotype threat, and test anxiety in young children in the
United States.
Demographic Patterns

Putwain (2007) recently examined demographic patterns in the expression of test
anxiety among 1348 14-16 year-old students in the United Kingdom. Putwain examined
students during their last two years of compulsory schooling because they take
examinations that greatly impact their educational and occupational futures during this
time. Students’ gender, socioeconomic status, ethnicity, and language of origin were all
related to the level of test anxiety experienced by students. First, females reported
significantly greater overall test anxiety, worry, and emotionality than males (Putwain).
This finding replicates previous studies that have found females consistently report
higher test anxiety than males. Hembree (1988) reported that gender differences in test
anxiety emerge around first grade and steadily increase before peaking in the middle and

high school years. Differences in test anxiety between females and males were moderate

in grades 5-10 (Effect size; ES = 0.43) and small in grades 1 and 2 (ES = 0.14), grades 3
and 4 (ES = 0.28), and grades 11, 12, and postseoondary (ES = 0.27; Hembree). The
question remains whether this difference may be attributed to differences in socialization
processes that result in females more openly disclosing anxious symptoms or whether
female students actually experience greater levels of anxiety (Wigfield & Eccles, 1989).
Second, Black, Asian, and ethnic minority students reported significantly greater
overall test anxiety, worry, and emotionality than Caucasian students (Putwain, 2007).
This finding is supported by the findings of Turner and colleagues (1993) and Wren and
Benson (2004) who found higher rates of test anxiety among African American
elementary school children, including more autonomic reactions and worry-related
thoughts. Alternatively, Beidel and colleagues (1994) found that rates of test anxiety did
not differ significantly by racial group, although African American children with test
anxiety had greater rates of social phobia diagnoses than Caucasian children with test
anxiety. In interpreting these different findings, is important to recognize that the cultural
and experiential differences between the samples from the United States and the United
Kingdom might affect the relationship between ethnicity and test anxiety (Bodas &
Ollendick, 2005). Similarly, the different ages of the children examined across these
studies highlights the. need for further examination of the relationship between ethnicity
and test anxiety across all school-aged children. Hembree (1988) found an interaction
between age and ethnicity in the expression of test anxiety, with significantly greater
rates of test anxiety among Black children in grades two to four (ES = 0.52) and grades
five to eight (ES = 0.21) than Whites, but no differences between older Black and White

high school students. Alternatively, Hispanic students in grades 4-12 reported

significantly greater test anxiety than White students (ES = 0.36; Hembree).

Finally, Putwain (2007) examined differences in test anxiety in relation to high-
stakes testing among students fi'om different socioeconomic backgrounds (SES),
including those with a head of household employed in routine/manual, intermediate,
managerial/professional, or not classified jobs such as retirement or fiJll-time study.
Students from routine/manual socioeconomic backgrounds reported significantly higher
overall test anxiety, worry, and emotionality than students from managerial/professional
backgrounds. Although students from intermediate and not classified backgrounds did
not report significantly different overall test anxiety, both reported significantly greater
worry than students from managerial/professional backgrounds (Putwain). Putwain’s
findings are consistent with Hembree (1988)’s meta-analytic finding that there is a small
but significant relationship between SES and test anxiety (ES = -.13). These findings
suggest that the socioeconomic status of students impacts the expression of test anxiety,
with particular differences found between low and high socioeconomic groups.
However, differences in cognitive responses to testing also exist between middle and
upper socioeconomic groups.

Other researchers have examined the relationship between age and test anxiety.
Rates of test anxiety appear to increase over the course of elementary school and become
more stable over the middle and high school years (Hembree, 1988, Hill & Sarason,
1966; McDonald, 2001; Sarason et al., 1960). According to a meta-analysis of test
anxiety differences between students in adjacent grades from 2 to 12, test anxiety
becomes established by second grade and rates increase steadily until fifth grade before

remaining relatively constant through the end of high school and then declining during

 

college (Hembree). However, the reported decline in test anxiety during the college years
may be the result of increased attrition among students with greater test anxiety
(Spielberger, 1966).

Increases in test anxiety over the school-age years may be related to children’s
increased exposure to testing over time and to the conditioning that occurs following
repeated concern about test failure, pressure to perform well, or feedback about poor
performance (King & Ollendick, 1989). Given the ever-increasing focus on using testing
performance to make high-stakes decisions, it is reasonable to hypothesize that the
overall prevalence and absolute level of test-anxiety will continue to increase among
young students over time (Hill & Wigfield, 1984). Students are now exposed to more
frequent assessments, assessments begin at a younger age, and assessments are directly
linked to rewards and sanctions. Thus, in this age of educational accountability,
prevalence rates and demographic patterns of test anxiety among young children need to
be reexamined.

Eflects on Performance and Educational Outcomes

The detrimental outcomes associated with test anxiety are extensive, including
impaired cognitive functioning, scholastic underperformance, and limited educational and
occupational outcomes (Zeidner, 1998). In comparison to low test-anxious individuals,
high test-anxious individuals experience significant impairment in evaluative testing
situations that arouse worry. In laboratory settings, a strong inverse relationship exists
between task performance and the amount of time spent worrying during the task
(Holroyd, Westbrook, Wolf, & Badhom, 1978). This relationship has also been found in

applied settings. For example, in a meta-analysis on 1112 high school and college

 

 

students, there was a significant correlation between worry and impaired performance on
aptitude and achievement assessments (r = -0.31) and on course grades (r = -0.26;
Hembree, 1988). Emotionality was also correlated to a lesser degree with significant
levels of impairment on aptitude and achievement assessments (r = -0.15) and on course
grades (r = -0. 19; Hembree). These findings reflect the relationship that exists between
cognitive and physiological aspects of test anxiety and test performance. Similar findings
by other researchers suggest high test-anxious students would be expected to score
significantly lower than low test-anxious students on real-world achievement tests (Seipp,
1991; Zeidner).

Over time, the relationship between test anxiety and test performance translates
into impaired educational and occupational outcomes for test-anxious students. A meta-
analysis of over 6000 students across all grade levels found that high versus low test
anxiety scores were related to significantly lower soores on intelligence (IQ),
achievement, and aptitude measures (ES = -0.48; Hembree, 1988). Although lower IQ
scores could be suggestive of lower ability in students, it is important to note that
intelligence assessments require testing, thus the lower scores may be more reflective of
impairment from test anxiety than lower innate ability among these students. For
example, Spielberger (1966) found that although students’ grade point averages (GPA) in
a single semester did not differ significantly, suggesting that high and low anxious
students had commensurate academic ability, more than 20% of the high-anxious college
students eventually dropped out of college as a result of academic failure, whereas fewer
than six percent of low anxious students experienced the same outcome. Spielberger’s

findings on students falling in the upper and lower 20% of the distribution on a measure

 

of manifest anxiety, suggest that there is a cumulative negative effect of anxiety on
academic performance. These studies are important because they suggest that test anxiety
is associated with multiple types of impairment in educational functioning and
attainment. Unfortunately, the majority of test anxiety research has been conducted on
college students, and relatively less is known about test anxiety in young children. This
pattern of research limits our understanding about the short- and long- term impact that
test anxiety has on young children.

Seymour Sarason led early research efforts in the field of test anxiety, and his
work on test anxiety among elementary school children has remained highly influential.
In 1964, Sarason, Hill, and Zimbardo conducted a seminal, longitudinal study on the
relationship between test anxiety and student performance on intellectual and
achievement tests. Sarason and colleagues (1964) reported on 713 students who were
followed from first and second grade until fourth and fifth grade, respectively. Their data
convincingly showed that test anxiety had a significant detrimental impact on students’
acquisition of knowledge and students’ utilization of knowledge on evaluative tasks
(Sarason et al.). Additionally, they found that children’s level of test anxiety increased
with age and that high test-anxious children experienced less growth over time in
academic achievement than low test-anxious children (Sarason et al.). For example, when
children were categorized into high and low test anxiety groups on the basis of their third
grade test anxiety scores, the high test-anxious group’s reading achievement scores were
eight to fourteen months behind the low test-anxious group’s scores.

When Sarason, Hill, and Zimbardo (1964) examined the relationship between

students’ verbal intelligence using the Lorge-Thorndike Intelligence Scale, achievement

 

.‘J -;l‘i I...

measures, and test anxiety, they found that the impairment high test-anxious children
showed on achievement measures was not fully explained away by differences in
intelligence. This finding strongly suggests that test anxiety impairs students’ initial
learning and/or utilization of skills, resulting in impaired academic performance.
Similarly, in a meta-analysis of early childhood studies, Hembree (1988) found that
second grade achievement in reading and English was significantly and inversely related
to test anxiety (r = -0.15), although there was no significant relationship between math
and test anxiety in grade two. In a much larger sample of older students fi'om grade three
to postsecondary school, both reading and mathematics achievement was significantly
and inversely related to test anxiety (r = -0.24 and r = -0.22, respectively). These
findings highlight the need for careful examination of how test anxiety affects students’
skill development and skill utilization during evaluative situations. Additional research
that considers student ability while also examining the impact of test anxiety on
performance is needed.

In their follow-up report, Hill and Sarason (1966) analyzed five years of
longitudinal data on 670 students in order to examine test anxiety stability and the
relationship between test anxiety and school achievement. While there was low stability
in test anxiety status from first to fifth grade (rs = 0.15 and 0.20 for boys and girls,
respectively), there was significantly greater stability in test anxiety status by late-
elementary school fi‘om fourth to sixth grade, (rs = 0.58 and 0.49 for boys and girls,
respectively; Hill & Sarason). Hill and Sarason also found that the absolute level of test
anxiety significantly increased among children from first to fifth grade. Similarly,

Hembree (1988) found that there was a significant relationship between test anxiety level

A

and grade level, with test anxiety levels increasing as students advanced in elementary
school from grade to grade (ES = -0.24). This pattern of increasing test anxiety in both
level and stability over the elementary school years suggests that the impairment
associated with test anxiety may similarly increase over time.

Further, Hill and Sarason (1966) found that the strength of the negative
correlation between test anxiety and achievement increased over the elementary school
years. Students’ academic achievement was significantly more impaired by test anxiety
as students advanced in elementary school. Even when high and low test-anxious
children were matched on the basis of intelligence scores, low test-anxious boys
performed significantly better on measures of overall achievement, reading, arithmetic-
comprehension, and arithmetic-problem-solving, and low test-anxious girls scored
significantly better on measures of reading achievement (Hill & Sarason). Children’s test
anxiety scores were also significantly related to their grades, with low test-anxious
children receiving better grades than high test-anxious children (Hill & Sarason). The
data showing that intelligence does not explain away differences between high and low
test-anxious children’s academic achievement strongly suggests that students with high
test anxiety experience significant impairment in their educational development over the
course of their education as a result of test anxiety. These data highlight the need for
further study of test anxiety among young children as well as examination of early
intervention techniques that may intercede in the developmental trajectory of test anxiety
in young children.

Educational Accountability Movement

In light of the established relationship between test anxiety and impaired

academic achievement, it is important to consider how the educational accountability
movement in the United States has resulted in an increased use and focus on test
performance in schools. The ongoing national focus on educational reform and
accountability developed in response to the 1983 publication of A Nation at Risk. In
response to concerns that students in the United States were lagging behind international
competitors in mathematics and science, the National Commission on Excellence in
Education (N CEE) provided a scathing condemnation of the then current educational
practices. The report found that the educational system in the United States was lacking
essential quality features and that students’ performance lagged behind other
industrialized countries in core content areas (NCEE, 1983). The NCEE concluded the
report with a series of recommendations for educational reform that were designed to
enhance student learning and performance. A Nation at Risk stimulated legislative action
that fundamentally shifted the role that the federal government took in overseeing
education in the United States. It marked the beginning of a national focus on standards-
based reform in education by calling for rigorous and measurable achievement standards
that could be monitored and evaluated.

Historically, the development, administration, implementation, and evaluation of
educational services have been duties performed by state and local educational agencies.
The federal govemment’s role in education was focused on ensuring that states complied
with federal procedural rules and regulations designed to provide students with
appropriate access to educational services. Following A Nation at Risk, however, the
federal government took an increasingly active role in developing performance-based

accountability policies to ensure that educational reform measures were implemented

20

(Wong & Nicotera, 2007). For example, in 1989, President George H. Bush and the
National Governor’s Association met for a summit on education and began to develop
educational performance standards. The work accomplished at this summit eventually
became the basis for the Goals 2000: Educate America Act, which delineated national
education goals for what students should know and be able to do (U .8. Congress, 1994a).
These goals reflected a shift in national educational philosophy from one that dictated
minimum competency standards to one that expected all students to meet high
educational standards when provided with appropriate opportunities to learn. Further,
Goals 2000 mandated national assessment practices to evaluate student achievement in
comparison to national and state performance goals.

AdditiOnally, the federal government increasingly mandated performance-based
educational reform measures through successive reauthorizations of the Elementary and
Secondary Education Act (ESEA). The 1994 reauthorization of ESEA known as the
Improving America’s Schools Act (IASA) focused on developing stronger accountability
measures for student achievement, including a specific focus on at-risk students through a
redesign of Title I (US. Congress, 1994b). IASA mandated the use of statewide
assessment programs to measure students’ annual yearly progress (AYP). Measuring
students’ AYP was designed to enhance academic and curriculum standards in schools by
focusing local and national attention on school performance. Most recently, when ESEA
was reauthorized in January 2002, the influence and importance of statewide academic
achievement testing programs were further increased by attaching financial and other

high-stakes sanctions and rewards to schools’ performance on the assessments.

21

No Child Lefi Behind Act of 2001

The No Child Left Behind Act of 2001 (NCLB), or the 2002 reauthorization of
ESEA, mandated the statewide administration of annual academic performance
assessments in order to hold public schools accountable for the education of school
children and to “close the achievement gap with accountability, flexibility, and choice, so
that no child is left behind” (US. Congress, 2002, p. 1). In order to achieve this goal,
NCLB sought to enhance educational services by promoting alignment between state
academic achievement standards and state academic assessments that are used to make
decisions about school funding, employment, and other major administrative functions.

The state assessment programs required by NCLB are specifically designed for
“identifying and turning around low-performing schools that have failed to provide a
high-quality education to their students” (US. Congress, 2002, p. 16). While NCLB did
not establish a national curriculum for schools, all states were mandated to develop
educational accountability plans. Additionally, states were obligated to develop a single
assessment system that measures the annual yearly progress of all students on the state’s
academic and curricular standards. States must “include sanctions and rewards, such as
bonuses and recognition” to hold local schools accountable for student achievement in
the areas of reading or language arts, mathematics, and science (US. Congress, 2002, p.
16). The requirement that test scores be publicly reported and linked to rewards and
sanctions such as school funding and employment decisions makes this testing high-
stakes in nature for educators and communities. Although individual students’ scores are
not linked directly to sanctions and rewards, the high stakes associated with this testing

for educators may result in teachers and schools placing increased pressure on students to

22

perform well on the test.

Reading and mathematics assessments must be administered annually to students
in grades 3-8 and once during grades 10-12. Science assessments must be administered at
least once during grades 3-5, grades 6-9, and grades 10-12. In addition, states are free to
assess other content areas such as social studies. Hence, NCLB requires students to take
no fewer than 17 assessments that are directly linked to high-stakes decisions over the
course of their elementary and secondary education. When students’ scores are reported
on these frequent assessments, scOres are disaggregated into groups on the basis of
students’ major social characteristics, including major racial and ethnic groups, economic
status, disability status, and limited English proficiency status. Decisions about schools
are made by examining student achievement data in total and in disaggregate. By
disaggregating the scores, the stakes for students who traditionally have been underserved
by the educational system are effectively increased. since schools may receive sanctions
for failing to meet annual yearly progress standards for any of the disaggregated groups.

Additionally, summary reports of disaggregated assessment data must be publicly
reported, and itemized scores are reported to parents, teachers, principals, and
administrators. Although major educational organizations including the National
Association of School Psychologists (N ASP) and the American Educational Research
Association (AERA) warn against using standardized assessments for individual
decision-making, the NCLB requirement that scores be reported individually for each
student implicitly encourages educators and parents to use the data to make individual
academic decisions about students (AERA, 2000; NASP, 2003). This reporting

requirement increases the testing stakes for students, who may be evaluated by parents

23

and teachers on the basis of their state assessment performance.
Michigan Educational Assessment Program

In the state of Michigan, the Michigan Educational Assessment Program (MEAP)
has been adapted to comply with the requirements of NCLB. The MEAP was initially
developed in 1969 to determine what students across the state knew and could do in
relation to the educational standards set by the state. The MEAP was traditionally
administered at key points in students’ academic careers. As a measure of students’ skills
in the core content areas of mathematics, reading, science, social studies, and writing, the
MEAP was designed to be an accountability measure for use in school improvement by
indicating curriculum and instructional needs at schools. The MEAP was specifically
designed to be interpreted in relation to other achievement measures and instructional
variables (State of Michigan, 2004a).

In its current form, the MEAP is a criterion-referenced test that is administered
during the fall of each academic year. Testing occurs over a period of two to three weeks
and it requires approximately eight hours to complete the assessment in English
Language Arts, Mathematics, and Science. MEAP scores are used as the key indicator of
student achievement in Michigan. In order to meet the Adequate Yearly Progress (AYP)
requirements of NCLB, student groups, schools, and districts are evaluated on their
MEAP performance in relation to annual performance goals. Scores are reported at the
individual student level through reports sent to parents and maintained in student files, at
the classroom level through class roster reports, at the school level through summary
reports disaggregated by student groups, and at the district and state levels. Student

performance in each content area is classified as advanced, proficient, partially proficient,

24

or not proficient, and shifts in performance levels from year to year are also monitored
and reported (Griffiths & Peterman, 2007).

Additionally, school and district level scores are posted on the State’s Department
of Education website and are also publicly reported on through the news media (e.g.,
newspaper, television, and intemet). The public reporting of MEAP scores at the school
and district levels has important implications for communities, including, but not limited
to influencing real estate values and decisions families make about where to live.
Similarly, in some cases, the individual student performance data that is reported to
schools and parents directly influences educational decisions made on behalf of children,
such as grade retention and special education referral decisions, (e.g., Fielding, 2004).

Although the Michigan Department of Education cautions against using the
MEAP as a sole measure of educational achievement, in accordance with the federal No
Child Left Behind Act, high-stakes outcomes are linked directly to MEAP scores in
Michigan. Sanctions associated with not meeting AYP standards increase with each
successive year of failure. For example, following two successive years of failure, Title I
schools must notify parents of their status, must offer school choice and pay for
transportation costs totaling up to twenty percent of Title I funds, must develop school
improvement plans, and must use ten percent of Title I funds for professional
development. Following three years of failure, schools must offer students
supplementary educational services in addition to the aforementioned sanctions.
Following four years of failure, schools must take corrective action such as staff
replacement, school restructuring, or new school management. Thus, the sanctions

associated with not meeting AYP significantly affect students, educators, and

25

communities; and the sanctions increase in magnitude with each successive year of
failure (State of Michigan, 2004b).

In sum, the educational accountability movement has resulted in an increased
emphasis on using performance-based measures to evaluate and make decisions about
students and schools. The MEAP has a significant impact on students, families, schools,
and communities. In terms of immediate impacts, students and schools must devote a
significant amount of time, energy, and funding to completing the MEAP. Following
administration of the MEAP, high-stakes consequences are associated with students’
performance on the MEAP. These consequences exist at individual, family, school, and
community levels, and by influencing decisions such as individual students’ educational
placements, family decisions about what communities to live in and send children to
school in, and school decisions about teacher employment, administrative structuring,
and fimding allotments.

Test Anxiety and High-Stakes Assessment

In 1984, Hill and Wigfield predicted that the trend towards using test scores to

evaluate educational programs and to make decisions about individual students would

result in increasing rates of test anxiety among students. The authors argued,

test scores have a strong influence in determining children’s school
progress and what they will be able to do later in life. . .. Because of
testing’s importance, we believe it is essential that all student’s test scores
accurately reflect what they do know and not be negatively influenced by
test anxiety and other test-taking factors (Hill & Wigfield, 1984, p. 122).

The federal mandate to use standardized, high-stakes testing programs to evaluate schools

26

and students rests on the assumption that students’ test scores accurately reflect students’
knowledge and skills. However, as Hill and Wigfield (1984) suggest, this assumption
must be demonstrated empirically since other extraneous factors, including test anxiety,
may differentially affect students’ performance on these assessments.

In light of the evidence that test anxiety is associated with impaired learning and
use of knowledge in evaluative situations, it follows that test anxiety may similarly
impact student performance on the standardized tests of academic achievement required
by NCLB (Cizek, 2001; Meijer, 2001). The following sections critically review the
existing literature on the impact of high-stakes assessments on children. Studies are
organized by methodology, with the few studies directly examining students’ perceptions
of assessments reviewed first and indirect studies examining the perceptions that
teachers, parents, and other school personnel have about the impact of testing on students
presented second. Gaps in the literature are also examined in the context of presenting a
rationale for further study of test anxiety and its relationship to academic achievement
assessments mandated by NCLB.

Direct Studies on the Impact of Testing on Children

Only four studies have directly examined children’s perceptions of achievement
assessments mandated by NCLB. These studies offer important insights into how
students experience these testing programs. Triplett and Barksdale (2005) examined 225
third through sixth graders’ drawings and written responses to probes that asked them to
first draw pictures about their testing experience and then describe their pictures through
writing. The authors concluded that the children overwhelmingly experienced negative

emotions, including nervousness and anger, in relation to testing. Students also expressed

27

feelings of isolation from their peers and teachers. In contrast to their hypothesis that the
high-stakes assessment experience would be “basically meaningless” (p. 257), testing
was a pervasively negative experience for students and the students exhibited an
awareness of and concern about the high-stakes associated with the assessment. The
authors suggest that the negative valance of the children’s creative products might stem
from test anxiety; however, the study did not directly measure test anxiety and the study
methodology required the researchers to interpret the meaning of children’s drawings and
writing samples, introducing the possibility of inaccurate interpretations. Therefore, while
this study provides interesting initial data suggesting that children experience significant
stress during testing, it needs to be followed up with a more direct examination of the
relationship between high-stakes assessment and test anxiety.

In contrast to the findings of Triplett and Barksdale, two studies conducted by
Mulvenon and colleagues (2001; 2005) suggest that the prevalence and impact of test
anxiety among students taking high-stakes assessments is minimal. In their initial study,
Mulvenon, Connors, and Lenares (2001) surveyed 283 fifth grade students who hadjust
completed state mandated Stanford Achievement Test- Ninth Edition (SAT-9) testing
about their attitudes towards testing. Seventy-five percent of students reported having
positive attitudes toward testing, and they did not exhibit stress or impaired performance.
Although there was a significant relationship between reported pressure and lower scores
on the SAT-9, pressure accounted for less than two percent of the variance in SAT-9
scores, suggesting that pressure was not a clinically significant predictor of test
performance.

In a later report, Mulvenon, Stegrnan, and Ritter (2005) expanded on their initial

28

report and examined the perceptions that teachers, principals, counselors, students, and
parents had about the relationship between high-stakes testing and student performance.
The authors reported that the pressure students perceive from parents and teachers, and
not test anxiety, was significantly related to poorer test performance. Additionally, the
authors found that teachers’ anxiety level was not significantly related to student
performance, which contrasts concerns raised by Triplett and Barksdale (2005) about the
negative impact of teachers’ anxious actions and statements. These studies suggest that
concern about test anxiety in relation to high-stakes testing may be unnecessary.
However, the studies conducted by Mulvenon and colleagues (2001; 2005) had a
number of methodological weaknesses that call their findings into question. First, the
studies used a 24-item investigator-developed survey to measure six constructs, including
test anxiety, rewards, pressure, attitude toward testing, and self-efficacy in reading and
math. While there were eleven questions about test anxiety, this subscale’s concurrent
validity was not established with established measures of test anxiety, and all of the other
constructs were measured using limited items ranging from one to three questions.
Additionally, when examining the relationship between test anxiety and test performance,
the authors controlled for student ability, which is important for isolating the effect of test
anxiety on test performance. However, the authors controlled for student ability using
students’ scores on a criterion-referenced examination from the previous spring. Since the
SAT-9 and the spring exam were both standardized, group-administered assessments,
students may have also perceived the spring assessment to be anxiety-provoking, which
would invalidate the use of this assessment as a measure of ability unaffected by test

anxiety. These studies findings need to be replicated using an empirically validated

29

measure of test anxiety and a more representative measure of student ability.

Most recently, Putwain (2008) examined whether the stakes of an examination
moderated the relationship between test anxiety and performance among 615 teenagers in
the United Kingdom. Students were split into three groups of those taking a low-stakes,
mock examination, those taking a mid-stakes, modular examination that is one of a series
of tests taken over the final 18 months of schooling, and those taking a high-stakes,
terminal examination. Putwain found that contrary to prediction, students taking the
mock exam reported significantly greater test anxiety than students taking either the
modular or terminal exam. This finding does not support the hypothesis that higher stakes
are associated with a greater degree of perceived threat by children. However, Putwain
argued that students taking the mock exam were experiencing their first exposure to this
type of testing, whereas students taking the mid- and high- stakes exams had previous
experience with testing (see Figure 1). Hence, the greater level of test anxiety found in
the low-stakes group may have been an artifact of this study design.

Putwain (2008) also found that there were small but significant negative
relationships between test anxiety, its components, and performance on all three exam
types. The correlations between examination grade and the various measures of test
anxiety including test anxiety, worry, and emotionality were rs = -O.19, -O.18, -O.14,
respectively (Putwain). When students were dichotomized into low- and high-test
anxious groups using test anxiety scores i 1 standard deviation from the sample mean,
high test-anxious students performed significantly lower than their low test-anxious peers
on the mock and modular examinations. Alternatively, there was a small facilitating

relationship between high test anxiety and examination performance on the terminal

30

examination. Putwain cautioned against the over-interpretation of these findings and
called for replication of this study with a more robust methodology. In sum, this study
raises important questions about the relationship between assessment stakes, test anxiety,
and student performance that merit further study.
Indirect Studies of the Impact of Testing on Children

Jones and colleagues (1999) interviewed 236 teachers about their perceptions of
the impact of high-stakes testing on students, themselves, and their instructional
practices. Teachers were from North Carolina where a state accountability program that
linked assessment results to employment decisions and financial incentives for teachers
had been recently implemented. Teachers reported that instructional patterns had changed
as a result of the accountability program, with more than 70% of teachers reporting
spending more time practicing for end-of-grade tests and 80% of teachers reporting that
they spent at least 20% of instructional time practicing for tests. Teachers also reported
that these instructional changes and the testing program resulted in negative outcomes for
children, with 61% of teachers reporting students felt more anxiety and 24% of teachers
reporting that students were less confident. These reported impacts on children must be
interpreted with considerable caution, however, since the authors did not directly assess
how the assessment program affected students.

Another study of 708 teachers and 325 administrators in Florida found that
teachers have greater concerns about the negative impact of high-stakes assessments on
instructional practices and on students than administrators (Jones & Egley, 2006). While
46% of teachers reported that the testing program had negative effects on student and

teacher motivation, only 27% of administrators reported this same belief. When these

31

same teachers responded to an open-ended question asking them to describe how the state
assessment was impacting Florida in a positive or negative way, Jones and Egley (2004)
described teachers’ responses as being overwhelmingly negative, with teachers providing
89% negative reasons and only 8% positive reasons. Twenty-five percent of responses
indicated that the testing program placed too much pressure on teachers and students,
resulting in stress, anxiety, worry, or fear. However, the survey required teachers to
decide if the state program took Florida in a positive or negative direction, leading
teachers to focus on either positive or negative outcomes rather than providing a more
complete description of both positive and negative outcomes of the program.

Similarly, Barksdale-Ladd and Thomas (2000) reported that teachers and parents
feel that high-stakes assessments have negative impacts on students and teachers. In
interviews with 59 teachers and 20 parents in two states with high-stakes assessment
programs tied to monetary consequences for schools and grade retention for students,
parents and teachers reported that the testing programs resulted in instructional changes
and student and teacher stress and anxiety. Teachers described feeling pressure to ensure
that students perform well, with some estimating that two to three months of the school
year was focused on preparing for the test. Teachers described concern that their own
stress and anxiety about the assessment program was transferred to students and that the
children worried about testing and its consequences. Similarly, although parents reported
more positive beliefs about the uses of testing, they reported that children experienced
high levels of anxiety and nervousness prior to the tests and they expressed
disappointment afier learning how they scored. These four reports suggest that teachers

have a number of concerns about the negative impact of testing on students and teachers,

32

and they emphasize the need to study the impact of testing on teachers and students
directly.

Additionally, in a national study of teachers’ opinions about state testing
programs, Abrams, Pedulla, and Madaus (2003) compared testing programs with high
versus low stakes, with high-stakes being defined as tests linked to significant
consequences such as accreditation, financial rewards, or school autonomy. Teachers
reported that the stakes associated with testing programs significantly impacted
instructional practices and student stress. For example, 43% of teachers in high-stakes
testing states, as opposed to only 17% in low-stakes states reported that time spent on
instruction in tested areas increased a great deal, with decreased time spent in class
enrichment or extracurricular activities. Forty-one percent of teachers in hi gh-stakes
states versus 18% in low-stakes states reported that there is so much pressure for students
to score highly on the tests that they have little time to teach anything not on the test. In
addition, 80% of teachers in high-stakes states perceived that students were under intense
pressure to perform well, with 35% reporting that students were extremely anxious about
the state test. In low-stakes states, 49% of teachers perceived that students were under
intense pressure to perform well, with 20% reporting that students were extremely
anxious about the state test. Therefore, the stakes associated with state testing programs
appear to be directly related to the impact that testing programs have on students and
schools.

Alternatively, other reports suggest that educational stakeholders such as school
counselors and parents do not perceive high-stakes testing to have pervasively negative

impacts on students. Brown, Galassi, and Akos (2004) surveyed 280 North Carolinian

33

school counselors and found that while 65% of school counselors reported concern that
testing resulted in student stress, anxiety, fear, or pressure, only 4% reported that students
experienced lowered self-confidence or self-esteem and only 2% reported that students
experienced feelings of failure or incompetence due to testing. Also, only 24% of school
counselors reported that the program had no positive student outcomes, with the others
reporting positive outcomes in the form of increased student focus on schoolwork or in
increased accountability. These results suggest that while anxiety or stress may be present
during the testing situation, school counselors perceive this distress as transitory and as
not having lasting impacts on students.

Similarly, Mulvenon, Stegman, and Ritter (2005) reported that parents were
generally supportive of standardized assessment programs. Only 24% of parents reported
that testing was significantly or extremely stressful for their children and only 13%
reported that they felt significant or extreme pressure to help their children score well,
suggesting that in general parents do not project high levels of stress to their children
about the testing process. Therefore, these two studies suggest that parents and school
counselors may have more positive views of testing and fewer concerns that testing has
negative psychological consequences for students than teachers. In turn, the ways that
different adults perceive testing and communicate with students about testing may affect
how students themselves respond to testing.

In sum, these indirect and direct reports by students, teachers, parents,
administrators, and school counselors provide the basis for further study of the impact of
testing programs on students. A number of studies suggest that state testing programs

result in heightened student anxiety, stress, and lowered motivation (Abrams et al., 2003;

34

Barksdale-Ladd & Thomas, 2000; Jones & Egley, 2004; 2006; Jones et al., 1999; Triplett
& Barksdale, 2005). Alternatively, Mulvenon and colleagues (2001; 2005) found that
most students felt positively about testing, did not experience stress as a result of testing,
and their performance was not impaired by test anxiety. These studies provide important
initial data about how various stakeholders perceive the impact of testing on students;
however, there is a clear need to for additional studies that directly assess the impact of
testing programs on children in light of the contradictory findings in the literature and the
methodological weaknesses of previous studies.
Current Study

As illustrated by the literature review above, relatively little research directly
examining test anxiety in children taking high-stakes tests has been conducted and little is
known about the effect of test anxiety on student performance on high-stakes
assessments. NCLB legislation has placed unprecedented weight on using students’ test
scores on statewide achievement assessments to evaluate and make decisions about
public schools. Given meta-analytic data indicating test anxiety inversely affects student
performance on achievement assessments, impairs academic performance in school, and
results in lower educational attainment (Hembree, 1988), it is essential to examine the
relationship between test anxiety and high-stakes testing. The existing literature
predominantly used indirect methodology to study the impact of testing on students and
there is a need for studies that directly examine the impact of testing on students. This
dissertation study critically examined the impact that test anxiety has on students taking
high-stakes achievement assessments and it filled gaps in the research literature on this

topic. This goal was achieved by directly measuring students’ test anxiety in relation to

35

high-stakes testing and classroom testing and examining the relationship between test
anxiety, psychosocial variables, and test performance.

The subjects in this study were comprised of school children taking state-
mandated achievement assessments required by NCLB. Young students, who have
traditionally not been subject to standardized testing, have been most impacted by the
annual testing requirements of NCLB. Therefore, students in elementary schools were the
focus of this study in order to add to the limited literature on the expression and the
impact of test anxiety among young children. Figure 2 presents the current study’s
methodological organization. The current study assessed the key biopsychosocial factors
contained in this study’s conceptual framework by including the measurement of overall
test anxiety, worry, emotionality, and test behavior; academic achievement; test
performance; and multiple psychosocial factors. Using this methodology, this
dissertation study significantly expands upon the current test anxiety literature by
critically examining how young children respond to high-stakes achievement testing.

While additional person-specific variables, including prior testing experience, trait
anxiety, achievement need, social-emotional functioning, self-efficacy, social skills, and
perception of the task, and testing/environmental variables, including testing atmosphere,
examiner characteristics, task difficulty, time constraints, and mode of administration, are
hypothesized to affect student test anxiety (see Figure 1), these factors were not included
in the current study. The additional person-specific variables were not included due to
methodological constraints (e.g., time and financial). The testing/environmental variables
were not included in the study since the mandated achievement testing administration is

standardized, using identical assessments, instructions, and administration dates.

36

 

   
  

   
 

  
  

Test
. Behavior

Cognition/
Worry

 

       
 
  

Test
performance

  
  

Academic
abilityI

  
 

Physiology/
Emotionality

 

 

 

 
   
 

Overall Test Anxiety

Psychosocial
Influences

 

 

Figure 2: Methodological Design of the Current Study

37

 

Therefore, variability in testing/environmental variables is minimized in this study.
Research Questions and Hypotheses

Question One. Are there differences in test anxiety levels and rates associated
with high-stakes testing and classroom testing? The first research question examined the
differences in the level of reported test anxiety experienced by students taking state-
mandated educational assessments as well as classroom tests. Additionally, differences in
rates of low, moderate, and high test anxiety across the test conditions were examined.
Previous test anxiety prevalence estimates have ranged from as low as 10 to as high as
41% (King & Ollendick, 1989; Turner et al., 1993). Although these rates have been
determined using different operational criteria, the rates all include students who respond
to testing situations with heightened anxiety and worry. Studies indicate that when
students perceive assessments to have high personal relevance or when assessments are
presented as measures of ability, they experience greater anxiety (Zeidner, 1998). The No
Child Left Behind Act of 2001 has increased the frequency and the stakes of state
educational assessments by linking essential funding, administrative policies, and
decisions to student performance data (U.S. Congress, 2002). This federal mandate
increased the direct relevance that state assessments have for educators, parents,
communities, and schools. In turn, the high-stakes associated with state assessments may
increase the prevalence of test anxiety in children. The researchers hypothesized that test
anxiety levels among students would be significantly greater for high-stakes assessments
than classroom testing. Additionally, the differences in rates of low, moderate, and high
test anxiety across testing conditions were explored using different classification systems

for test anxiety. The researchers hypothesized that rates of high test anxiety would be

38

significantly greater for high-stakes assessments than classroom testing.

Question Two. Are there differences in classroom teachers ’ perceptions of
students’ test anxiety responses and performance in relation to high-stakes testing and
classroom testing? Are there differences in classroom teachers’ anxious responses to
high-stakes testing and classroom testing? Previous studies suggest that teachers believe
high-stakes testing programs result in heightened student and teacher anxiety and stress
(Abrams et al., 2003; Barksdale-Ladd & Thomas, 2000; Jones & Egley, 2004; 2006;
Jones et al., 1999). However, differences in teacher perceptions across different test
conditions have not been previously examined. The second research question examined
the differences in teachers’ perceptions of both their own and students’ anxiety responses
to testing and associated test performance across testing conditions. The researchers
hypothesized that teachers would perceive students to have more test anxiety in relation
to high-stakes assessments than classroom testing. Similarly, the researchers
hypothesized that teachers would report more anxiety themselves about hi gh-stakes
assessments than classroom testing. Finally, the researchers hypothesized that teachers
would report that student performance was more impaired on high-stakes assessments
than classroom tests.

Question Three. Do students with different test anxiety classifications perform
differently on high-stakes testing? Does test anxiety predict testing performance? Are
there other predictors of test performance? Perhaps most central to the argument behind
NCLB is the assumption that student test scores reliably and accurately represent
students’ educational experiences and knowledge. The use of assessments to make

decisions about school funding, employment, and curriculum is based on the assumption

39

that the tests are indicative of the quality of the educational services being provided to
students. However, research suggests that student performance on standardized
assessments is related to student test anxiety (Hembree, 1988; Zeidner, 1998). The
researchers hypothesized that test performance would differ on the basis of students’ test
anxiety level. Additionally, the researchers hypothesized that even after controlling for
student achievement using academic indicators and student demographic characteristics,
test anxiety would be a significant predictor of performance on the state assessment.
Student achievement needs to be controlled for in order to account for pre-existing
individual differences among students that also affect test scores. The test anxiety
literature also suggests that impaired reading scores are more strongly associated with test
anxiety than other academic subject areas (Sarason et al., 1960), leading the researchers
to hypothesize that test anxiety would be more strongly related to impaired reading scores
than mathematics scores. If test anxiety is found to be significantly related to test
outcomes on the high-stakes assessment, this relationship may call into question the
validity of making high-stakes decisions based on a sole indicator of student
performance.

Question Four. D0 students with diflerent demographic characteristics report
significant different levels test anxiety? How predictive are student demographic
characteristics of test anxiety across the testing conditions? The test anxiety literature
suggests that numerous demographic variables are associated with test anxiety.
However, these variables have yet to be examined in relation to high-stakes tests. This
research question examined whether or not, and to what degree, demographic

characteristics predicted test anxiety for both classroom testing situations and high-stakes

40

testing situations. Since NCLB requires that student performance data be reported in
groups that are disaggregated by social characteristics, including race, economic status,
and disability status, among others, it is essential to examine if these variables are
associated with test anxiety. As a result of student data being reported separately each of
these student subgroups, greater pressure may be placed on these students to perform well
on assessments, resulting in greater student anxiety and worry (Steele & Aronson, 1995).
If certain groups of children respond to the assessment in a way that hinders their ability
to perform, the use of their scores to make high-stakes decisions needs to be critically
reexamined. Based on theory and previous findings, the researchers in the current study
hypothesized that students receiving special education services, female students, older
students, racial minority students, and economically disadvantaged students would
experience greater test anxiety than other students on the high-stakes assessment
(Hembree, 1988; Putwain, 2007). Additionally, the researchers in the current study
hypothesized that female students, older students, racial minority students, and
economically disadvantaged students would report greater levels of test anxiety on
classroom tests (Hembree).

Question Five. What is the strength of the relationship between the Children’s
Test Anxiety Scale, the Behavior Assessment Scale for Children-2’"! Edition-T est Anxiety
Subscale, and the Test Anxiety Scale for Children within testing conditions? How closely
do classroom testing test anxiety scores relate to high-stakes assessment test anxiety
across measures? The most commonly used measure of test anxiety among children is
the Test Anxiety Scale for Children (TASC, Sarason et al., 1960). It has been used

extensively in studying test anxiety among young school children. However, questions

41

about the developmental appropriateness of the scale have been raised because the items
are wordy and include complex sentence structures. Although designed for children as
young as first grade, the scale’s linguistic complexity raises the question of whether or
not children are able to fully understand the questions. Therefore, it is important to
examine more contemporary scales of test anxiety to determine their construct validity in
relation to this established scale. The Children’s Test Anxiety Scale (CTAS; Wren &
Benson, 2004) and the Behavior Assessment Scale for Children, Second Edition, Test
Anxiety subscale (BASC-Z-TA; Reynolds & Kamphaus, 2004) are both recently
developed measures of test anxiety. The relationship between the CTAS, BASC-2-TA,
and TASC scales were examined in order to provide preliminary data about the newer
scales’ construct validity. On the basis of the scales’ similar theoretical constructs, the
researchers hypothesized that student’s scores on the three scales would correlate
moderately to highly with one another (rs 2 0.60). Additionally, the researchers
hypothesized that student test anxiety in relation to classroom testing would be strongly

correlated with student test anxiety in relation to high-stakes testing across all measures.

42

CHAPTER 3
METHOD

Subjects

School children in grades three through five in three elementary schools from a
suburban school district in the state of Michigan were invited to participate in the study.
This age population was selected based on the design of No Child Left Behind (N CLB)
testing programs which require annual achievement testing beginning in grade three as
well as data suggesting that test anxiety stability becomes significantly greater by late-
elementary school (Hill & Sarason, 1966). All students taking the Michigan Educational
Achievement Program (MEAP) were eligible for the study. Students who took the
alternative assessment were not included; NCLB allows up to one percent of students to
take an alternative assessment due to severe cognitive disability. Decisions about the
alternative assessment are made by a student’s Individualized Education Program Team
when, due to severe cognitive impairment, the MEAP, even with accommodations, is
deemed to be an inappropriate assessment.

A total of 617 children from 25 classrooms were invited to participate in the
study. Parental consent forms permitting children to participate in the study were
returned for 363 children. Among the three participating schools, 121 children received
parental consent to participate in school A (54% of invited), 65 in school B (48% of
invited), and 177 in school C (69% of invited). This overall 58% rate of student
participation is excellent for survey research (Hopkins & Gullickson, 1992).

Following survey administrations, the total number of students who completed at
least one survey was 335 students (92% of those with consent for an overall participation

rate of 54% among all invited students). Children who did not participate in at least one

43

of the surveys were dropped from the study. One hundred and nineteen children
completed surveys in school A (98% of those with consent). 50 in school B (78% of
those with consent), and 166 in school C (94% of those with consent). Differences in
participation rates across schools were related to student absences from school and
student participation in alternative activities at the time of survey administration. Among
the 25 teachers invited to complete teacher questionnaires, 88% participated in the first
questionnaire and 60% participated in the second questionnaire. Differences in teacher
completion of questionnaires may have been related to the increased length of the second
questionnaire.

Table 1 summarizes the demographic characteristics of student participants. In
some cases, the sample N does not tetal to 335, indicating that some demographic
information was unavailable for participants. There were 151 males and 180 female
students who ranged in age from 7 to 11 (M = 9.10 years, SD = 0.92), with 10.1%
receiving special education services. On average, parents completed between one and
four years of college and were from the middle class (SES: M = 2.10, SD = 0.86). The
racial/ethnic makeup of the overall sample was 82.1% Caucasian, 4.5% African
American, 7.2%, Hispanic or Latino, 3.0% Asian, and 2.4% Other.

The demographic characteristics of the total sample were examined to determine
if participants in the three schools were representative of the actual schools’ populations
and if the samples differed significantly across schools. Differences in demographic
characteristics between the overall population in the three schools and the overall sample
were examined with Chi-square tests, using a Bonferonni correction of 0.125 for four

comparisons. Chi-square goodness of fit tests indicated that there were no significant

44

Table 1

Demographic Characteristics of Student Participants

 

 

 

 

Sample Population
Demographic Characteristic % N %
Grade 1.41 .49
3 119 35.5 203 32.9
4 124 37.0 229 37.1
5 92 27.5 185 30.0
Gender 5.85 .02
Male 151 45.1 682 52.3
Female 180 53.7 623 47.7
Race 6.71 .08
Caucasian 275 82.1 1098 84.1
African American 15 4.5 81 6.2
Latino/Hispanic 24 7.2 63 4.8
Asian 10 3.0 55 4.2
Other 8 2.4
Educational Verification 3.15 .08
General Education 289 86.3 1 123 86.1
Special Education 34 10.1 182 13 .9
Total N 335 1305

 

45

differences between the sample by grade, sex, ethnicity, or educational verification and
the overall population of the three schools.

Additionally, differences in demographic characteristics between the overall
school populations and the school samples were examined using Chi-square tests, using a
Bonferonni correction of 0.17 for three comparisons. Chi-square goodness of fit tests
indicated that there were no differences between the racial/ethnic or gender makeup of
study samples from schools A, B, and C and the actual population of those schools.
Similarly, there was no difference between the rate of students receiving special
education services in the current study sample from schools A or B and the population of
schools A and B. In school C, significantly fewer students receiving special education

services were included in the current sample (10%) as compared to the school population
proportion of 18%, 12 (1, n = 165) = 6.18,p = .01.

Table 2 summarizes the demographic characteristics of student participants for
each of the participating schools. Differences in demographic characteristics across the
samples from the three schools were also examined using analyses of variance and Chi-
square tests, with a Bonferonni correction of p S 0.008 for six comparisons. A one-way
between-groups analysis of variance was conducted to explore whether or not there were
differences in parental education by school. There was a significant difference in parental
education levels for the three schools: F (2, 330) = 6.70, p = .001. Parents in school B
reported less education than parents in schools A and C. A one-way between-groups
analysis of variance was also conducted to explore whether or not there were differences
in family socioeconomic status by school. There was a significant difference in

socioeconomic status levels for the three schools: F (2, 305) = 5.29, p = .006. Parents in

46

Table 2

Demographic Characteristics of Student Participants by School

 

 

 

 

 

School A School B School C
Characteristic M SD M SD M SD F p
Parent Educationa 5.53 .77 5.16 .74 5.63 .83 6.70 .001
SES Codea 2.06 .82 2.46 .80 2.01 .88 5.29 .006
N % N % N % 12 p
Grade 1.61 .81
3 43 36.4 20 39.2 56 33.7
4 44 37.3 20 39.2 60 36.1
5 31 26.3 11 21.6 50 30.1
Gender .35 .84
Male 51 44.0 24 49.0 76 45.8
Female 65 56.0 25 51.0 90 54.2
Race 5.37 .07
Caucasian 96 82.1 47 94.0 132 80.0
Non-Caucasian 21 17.9 3 6.0 33 20.0
Educational Verification 1.33 .52
General Education 104 91 .2 40 85. 1 145 89.5
Special Education 10 8.8 7 14.9 17 10.5

 

a’Parental Education: 2: 9th grade or less, 3: 10th or 11th grade , 4: High School Graduate,

5: Partial College , 6: College Graduation, 7: Graduate Training; SES: Hollingshead

Socioeconomic Codes: 1: High Class, 2: Middle Class, 3: Working Class, 4 and 5: Low

Class

47

school B reported lower socioeconomic status than parents in schools A and C. Chi-

square tests for independence indicated no significant associations between school and

grade: 12 (4, n = 335) = 1.61,p = .81, school and gender: 12 (2, n = 331) = .35,p = .84,
school and race: 12 (2, n = 332) = 5.37, p = .07, and school and educational status: 12 (2,

n = 323) = 1.33, p = .52. Based on these analyses, which indicate that with the exception
of socioeconomic status and parental education, participants’ demographic
characteristics did not differ significantly across schools, and the analyses indicating that
the samples from each school were representative of the schools’ actual populations, the
data across all schools were combined for all subsequent data analyses.
Measures

Two survey packets were developed for use in this study. The first survey was
made up of the Children’s Test Anxiety Scale (CTAS) and the Behavior Assessment
Scale for Children, Second Edition, Test Anxiety subscale (BASC-2-TA). The second
survey contained the original measures and also included the Test Anxiety Scale for
Children. The following sections describe the test anxiety measures contained in the
survey packets:
Children ’s Test Anxiety Scale

The Children’s Test Anxiety Scale (CTAS) is a 30-item self-report scale designed
to measure test anxiety in children in grades three through six (Wren & Benson, 2004).
The scale includes items that assess children’s thoughts, off-task behaviors, and
physiological reactions to testing. Each item is scored on a four-point Likert scale that
ranges from almost never to almost always. Scores on the CTAS range from 30 to 120.

The scale was constructed and validated on a sample of 261 children. Internal consistency

48

estimates for the CTAS were satisfactory for the overall scale (0.92), and for off-task
behavior (0.76), physiological reactions (0.82), and thoughts (0.89) subscales.
Administration time ranged from 5 to 12 minutes. Means and standard deviations for the
total sample were 61.97 (16.49) on the overall scale, 16.89 (5.14) on the off-task behavior
subscale, 15.96 (5.63) on the physiological reactions subscale, and 29.12 (8.79) on the
thoughts subscale. Girls reported significantly greater overall test anxiety than boys, with
mean scores of 65.78 (16.93) versus 57.71 (14.94). No classification system for clinical
levels of test anxiety on the CTAS has been developed, although the test developer
recommends classifying scores falling one standard deviation above the mean of the
standardization sample as high test-anxious and scores falling one standard deviation
below the mean as low test-anxious (Wren, personal communication, March 14, 2008).
Behavior Assessment Scale for Children, Second Edition, Test Anxiety Subscale

The test anxiety content subscale of the Behavior Assessment Scale for Children,
Second Edition (BASC-2-TA) is made up of seven items on the self-report of personality
(Reynolds & Kamphaus, 2004). The subscale measures a child’s propensity for irrational
worry and fear regarding test-taking. Scores on the subscale range from 0 to 18. The scale
has been normed for children aged 12 to 18. In a sample of 884 12-14 year olds in the
normative sample, the test anxiety subscale had an internal consistency alpha of 0.81.
Means and standard deviations for the sample were 6.74 (3.67) when the sample was
combined, 7.52 (3.61) for females, and 5.95 (3.57) for males. The BASC-2 classifies
scores falling one standard deviation above the mean of the standardization sample as At-
Risk and scores falling two standard deviations above the mean as Clinically Significant.

No normative data for children aged 7 to 11 is available for the BASC-2-TA.

49

Test Anxiety Scale for Children

The Test Anxiety Scale for Children (TASC) is a 30-item self-report scale for
children in grades one or higher (Sarason et al., 1960). The scale is designed for either
researcher or teacher administration, with no differences in outcomes observed between
these two conditions (Sarason eta1.). The scale’s construct validity has been
demonstrated through studies demonstrating a significant correlation between the TASC
and teacher ratings of student anxiety as well as consistently positive correlations
(ranging from 0.47 to 0.69) found between the TASC and the General Anxiety Scale for
Children (GASC) in American school children (Sarason et al.). The scale has a four-
month test-retest reliability of 0.67 (Sarason et al.). Children who score above or below
established cutpoints are classified as being high or low test-anxious. Girls with scores
above 16 and boys with scores above 12 are considered to be high test-anxious. Girls
with scores below 10 and boys with scores below 7 are considered to be low test-anxious
(Beidel & Turner, 1988).

In addition to the test anxiety surveys, student demographic, academic
achievement, and MEAP test performance data were collected. The following sections
describe each area of data collection.

Demographic Information Form

Demographic information was collected through a form that was included with
the parental consent form. Parents reported their child’s date of birth, age, grade, sex,
ethnicity, and educational verification (i.e., general education versus special education).
Parents also reported on their educational attainment and employment status.

Socioeconomic status (SES) was calculated using the Hollingshead Four Factor Index of

50

Social Status (Hollingshead, 1975). The Demographic lnforrnation Form is included in
Appendix A.
Academic Performance Form

In order to examine whether or not student performance on the state assessment
was affected by test anxiety, it was necessary to control for students’ prior achievement.
Some previous researchers have attempted to do this using student scores on previous
standardized assessments (e.g., Mulvenon et al., 2001; 2005); however, these scores are
likely to be similarly affected by test anxiety. Therefore, students’ course grades were
used as proxies for student achievement. Although course grades have some stakes
associated with them, the grades children receive are based on a diverse set of course
assignments, increasing the likelihood that the grades are accurate representations of
students’ true achievement. Student grades in English language arts, mathematics, and
science were collected for the first grading period of the academic year. Science grades
were also collected for current fifth grade students who took the MEAP subject test in
science. The Academic Performance Form is included in Appendix B.
Michigan Educational Assessment Program Data

The Michigan Educational Assessment Program (MEAP) is the state
accountability testing program used to comply with the requirements of the No Child Left
Behind Act of 2001 (N CLB) in the state of Michigan. The MEAP is a criterion-
referenced test that measures students’ skills in the core content areas of mathematics,
science, social studies, and English language arts. Testing occurred during the last three
weeks of October 2008. Third, fourth, and fifth grade students took the MEAP

assessment in the areas of Reading and Writing (English Language Arts) and

51

Mathematics. Fifth grade students also took the MEAP Science content test. Test data
were reported through class grade roster reports and entered into the database by the
researchers.

Scale scores are reported for each content area. For third grade students, a score of
300 is the lowest score in the proficient range. The same pattern is true for all grades, i.e.,
400 for fourth grade, 500 for fifth grade, etc. are the lowest scores in the proficient range
for these grades; hence, scale scores are not comparable across grades. A third grade
student who scores 400 is not performing at a fourth grade level. Scores are also
classified as ranging from Level 1 to Level 4. Level 1 indicates an Advanced Student,
Level 2 indicates a Proficient student, Level 3 indicates a Partially Proficient student, and
Level 4 indicates a Not Proficient student (State of Michigan, 2007). Additionally,
students are classified as performing in the low, middle, or high range of each Level.
Thus, student performance level is reported on a 12-point scale, with students classified
as being low, middle, or high non-proficient (1 - 3), low, middle or high partially
proficient (4 - 6), low, middle or high proficient (7 - 9), or low, middle, or high advanced
(10 - 12).
Teacher Perception Questionnaires

Teacher perceptions of student test anxiety, teacher anxiety, and student
performance were collected through two brief teacher questionnaires. Teachers reported
on their perceptions of students’ anxiety before and during testing conditions as well as
their own anxiety related to student performance. Anxiety was rated on a 4-point Likert
scale ranging from “not at all” to “very” anxious, with higher scores indicating higher

anxiety. Finally, teachers reported on their perceptions of how students performed during

52

the two types of testing conditions. Performance was rated on a 5-point Likert scale
ranging from “very much below their ability” to “very much above their ability,” with
higher scores indicating higher performance. The Teacher Questionnaires are included in
Appendix C and D.
Operational Definitions of Test Anxiety

Students were classified as having low, moderate, or high test anxiety using three
definitions of test anxiety. Three definitions of test anxiety were used in order to examine
whether or not there were classification differences between the TASC, a frequently used
and well-established measure of test anxiety, and the CTAS, a contemporary measure of
test anxiety based on a biopsychosocial model. No test anxiety classifications were made
using students BASC-2-TA scores due to age differences between the current sample and
the normative sample for the BASC-2-TA.
Definition One

The first test anxiety definition was based on students’ scores on the Children’s
Test Anxiety Scale, which examines behavioral, cognitive, and physiological aspects of
test anxiety. Students were classified as having high test anxiety if their overall CTAS
score was one standard deviation above the mean of the standardization sample (78.46)
and as having low test anxiety if their CTAS score was one standard deviation below the
mean (45.48, Wren & Benson, 2004; Wren, personal communication, March 14, 2008).
Students whose test anxiety scores fall within one standard deviation of the mean were
classified as having moderate test anxiety.
Definition Two

The second test anxiety definition was based on students’ scores on the Test

53

Anxiety Scale for Children, which was developed by Sarason and colleagues in 1960, and
has been the most extensively used tool to measure childhood test anxiety. Cutpoints
established by Beidel and Turner (1988) were used to determine students’ test anxiety
classification. Girls with scores above 16 and boys with scores above 12 were classified
as high test-anxious. Girls with scores below 10 and boys with scores below 7 were
classified as low test-anxious. Girls with scores between 10 and 16 and boys with scores
between 7 and 12 were classified as having moderate test anxiety.
Definition Three

The third and most conservative test anxiety definition was based on students’
scores on both the Children’s Test Anxiety Scale and the Test Anxiety Scale for
Children. Students classified as having high test anxiety by both Definitions One and
Two were classified as high test-anxious and students classified as having low test
anxiety by both Definitions One and Two were classified as being low test-anxious.
Students classified as having moderate test anxiety by both Definitions One and Two
were classified as moderately test-anxious. Additionally, students who did not have
stable test anxiety classifications across the Definitions One and Two were classified as
having moderate test anxiety.

Data Collection Procedures

Recruitment, Consent, and Assent

The elementary schools where the study took place were selected by contacting
principals and explaining the study and its purpose. Schools were selected as a
convenience sample on the basis of the schools’ interest and commitment to the study.

Three schools in a suburban school district, with a total population of 61 7 students in

54

grades three through five, agreed to participate. The school district reported that it did not
engage in any specific test preparation for the MEAP, although teachers are encouraged
to format some tests using a multiple choice format so that students become familiar with
this testing format. All schools in the study had a history of meeting Annual Yearly
Progress (AYP) goals for the past five years except for one school, which did not meet
AYP in 2007-2008 because it only assessed 94% of the students classified as
economically disadvantaged on one portion of the MEAP and AYP standards require that
95% of students are assessed.

An informational letter was mailed home to the parents of children in the
participating schools that described the study and its purpose, and extended an invitation
for voluntary participation (see Appendix E). Consent forms were included in the
mailing, along with privacy envelopes to return consent forms to the school (see
Appendix F). Letters were sent to parents approximately three weeks prior to the data
collection. Prior to data collection, children were also informed that they had the right to
choose not to participate in the study. Students were asked to sign a statement of assent
on the cover page of each survey packet that they completed (see Appendix G and H).
Participating students received nominal items, including pencils, stickers, and erasers,
following each survey.

MEAP Test Anxiety Data Collection

Prior to data collection, researchers read standardized instructions explaining
how the students should complete the measures (see Appendix 1). Researchers
administered the first survey (see Appendix G) within 1 to 3 days after the completion of

the Michigan Educational Assessment Program (MEAP) administration. Measures

55

included the adapted Children’s Test Anxiety Scale and adapted Behavior Assessment
Scale for Children Test Anxiety Subscale. The CTAS and the BASC-2-TA measures
were adapted for use in the first survey by altering the word test to MEAP throughout
survey items. Measures were randomly ordered by classroom in order to control for
order effects.
Classroom Test Anxiety Data Collection

Researchers administered the second survey (see Appendix H) 28 days after the
first survey. This survey focused on children’s responses to classroom tests. The delay
between surveys was designed to help students clearly differentiate between the two
surveys’ foci, the first being MEAP testing and the second being classroom testing.
Measures on the second survey included the Children’s Test Anxiety Scale, the Behavior
Assessment Scale for Children- Second Edition Test Anxiety Subscale, and the Test
Anxiety Scale for Children. Researchers read a set of standardized instructions explaining
how the students should complete the measures and reminded students of all classroom
tests they had taken in the past two weeks. Measures were randomly ordered by
classroom in order to control for order effects.

Data Analysis

Data were entered and stored in a computer database that was password protected.
No identifying information was recorded in the database. All raw data were examined for
any errors in coding through examination of descriptive statistics and visual presentations
of data. Twenty-four percent of all surveys (n= 151) were receded to assess interrater
reliability of subscale and total scale scores across all three test anxiety measures. High

interrater reliability was found between coders, with 98% agreement. Following data-

56

checking procedures, statistical analyses were used to examine response patterns and
relationships in the data.
Question One

Question One had two primary aims: (1) examining differences in reported test
anxiety between testing conditions and (2) examining differences in how the three
definitions of test anxiety classified children as low, moderate, or high test anxious.

Aim 1. Descriptive statistics of reported test anxiety in relation to the MEAP and
classroom tests on the CTAS, BASC-2-TA, and TASC were examined. Wilcoxon Signed
Rank tests were used to determine whether or not students reported significantly different
levels of test anxiety for classroom testing and MEAP testing conditions. Nonparametric
statistics were conducted because the distributions of test anxiety scores were not
normally distributed. The distributions of scores on all test anxiety measures were
positively skewed. Test anxiety was considered a continuous variable for the purpose of
these analyses. A Bonferroni correction of p s .01 was used to correct for multiple
comparisons. Effect sizes of scores that differed significantly across test types were
calculated, r = z / ‘12n.

Aim 2. Chi-square goodness of fit analyses were conducted to examine whether or
not students’ test anxiety classifications (Low, Moderate, High) as determined by the
three definitions of test anxiety differed significantly by testing condition (MEAP versus
classroom tests). A Bonferroni correction of p S .01 was used to correct for multiple
comparisons. Additional Chi-square goodness of fit analyses were conducted to examine
whether or not students’ test anxiety classifications (Low, Moderate, High) differed.

significantly across different definitions within the same testing condition.

57

Question Two

The primary aim of Question Two was to examine differences in teacher
perceptions of student test anxiety, teacher anxiety, and student performance between
testing conditions. Independent-samples t-tests were conducted to examine differences
because teacher perception data were collected anonymously and it was not possible to
pair teacher questionnaires across testing conditions. When Levene’s test for equality of
variances indicated a violation of the assumption of equal variance, t-tests without the
assumption of equal variance were examined.

Question Three

Question Three had two primary aims: (1) examining differences in MEAP test
performance across students identified as having low, moderate, and high test anxiety and
(2) examining predictors of high-stakes test performance.

Aim 1. Descriptive statistics of test performance on the English Language Arts
and Math portions of the MEAP were examined and one-way between- groups analyses of
variance with Tukey HSD post-hoe tests were used to determine whether or not students
with different test anxiety classifications performed differently on the MEAP. For the
purpose of this analysis, student performance level on the MEAP was used so that student
performance data across all three grades could be combined. Effect sizes of performance
levels that differed significantly across test anxiety classifications were calculated, eta
squared = (Sum of squares between-groups) / (Total Sum of Squares).

Aim 2. Hierarchical multiple regression analyses were used to assess the ability of
students’ test anxiety scores to predict test performance on the MEAP. Student scholastic

achievement was controlled for by including students’ academic grades in each

58

regression analysis. Demographic variables, including student sex, SES, educational
verification, and ethnicity, were also included in the regression analyses to examine the
influence of these variables on student performance. Test anxiety was considered a
continuous variable for the purpose of these analyses. Separate analyses were conducted
for each grade level because student academic grades and MEAP scores were reported
using different scales across grades three, four, and five.
Question Four

Question Four had two primary aims: (1) examining if there were systematic
differences between students’ test anxiety scores when students were grouped by
different demographic variables and (2) examining if demographic variables significantly
predicted test anxiety measure scores.

Aim 1. Nonparametric statistics were used to examine if there were significant
differences in students’ test anxiety scores as a result of differences in students by
demographic variables. Nonparametric analyses were conducted because the distributions
of test anxiety scores across all measures were not normally distributed. Test anxiety was
considered a continuous variable for the purpose of these analyses. Mann-Whitney U
tests were used to examine differences in dichotomous demographic variables, including
sex, ethnicity, and educational verification. Kruskal-Wallis Tests were used to examine
differences in demographic variables with three or more categories, including
socioeconomic status and grade. A Bonferroni correction of p S .01 was used to correct
for multiple comparisons. Effect sizes of scores that differed significantly across test
types were calculated, r = 2/ \ln.

Aim 2. Hierarchical multiple regression analyses were used to assess the ability of

59

demographic variables, including student grade, sex, SES, educational verification, and
ethnicity, to predict test anxiety. Data transformations of test anxiety scores were
completed to account for the non-normal distribution of survey data, which was
positively skewed. Test anxiety was considered a continuous variable for the purpose of
these analyses. The relationship between demographic variables and test anxiety were
examined in relation to both MEAP and classroom testing conditions.

Question 5

Question Five had two primary aims: (l) examining the relationship between the
different test anxiety measures within testing conditions and (2) examining the
relationship between the measures across the two testing conditions.

Aim 1. Correlation analyses were conducted to examine the relationship between
test anxiety scores on the CTAS, BASC-2-TA, and TASC for both testing conditions.
Nonparametric, two-tailed Spearman rank order correlation (rho) analyses were used
because children’s reported test anxiety across measures was not normally distributed.

Test anxiety was considered a continuous variable for the purpose of these analyses.

Correlations were considered to be strong if r, 2 .60, moderate if .40 S rs S .59, and low if

r, S .40. Additionally, a Kappa Measure of Agreement was calculated for children

classified using the TASC, the “gold standar ” measure of test anxiety and the CTAS, a
more contemporary measure of test anxiety designed to assess children’s psychological,
behavioral, and physiological response to testing, to examine how similarly the two
measures’ classified students as anxious or non-anxious.

Aim 2. Correlation analyses were conducted to examine the relationship between

children’s scores on test anxiety measures across testing conditions. Nonparametric, two-

60

tailed Spearman rank order correlation (rho) analyses were used because children’s
reported test anxiety across measures was not normally distributed. Test anxiety was
considered a continuous variable for the purpose of these analyses. Analyses were

conducted for the three different test anxiety scales and the subscales of the CTAS.

Correlations between test conditions were considered to be strong if rs 2 .60, moderate if

.40 S r, S .59, and low if rs S .40.

61

CHAPTER 4
RESULTS
Question One
Aim 1

Wilcoxon Signed Rank Tests were used to examine the differences between test
anxiety scores reported for MEAP and classroom tests across measures using a
Bonferroni correction of p S .01. Table 3 summarizes the differences identified between
reported test anxiety on the MEAP assessment and classroom tests across measures.
There was a statistically significant difference between students’ reported test anxiety on
the CTAS-Total scale across the MEAP and classroom tests, 2 = -5.04, p < .0005, with a
small effect size (r = 021). Students reported significantly greater test anxiety on the
CTAS-Total scale in relation to the MEAP assessment than in relation to classroom tests
(Md = 52 versus Md = 47, respectively). There was also a statistically significant
difference in students’ BASC-2-TA test anxiety scores between the MEAP and classroom
tests, 2 = -2.47, p = .01, with a small effect size (r = -0.10). Students reported
significantly greater test anxiety on the BASC-2-TA in relation to the MEAP than in
relation to classroom tests (Md = 4 versus Md = 4, respectively).

Wilcoxon Signed Rank tests also identified significant differences between
reported test anxiety on two CTAS subscales in relation to the MEAP and classroom
tests. There was a statistically significant difference in students’ CTAS-Thought subscale
scores between the MEAP and classroom tests, 2 = -4.87, p < .0005, with a small effect
size (r = -0.20). Students reported significantly greater test anxiety on the CTAS-Thought

subscale in relation to the MEAP than in relation to classroom tests (Md = 24 versus Md

62

.mcoflshgoo 035:8 com S. w a mo corooboo Economcomn

coke—EU com 23m beige. amok “Oman—r mo_aomn:m beta/w “mob .coEnm vacuum .aBEEU com 28m

82:333. Eomrfifiom n<._.-~-0m<m 618m .30me “web Macaw—EU ”mar—.0 ”Swamp—m 808383. Reagan—um SawEomE ”madam;

 

 

 

 

as a; 8m 85

:3 5%. am 3m .9; 4 mmm 8.4 3.4 6 8m 33-8%
88. v :3- mom m; 3.2 S mmm was 2.2 3 8m 28833..
85 mod- am as an: 2 4mm mam 2.2 3 8m sauce
88. v 5.... EN 4.3 was a 8». N; 0me 4m 8m afiéfi
38. v 36. man 5: 5mm 3 w: 2.2 8.2 mm 8m .33

m<eu

an o 2 a. 2 o: a 9. 3 so: = soaaoz

mumor—r 800.5920 sewer—r mdme—Z

 

MEENN Booties 35c BeEfiman KENS 2: 2: 339m. 553%.. ads

m 033.

63

= 21, respectively). There was also a statistically significant difference in students’
CTAS-Autonomic subscale scores between the MEAP and classroom tests, 2 = -5.87, p <
.0005, with a small effect size (r = -0.24). Students reported significantly greater test
anxiety on the CTAS-Autonomic subscale in relation to the MEAP than in relation to
classroom tests (Md = 14 versus Md = 12, respectively). There was no significant
difference in students’ CTAS-Off-task subscale scores between the MEAP and classroom
tests, = -0.08, p = .93.
Aim 2

Chi square goodness of fit tests were conducted to examine if there were
significant differences between the proportions of students identified as low, moderate, or
high test-anxious across the MEAP and classroom testing conditions using a Bonferroni
correction of p S .01. Table 4 summarizes the differences identified between students’
test anxiety classifications across the MEAP and classroom testing conditions. Using
students’ CTAS-Total scale scores to classify students, there was a significant difference
in the proportions of students with low, moderate, and high levels of test anxiety in the

classroom test condition as compared to the values of 32.2%, 58.6%, and 9.1%,

respectively that were obtained in the MEAP condition, 12 (2, n = 318) = 27.90, p <

.0005. More students reported low (143 vs. 102.5 expected) and high (34 vs. 29
expected) test anxiety and fewer students reported moderate test anxiety (141 vs. 186.4
expected) in the classroom testing condition than the MEAP condition.

Chi square goodness of fit tests also identified significant differences in students’
test anxiety classifications using two of the CTAS subscales to classify students’ test

anxiety across the MEAP and classroom testing conditions. Using students’ CTAS-

64

388388 23:3: com .o. .v. «Co :osootoo Ecobmcoma

56:6 6.2

2an 53x56. amok ”Um/E. ”23m bomxc< “mob Macaw—EU Hm<HU mgcwoi EoEmmomm< Encamosem gwfioaz HAZE/r

 

 

 

88. v 3.? em a: a: 2 m2 3 Omfiafio 6393 m 8:58
8 s a: GEE N 8:58
88. v N3. 3. m: a: we 42 8 6:853.
S. we.” 3 ”2 N2 8 ca 3 usage
88. v 02; a o: 3 mm 2: 8 mzwsofi
38. v 8.2 X :2 m: ”N s: a .93
65.8 _ 8:58
an as as: 82 33 as: 82 33 «65362
_o>ox_ .0253 L264 302?.
warmed. Seemed—U EoEmmomm/x m<m2

 

neEEEQ \3 uneacmhteb

35.85 Eggheads SSE RENE «$953 babe? :65 «RE has Seamus: $8M Mataommx Sathmsxe x3252 2: S amoamngQ

v 2an

65

Thought subscale scores to classify students, there was a significant difference in the
proportions of students with low, moderate, and high levels of test anxiety in the

classroom testing condition as compared to the values of 32.2%, 58.6%, and 9.1%,
respectively that were obtained in the MEAP condition, 12 (2, n = 320) = 38.36, p <

.0005. More students reported low (153 vs. 103.2 expected) and high (31 vs. 29.2
expected) test anxiety and fewer students reported moderate test anxiety (136 vs. 187.6
expected) in the classroom testing condition than the MEAP condition. Using the CTAS-
Autonomic subscale scores to classify students, there was also a significant difference in
the proportions of students with low, moderate, and high levels of test anxiety in the

classroom testing condition as compared to the values of 21 .2%, 63.2%, and 15.6%,

respectively that were obtained in the MEAP condition, 2'2 (2, n = 322) = 43.72, p <

.0005. More students reported low (116 vs. 68.2 expected) test anxiety and fewer
students reported moderate (172 vs. 203.5 expected) and high (34 vs. 50.3 expected) test
anxiety in the classroom testing condition than the MEAP condition. Using the CTAS-
Off-task subscale scores to classify students, there was no significant difference in the
proportions of students with low, moderate, and high levels of test anxiety in the

classroom testing condition as compared to the values of 27.8%, 62.1%, and 10.1%,

respectively that were obtained in the MEAP condition, 12 (2, n = 324) = 8.08, p = .02.
Finally, using students’ stable CTAS-Total and TASC scores to determine student

test anxiety levels, a Chi square goodness of fit test indicated that there was a significant

difference in the proportions of students with low, moderate, and high levels of test

anxiety in the classroom testing condition as compared to the value of 28.3%, 67.2%, and

4.5%, respectively that was obtained in the MEAP condition, 12 (2, n = 290) = 42.24, p <

66

.0005. More students reported low (1 19 vs. 82.1 expected) and high (26 vs. 13.0
expected) test anxiety and fewer students reported moderate (145 vs. 194.9 expected) test
anxiety in the classroom testing condition than in the MEAP condition.

Figure 3 illustrates the differences identified in test anxiety classifications by
Definition and test condition type, with the CTAS-Total classification proportions
represented in the Definition 1 columns, the TASC classification proportions represented
in the Definition 2 column, and the stable CTAS-Total and TASC classification
proportions represented in the Definition 3 columns. Letters A-E below refer to the
different columns represented on Figure 3. As discussed previously, students differed
significantly in Definition 1 classifications across the MEAP (A) and classroom testing
(B) conditions, p < .0005. Students also differed significantly in Definition 3
classifications across the MEAP (D) and classroom testing (E) conditions, p < .0005.

Additional Chi-square goodness of fit analyses were conducted to examine if
there were significant differences in classification patterns across the different test
anxiety classification definitions. First, differences were examined among students’ test
anxiety classifications on the MEAP assessment. There was a significant difference in the
proportions of students with low, moderate, and high levels of test anxiety on the MEAP
across the classifications established by Definition 1 and Definition 3. The students

classified by the Definition 1 (A) significantly differed from the proportions of 28.3%,
67.2%, and 4.5%, respectively obtained by Definition 3 (D), 12 (2, n = 307) = 19.61, p <

.0005. More students reported low (99 vs. 86.9 expected) and high (28 vs. 13.8 expected)
test anxiety and fewer students reported moderate test anxiety (180 vs. 206.3 expected)

when classified using CTAS-Total scores versus stable CTAS-Total and TASC scores.

67

1000/0 *

90%
800/0

h‘ k. ._

700/0

I Moderate

I Low

50% 1
40% T
30% T
20% ‘
10% T

0°/o T2“

 

MEAP Tests Tests MEAP Tests
Def. 1 Def. 1 Def. 2 Def. 3 Def. 3
(A) (B) (C) (D) (E)

Test Anxiety Definition

Figure 3: Rates of Test Anxiety by Classification Definitions

68

Next, differences between students’ test anxiety classifications on the Classroom
tests were examined across the different Definitions. There was a significant difference in
the proportions of students identified with low, moderate, and high levels of test anxiety
on classroom tests across the Definition 1 and Definition 2. The students classified by

Definition 1 (B) significantly differed from the proportions of 53.3%, 25.9%, and 20.8%,

respectively obtained by Definition 2 (C), 12 (2, n = 318) = 61.77, p < .0005. More

children reported moderate (141 vs. 82.3 expected) test anxiety and fewer children
reported low test anxiety (141 vs. 169.5 expected) and high (34 vs. 66.2 expected) when
classified using CTAS-Total scores rather than TASC scores. The students classified by

Definition 2 (C) also significantly differed from the proportions of 41 .0%, 50.0%, and

9.0%, respectively obtained by Definition 3 (E), 12 (2, n = 317) = 98.21, p < .0005. More

students reported high (66 vs. 28.4 expected) and low test anxiety (169 vs. 130.1
expected) test anxiety and fewer students reported moderate (82 vs. 158.5 expected)
when classified using TASC scores rather than stable CTAS-Total and TASC scores.

The students classified by Definition 1 (B) did not differ significantly from the
proportions of 41 .0%, 50.0%, and 9.0%, respectively obtained by Definition 3 (E), 12 (2,

n = 318) = 4.29,p = .12.
Question Two
Teacher questionnaire data were analyzed to examine differences in teachers’
perceptions of student anxiety, teacher anxiety, and student performance based on the
testing condition. Table 5 summarizes differences in perceived anxiety and performance.
There was a significant difference in teachers’ perceptions of students’ anticipatory test

anxiety for the MEAP (M = 2.36, SD = .73) and classroom testing (M = 1.93, SD = .46); t

69

 

 

 

 

 

 

8. .8. 7 R. a: an. own 282682 so Renata 382a 36» 6E5 8» 8 =63 36m .4
So. m S .m we. ow; ow. omd mmeoumweofl so ow 2:95 359% So» :63 Be: Sons a 22» £935 263 .m
m _. mm.“ mm. cod no. mmd ammo. 2t wage FoEwcooH Sons 35me So» 86>» @538 253 .N
8. SN we. no.“ mm. cmN «Eons. xloogofi 882. F0238. Sons 35an So» 083 Satan Bo: .—
m a 9m. 3 9w E 538:0
memo... EoEmmomm<
Became—U med—2

 

ENSSSD M3365 3 meEV Ems assoefl SEE Bonnie? weemimoamm x2335
m 2an

70

(34.85) = 2.21, p = .03 (two-tailed). There was no significant difference in teachers’
perceptions of students’ test anxiety during the MEAP (M = 2.32, SD = .65) or classroom
testing (M = 2.00, SD = .53); t(35) = 1.57, p = .13 (two-tailed), although there was a
trend toward rating students as more anxious during the MEAP. There was a significant
difference in teachers’ perceptions of their own anxiety about student performance on the
MEAP (M= 2.59, SD = .80) and classroom tests (M= 1.80, SD = .68); t (35) = 3.15, p =
.003 (two-tailed). There was no significant difference in teachers’ perceptions of
students’ performance on the MEAP (M = 2.50, SD = .58) and classroom tests (M = 2.80,
SD = .37); t (34.90) = -1.93, p = .06 (two-tailed), although there was a trend toward rating
students as performing better on classroom tests than on the MEAP.
Question Three

Aim 1

One-way between-groups analyses of variance were conducted to examine
whether or not students with different test anxiety classifications, as defined by CTAS-
Total scores (Definition 1) and the TASC scores (Definition 2), performed significantly
differently on the MEAP in the areas of English Language Arts, Math, and Science for
students in grades three through five. First, analyses were conducted with students
divided into low, moderate, and high test anxiety groups on the basis of their CTAS-Total
scores. There was a significant difference between student performance level on the
English Language Arts section of the MEAP for the three test anxiety groups: F (2, 297)
= 7.07, p = .001. The effect size of the difference between groups was small, eta squared
= .05. Post-hoe comparisons using the Tukey HSD indicated that the mean performance

of the low test anxiety group (M = 8.81, SD = 1.19) was significantly greater than the

71

moderate test anxiety group (M= 8.17, SD = 1.47). The high test anxiety group (M= 8.20,
SD = 1.35) did not differ significantly from either the moderate or low test anxiety
groups. There was also a significant difference between student performance level on the
Math section of the MEAP for the three test anxiety groups: F (2, 292) = 18.60, p <
.0005. The effect size of the difference between groups was moderate, eta squared = .1 1.
Post-hoe comparisons indicated that the mean performance of the low test anxiety group
(M = 9.86, SD = 1.26) was significantly greater than the moderate test anxiety group (M=
8.87, SD = 1.56) and the high test anxiety group (M= 8.24, SD = 1.76). The high test
anxiety group did not differ significantly from the moderate test anxiety group.

Next, analyses were conducted with students divided into low, moderate, and high
test anxiety groups on the basis of their TASC scores. There was a significant difference
found between student performance level on the English Language Arts section of the
MEAP for the three test anxiety groups: F (2, 308) = 6.24, p = .002. The effect size of the
difference between groups was small, eta squared = .04. Post-hoe comparisons indicated
that the mean performance of the low test anxiety group (M = 8.65, SD = 1.36) was
significantly greater than moderate test anxiety group (M = 8.13, SD = 1.42) and the high
test anxiety group (M= 8.05, SD = 1.47). The high test anxiety group did not differ
significantly from the moderate test anxiety group. There was also a significant
difference between student performance level on the Math section of the MEAP for the
three test anxiety groups: F (2, 304) = 10.01, p < .0005 . The effect size of the difference
between groups was moderate, eta squared = .06. Post-hoc comparisons indicated that the
mean performance of the low test anxiety group (M = 9.53, SD = 1.41) was significantly

greater than the moderate test anxiety group (M= 8.66, SD = 1.96) and the high test

72

anxiety group (M= 8.80, SD = 1.52). The high test anxiety group did not differ
significantly from moderate test anxiety group. Table 6 summarizes the differences in
MEAP performance level across low, moderate, and high test-anxious students.
Aim 2

Hierarchical multiple regression analyses were used to assess the ability of test
anxiety in relation to the MEAP to predict MEAP performance. Analyses were conducted
separately for students in each grade level. Students’ academic achievement was
controlled for by including a measure of scholastic achievement in each analysis.
Additionally, the ability of student demographic characteriétics to predict MEAP
performance was also examined by including these variables in the regression analyses.
Preliminary analyses were conducted for each regression analysis to ensure no violation
of the assumptions of multiple regression. Tolerance and variance inflation factor scores
(VIFs) were examined and no indications of multicollinearity were identified (Tols. > .10
and VIFs < 10). Normal probability plots of the regression standardized residuals and
scatterplots of the standardized residuals were also examined and no violations of
normality, linearity, homoscedasticity, or independence of residuals were identified for
any of the regression analyses. The CTAS-Total scale was used as the only measure of

test anxiety for these analyses and the BASC-2-TA scale was excluded to prevent
multicollinearity between the two highly correlated variables (rs = .71).

The first regression analysis for each grade examined the value of the CTAS-
Total scale and student demographic characteristics in predicting test performance on the
English Language Arts section of the MEAP when academic achievement was controlled

for using students’ reading grades. Academic achievement was entered in Step 1 and

73

coup—EU .8.“ 2QO

house “mob Um»; ”200m 30,—. 23m 523 an... Pepe—EU ”.5oh-m<._b ”Edcwoi EoEmmomm< Snowman—om gwfioaz Hm<m2a

 

 

 

 

 

 

 

 

88v .3: an. a; we 03 8.” R :2 mg 8. 552
a... 3.» E a... m; M: on. B.” E a}. 8.333

So. 4% fiacm

aeaocafiu 85.
88v 8.2 o: 43 mm of a; m: 3.. £6 8 522
so. 8.» 92 cm.» mm t... 2.” N: a: a; 8 32 owazwcfi
imam
8:85:86 38.230
m a 9.. E a a. 3 a ma. 2 2 83% a<m2
.3: 82 33

 

«ESQ bomxtw Bfi x3 36w 3:35.5me KVNS. E amonmxmtEQ

o 033.

74

included students’ Reading Grade, the test anxiety measure was entered in Step 2 and
included the CTAS-Total scale score, student sex was entered in Step 3 (female = 1, male
= 0), socioeconomic status was dummy coded in Step 4: SES II (SES-II = 1, other = 0),
SES-III (SES-III = 1, other = 0), and SES-IV-V (SES-IV-IV = 1, other = 0), educational
verification was entered in Step 5: (special education = 1, general education = 0), and
ethnicity was entered in Step 6: (Non-Caucasian = 1, Caucasian = O). The second
regression analysis for each grade examined the value of the CTAS-Total scale and
student demographic characteristics in predicting test performance on the Math section of
the MEAP when academic achievement was controlled for using students’ math grade.
The regression analysis was entered in the same manner as the previous analysis with the
exception of entering students’ math grades in Step 1.

Grade 3. The regression analysis of third grade students’ English Language Arts
MEAP scores as the dependent variable indicated that reading grade explained 36.0% of
the variance in third graders’ English Language Arts MEAP scores, F (1 , 94) = 52.98, p <
.0005. After the CTAS-Total was entered in the model, the total variance explained was
37.3%, R squared change = .013, F change (1, 93) = 1.92, p = .17, indicating that CTAS-
Total did not contribute significantly to the model. After the sex was entered in the
model, the total variance explained was 41.4%, R squared change = .040, F change (1 ,
92) = 6.30, p = .01. After SES was entered in the model, the total variance explained was
46.1%, R squared change = .048, F change (3, 89) = 2.62, p = .06. Educational
verification and ethnicity variables did not add significantly to the model, ps > .10. The
overall model was a significant predictor of English Language Arts scores on the MEAP,

F (8, 87) = 9.87, p < .0005. In the final model, reading grade, CTAS-Total score, sex, and

75

SES-III variables were significant predictors of English Language Arts MEAP scores,
with reading grade recording the highest beta value ([3 = .50, p < .0005) in comparison to
the other variables, SES-III (B = -.24, p = .02), sex ([3 = .21, p = .01), and CTAS-Total (B
= -.17,p = .04).

The regression analysis of third grade students’ Math MEAP scores as the
dependent variable indicated that math grade explained 9.7% of the variance in third
graders’ Math MEAP scores, F (1, 94) = 10.09, p = .002. After the CTAS-Total was
entered in the model, the total variance explained was 20.4%, R squared change = .107, F
change (1, 93) = 12.47, p = .001. After the educational verification was entered in the
model, the total variance explained was 32.7%, R squared change = .094, F change (1 ,
88) = 12.29, p = .001. Sex, SES, and ethnicity variables did not add significantly to the
model, ps > .10. The overall model was a significant predictor of MEAP Math scores, F
(8, 87) =5.40, p < .0005. In the final model, CTAS-Total and educational verification
variables were the only significant predictors of Math MEAP scores, with CTAS-Total (B
= -.35, p < .0005) recording a higher beta value than educational verification (B = -.32, p
= .002). Tables 7 summarizes how predictive students’ CTAS-Total scale scores,
academic grades, and demographic characteristics were in accounting for variance in
English Language Arts and Math MEAP scores for students in grade three.

Grade 4. The regression analysis of fourth grade students’ English Language Arts
MEAP scores as the dependent variable indicated that reading grade explained 14.1% of
the variance in fourth graders’ English Language Arts MEAP scores, F (1, 104) = 17.09,
p < .0005. After the CTAS-Total was entered in the model, the total variance explained

was 17.8%, R squared change = .037, F change (1 , 103) = 4.67, p = .033. After SES was

76

Table 7

Multiple Regression OfPredictors of 'MEAP Scores in Grade 3

 

 

 

 

 

 

MEAP
English Language Arts Math
. . a 2 a 2
Add1t1ve Model [5 AR F-A B AR F—A
- I! it it
+Ach1evement ...... .360 52.98 .097 10.09
Reading Grade _ 50
Math Grade _]7’r
+CTAS-T0tal -.17’ .013 1.92 -.35’" .107 12.47"
+Sex .21" .040 6.30 .04 .000 .01
+SES (dummy) .048 2.62' .029 1.12
SES 11 ~.03 .04
SES III -24" -.09
SES IV-V ‘03 .02
+Educational -.12 .014 2.28 -.32" .094 12.29"
Verification
+Ethnicity -.04 .001 .19 -.08 .005 .66
’13 in the final model

”’p < .0005, "p s .01, ’p g .05, ’p s .10

77

entered in the model, the total variance explained was 26.0%, R squared change = .074, F
change (3, 99) = 3.31, p = .023. Sex, educational verification, and ethnicity variables did
not add significantly to the model, ps > .10. The overall model was a significant predictor
of MEAP English Language Arts scores, F (8, 97) = 4.31, p < .0005. In the final model,
reading grade, CTAS-Total score, SES-II, and SES-III variables were significant
predictors of English Language Arts MEAP scores, with reading grade recording the
highest beta value (B = .31, p = .001) in comparison to the other variables, SES-II (B = -
.30, p = .008), SES-III (B = -.27, p = .011), and CTAS-Total (B = -.20, p = .038).

The regression analysis of fourth grade students’ Math MEAP scores as the
dependent variable indicated that math grade explained 6.5% of the variance in fourth
graders’ Math MEAP scores, F (1, 104) = 7.19, p = .009. After the CTAS-Total was
entered in the model, the total variance explained was 9.7%, R squared change = .032, F
change (1, 103) = 3.65, p = .059. Sex, SES, educational verification, and ethnicity
variables did not add significantly to the model, ps > .10. The overall model was a
significant predictor of MEAP Math scores, F (8, 97) = 2.28, p = .028. In the final model,
math grade and SES-III variables were significant predictors of Math MEAP scores, with
SES-III recording the highest beta value (B = -.24, p = .039) in comparison to math grade
(B = .23, p = .028). Table 8 summarizes how predictive students’ CTAS-Total scale,
academic grades, and demographic characteristics were in accounting for variance in
English Language Arts and Math MEAP scores for students in grade four.

Grade 5 The regression analysis of fifth grade students’ English Language Arts

MEAP scores as the dependent variable indicated that reading grade explained 19.0% of

78

Table 8

Multiple Regression 0f Predictors of MEAP Scores in Grade 4

 

MEAP

 

 

 

 

English Language Arts Math
. . a 2 a 2 .

Addrtlve Model B A R F-A B A R F -A
+Achievement .141 17.09“ .065 7.19"

Reading Grade 31"

Math Grade 23*
+CTAS-Total -20" .037 4.67’ -. 15 .032 3.65’
+Sex .15 .007 .90 -.06 .008 .96
+SES (dummy) .. .074 3.31' 4 .049 1.90

SES 11 -.30 -.20

SES 111 -.27 -.24

SES rv-v -. 13 -.07
+Educational -.03 .001 .10 -.04 .001 .12
Verification
+Ethnicity .04 .002 .22 -.O6 .003 .38
as m the final model

"’10 < .0005, "p s .01, ’p s .05, ’p s .10

79

the variance in fifth graders’ English Language Arts MEAP scores, F (1, 78) = 18.26, p <
.0005. When the CTAS-Total was entered in the model, the total variance explained was
22.6%, R squared change = .036, F change (1, 77) = 3.62, p = .061. After SES was
entered in the model, the total variance explained was 34.5%, R squared change = .102, F
change (3, 73) = 3.80, p = .014. After educational verification was entered in the model,
the total variance explained was 37.5%, R squared change = .029, F change (1, 72) =
3.40, p = .07. Sex and ethnicity variables did not add significantly to the model, ps > .10.
The overall model was a significant predictor of MEAP English Language Arts scores, F
(8, 71) = 5.31, p < .0005. In the final model, student reading grade and SES-III variables
were significant predictors of English Language Arts MEAP scores, with reading grade
recording the highest beta value (B = .36, p = .001) in comparison to SES-III (B = -.35, p
= .007).

The regression analysis of fifth grade students’ Math MEAP scores as the
dependent variable indicated that math grade explained 3.2% of the variance in fifth
graders’ Math MEAP scores, F (1, 78) = 2.60, p = .11. After the CTAS-Total was entered
in the model, the total variance explained was 11.2%, R squared change = .080, F change
(1 , 77) = 6.91 , p = .01. After sex was entered in the model, the total variance explained
was 18.1%, R squared change = .069, F change (1, 76) = 6.42, p = .01. After educational
verification was entered in the model, the total variance explained was 25.8%, R squared
change = .041, F change (1, 72) = 4.02, p = .05. SES and ethnicity variables did not add
significantly to the model, ps > .10. The overall model was a significant predictor of
MEAP Math scores, F (8, 71) = 3.10, p = .005. In the final model, sex and educational

verification were the only significant predictors of Math MEAP scores, with sex (B = -

80

.33, p = .004) recording a higher beta value than educational verification (B = -.21, p =
.05).

The regression analysis of fifth grade students’ Science MEAP scores as the
dependent variable indicated that science grade explained 13.9% of the variance in fifth
graders’ Science MEAP scores, F (1, 78) = 12.55, p = .001. When CTAS-Total was
entered in the model, the total variance explained was 15.0%, R squared change = .012, F
change (1, 77) = 1.07, p = .30, indicating that CTAS-Total did not add significantly to the
model. After sex was entered in the model, the total variance explained was 22.4%, R
squared change = .074, F change (1, 76) = 7.22, p = .009. SES, educational verification,
and ethnicity variables did not add significantly to the model, ps > .10. The overall model
was a significant predictor of MEAP Science scores, F (8, 71) = 3.48, p = .002. In the
final model, science grade and sex variables were significant predictors of Science
MEAP scores, with science grade recording a higher beta value (B = .43, p = .001) than
sex (B = -.27, p = .02). Tables 9 summarizes how predictive students’ CTAS-Total scale,
academic grades, and demographic characteristics were in accounting for variance in
English Language Arts, Math, and Science MEAP scores for students in grade five.

Question Four
Aim 1

Mann-Whitney U tests and Kruskal-Wallis tests were used to examine differences
in test anxiety resulting from differences in student demographic variables, including sex,
ethnicity, educational verification, socioeconomic status, and grade. Analyses were
conducted separately to examine differences in test anxiety scores in relation to the

MEAP and classroom testing. Table 10 summarizes differences in students’ test anxiety

81

S. w $.32 .52 .8806 fives mam 65:3.
I F III

 

 

 

 

8. 8o. 8.- 3. So. .5.- 8. 8o. 8.- 3.8%?
.34 2o. 2. ems.“d :5. MN. +ov.m amo. +w_ .- aogomto> fiaomwmosvm+

:.- 2 .- , 8.- >->~ ma

8.. S .- rm....- E ma

2. 8.- 8.- a mm
do. So. 2: mg. .23. 8.. $853 mm?
...-as 48. :R- .mee oeo. :2.- ef to. E .- 6m.-
B._ So. 3.- :36 23. m: .- ...-Sm e8. .2 .- Ede-2.8 +

Q. 2.80 00:25

..E. 635 :32

cm. 2:20 meme-mom
:32 32. SN 3.. ...eS: 8.. .. 2686626.}.

d-m am d an «m mm d an am mm d an .682 62:26....
ooaofim 532 mg amass fizwam
3&2

 

w $8.5 S 8.59m. ANY-“63% axodouhmi-G teammamfim 33:33
a 2an

82

Table 10

Diflerences in Test Anxiety in Relation to the MEAP Testing Condition by Demographic

 

 

 

 

 

Variablesa
CTAS Total BASC-2-TA

Measure n z pb n 2 pb

Sex 301 -4.35 <.0005 305 -4.09 <.0005
Ethnicity 302 -.60 .55 306 -1.18 .24
Educational Verification 295 -.73 .46 299 -.91 .36
2 b 2 b

n X P n X P

Grade 305 6.90 .03 309 1 1.46 .003
SES 282 2.23 .69 286 2.69 .61

 

aMEAP: Michigan Educational Assessment Program; CTAS: Children’s Test Anxiety

Scale; BASC-2-TA: Behavioral Assessment Scale for Children, Second Edition, Test
Anxiety Subscale; SES: Hollingshead Socioeconomic Codes: 1: High Class, 2: Middle

Class, 3: Working Class, 4 and 5: Low Class

bBonferroni correction for multiple comparisons: p S .01

83

scores on the MEAP due to demographic characteristic differences

Differences in students’ test anxiety scores due to student demographic
characteristics on the CTAS-Total scale in relation to the MEAP were examined first. A
Mann-Whitney U test revealed a significant difference in CTAS-Total scores of males
(Md= 48, n = 140) and females (Md = 56, n = 161), U= 7998.50, 2 = -4.35,p < .0005.
The effect size of this difference was small, r = -.25. There was no significant difference
in the CTAS-Total scores of Caucasians (Md = 53, n = 250) and Non-Caucasians (Md =
50.50, n = 52), U = 6154.50, 2 = -.60, p = .55. There was no significant difference in the
CTAS-Total scores of general education students (Md = 53, n = 268) and special
education students (Md = 49, n = 27), U = 3308.50, 2 = -.73, p = .46. A Kruskal-Wallis

Test revealed a marginally significant difference in CTAS-Total scores across three grade
levels (Gpl, n = 102: Grade 3, Gp2, n = 115: Grade 4, Gp3, n = 88: Grade 5), 12 (2, n =

305) = 6.90, p = .03. Mann-Whitney U tests between the three grade levels revealed a
significant difference in the CTAS-Total scores of grade 3 students (Md = 50.50, n =
102) and grade 4 students (Md = 56 n = 115), U= 4921,50, z = -2.05,p = .04, and
between grade 4 students (Md = 56, n = 115) and grade 5 students (Md = 49, n = 88), U =
4069.50, 2 = -2.39, p = .02. There was no significant difference between reported test
anxiety between third and fifth grade students. A Kruskal-Wallis Test also revealed no
significant difference in CTAS-Total scores across the five socioeconomic groups (Gpl ,
n = 65: SES 1, Gp2, n = 144: SES II, Gp3, n = 57: SES III, Gp4, n = 13: SES IV, Gp5, n

= 3: SES V), 22 (4, n = 282) = 2.23, p = .69.

Differences in students’ test anxiety scores due to demographic characteristics on

the BASC-2-TA scale in relation to the MEAP were examined second. A Mann-Whitney

84

U test revealed a significant difference in the BASC-2-TA scores of males (Md = 3, n =
144) and females (Md = 5, n = 161), U = 8460, 2 = -4.09, p < .0005. The effect size of
this difference was small, r = -.23. There was no significant difference in the BASC-2-
TA scores of Caucasians (Md = 4, n = 254) and non-Caucasians (Md = 3, n = 52), U =
5923.00, z = -1.18, p = .24. There was no significant difference in the BASC-2-TA scores
of general education students (Md = 4, n = 270) and special education students (Md = 3, n
= 29), U = 3515.50, 2 = -.91, p = .36. A Kruskal-Wallis Test revealed a significant

difference in BASC-2-TA scores across students at the three grade levels (Gpl , n = 103:
Grade 3, Gp2, n = 117: Grade 4, Gp3, n = 89: Grade 5), 12 (2, n = 309) = 11.46, p = .003.

Mann-Whitney U tests between the three grade levels revealed a significant difference in
the BASC-2-TA scores of grade 3 students (Md = 3, n = 103) and grade 4 students (Md =
5 n = 117), U = 4470.5, 2 = -3.32, p = .001. There was no significant difference in
reported test anxiety between fourth and fifth grade or third and fifth grade students.
There was also no significant difference in BASC-2-TA scores across the five
socioeconomic groups (Gpl, n = 65: SES 1, Gp2, n = 147: SES II, Gp3, n = 57: SES Ill,

Gp4, n = 14: SES IV, Gp5, n = 3: SES V), 12 (4, n = 286) = 2.69, p = .61.

Next, differences in students’ test anxiety scores in relation to classroom testing
were examined. Table 11 summarizes the differences in students’ test anxiety scores in
relation to classroom testing due to demographic characteristics. Differences in students’
test anxiety scores on the CTAS-Total scale in relation classroom testing were examined
first. A Mann-Whitney U test revealed a significant difference in the CTAS-Total scores

ofmales (Md= 44, n = 144) and females (Md= 51 n = 171), U= 9382, 2 = -3.64,p <

85

S. .v- Q ”Eammamaoo oar—:8 8.2 2282.80 Eobomcome

.845 33 a wee 4 .820 mega? "m .846 6:622

”N .mmflu :wmI L ”8260 omEocoooomoom umonmwcaom ”mmm ”22220 Sm 28m 5225.. 2mm..- Homeea- ”2335 ©2292

“mo-r .2oEum vacuum 4.22:5 8m 23m EoEmmomm< EoSEom ”<-_--N-Um<m ”2QO 223224 “mo-.- m.:o.6_EU ”me-Ue

 

 

 

 

 

 

E. 82 e8 2. 22 SN 2. was an ma
8. 82. am 8. 5.. mmm em. m2 2: 685
2 N 2 2. R 2 2. N 2
n N n N a m
we. 2... _. a E. 2.- N _ m mm. 8.. wow 8282:”; .262488
8. mn- omm Q. 3.. EM E. mm- :m €255
88v m3- 2». 83v one- am 880 $2..- 3 26m
R N 2 Q N 2 Q N . 2 2:302
e e e
85 firm-63m Lech-QB

 

832.2222 23222M2222Q 3 22.22.8229 M2235 222223.23 22 2 22.223% 2.2 52.2222 Nam-m 2.2 2222222§Q

_ 2 2an

86

.0005. The effect size of this difference was small, r = -.21. There was no significant
difference in the CTAS-Total scores of Caucasians (Md = 47, n = 263) and non-
Caucasians (Md = 46.50, n = 54), U = 6898.50, 2 = -.33, p = .74. There was no significant
difference in the CTAS-Total scores of general education students (Md = 48, n = 278)
and special education students (Md = 44.50, n = 30), U = 3718, 2 = -.7398 p = .33. A
Kruskal-Wallis Test revealed no significant difference in CTAS-Total scores across the

three grade levels (Gpl, n = 115: Grade 3, Gp2, n = 115: Grade 4, Gp3, n =88: Grade 5),
12 (2, n = 318) = 2.73, p = .26. There was also no significant difference in CTAS-Total

scores in relation to classroom testing across the five socioeconomic groups (Gpl , n = 70:

SES 1, Gp2, n = 152: SES II, Gp3, n = 54: SES III, Gp4, n = 12: SES 1v, Gp5, n = 5: SES
V), 12 (4, n = 293) = 7.22,p = .13.

Differences in students’ test anxiety scores on BASC-Z-TA scale in relation
classroom testing were examined next. A Mann-Whitney U test revealed a significant
difference in the BASC-Z-TA scores of males (Md = 2, n = 146) and females (Md = 5, n
= 173), U = 8883, 2 = -4.59, p < .0005. The effect size of this difference was small, r = -
.26. There was no significant difference in the BASC-Z-TA scores of Caucasians (Md =
4, n = 267) and non-Caucasians (Md = 3, n = 54), U = 6996.00, 2 = -.34, p = .73. There
was no significant difference in the BASC-Z-TA scores of general education students
(Md = 4, n = 281) and special education students (Md = 3, n = 31), U = 4180.50, 2 = -.37,
p = .71. A Kruskal-Wallis Test revealed no significant difference in BASC-2-TA scores
across the three grade levels (Gpl, n = 115: Grade 3, Gp2, n = 117: Grade 4, Gp3, n = 90:

Grade 5), 12 (2, n = 322) = 1.01, p = .60. There was also no significant difference in

BASC-Z-TA scores across the five socioeconomic groups (Gpl, n = 70: SES 1, Gp2, n =

87

154: SES II, Gp3, n = 55: SES III, Gp4, n = 13: SES IV, Gp5, n = 5: SES V), 12 (4, n =

297) = 2.18,p = .70.

Differences in students’ test anxiety scores on the TASC scale in relation
classroom testing were examined last. A Mann-Whitney U test revealed a significant
difference in the TASC scores of males (Md = 5, n = 147) and females (Md = 10, n =
171), U = 8710.50, 2 = -4.73, p < .0005. The effect size of this difference was small, r = -
.27. There was no significant difference in the TASC scores of Caucasians (Md = 7, n =
264) and non-Caucasians (Md = 7, n = 56), U = 7061.00, 2 = -.53, p = .60. There was no
significant difference in the TASC scores of general education students (Md = 8, n = 279)
and special education students (Md = 7, n = 32), U = 4236, 2 = -.47, p = .64. A Kruskal-

Wallis Test revealed no significant difference in TASC scores across the three grade
levels (Gpl, n = 116: Grade 3, Gp2, n = 116: Grade 4, Gp3, n = 89: Grade 5), 12 (2, n =

321) = 4.93, p = .09. There was also no significant difference in TASC scores across the
five socioeconomic groups (Gpl, n = 70: SES 1, Gp2, n = 153: SES II, Gp3, n = 55: SES
III, Gp4, n = 13: SES IV, Gp5, n = 5: SES V), 12 (4, n = 296) = 2.20,p = .70.
Aim 2

Hierarchical multiple regression analyses were used to assess the ability of
demographic variables, including student grade, sex, SES, educational verification, and
ethnicity, to predict scores on the different test anxiety measures. Data transformations of
test anxiety scores were completed to account for the non-normal distribution of test
anxiety scores. Following logarithmic data transformations of CTAS-Total variable and
square root data transformations of BASC-Z-TA and TASC variables (Pallant, 2007), the

test anxiety variables were found to more closely meet the assumption of normality.

88

Preliminary analyses were conducted for each regression analysis to ensure no
violation of the assumptions of multiple regression. Tolerance and variance inflation
factor scores (VIFs) were examined and no indications of multicollinearity were
identified (Tols. > .10 and VIFs < 10). Normal probability plots of the regression
standardized residuals and scatterplots of the standardized residuals were examined and
no violations of normality, linearity, homoscedasticity, or independence of residuals were
identified for any of the regression analyses. Table 12 summarizes how predictive
demographic variables were in accounting for variance in test anxiety scores by each test
anxiety measure across both testing conditions. In each regression analysis, sex was
entered in Step 1 (female = 1, male = 0), grade was dummy coded in Step 2: Grade4
(Grade 4 = 1, other = 0) and Grade5 (Grade 5 = 1, other = O), ethnicity was entered in
Step 3 (non-Caucasian = l, Caucasian = 0), educational verification was entered in Step 4
(special education = 1, general education = 0), and socioeconomic status was dummy
coded in Step 5: SES 11(SES-II = 1, other = 0), SES-III (SES-III = 1, other = 0), and
SES-IV—V (SES-IV-IV = 1, other = 0).

The value of demographic variables in predicting test anxiety scores in relation to
the MEAP were considered first. The first regression analysis of CTAS-Total as the
dependent variable indicated that sex explained 6.2% of the variance in CTAS-Total
scores, F (1, 280) = 18.40, p < .0005. After grade was entered in the model, the total
variance explained was 8.6%, R squared change = .024, F change (2, 278) = 3.63, p =
.03. Student ethnicity, educational verification, and SES did not add significantly to the

model. The final model was a significant predictor of CTAS-Total scores in relation to

89

8. v 2., .8. w 2... .88. v 2: £322, 8:28:22... M285 2:: as a a“

 

 

 

 

8. 8. 8. 2. 2. >8, ma
8. 8. .2. 8.- 8. E ma
8.- 8. 8. 8.- 8. = ma
8; 8. E. 8. .83 8. 8.2 8. a? 8. aeesammf
8. oo. 8. a. 8. 8. om. oo. 8.- 8. 8. 8.- 8. co. 8.- 8:82.23,
fiaogosvm+
2. 8. .8. 8. 8. 8. 8. 8. .8.- 82 8. S.- 8. 8. 8.- 8252-7-
wor 3. Nov 3. No: m 0280
8. 8. :. :8. 22. < 882
83 8. 8. 8. m2 8. .82. .8. .288 8. 2258:3889-
:28 8. r8. had-n. 8. ...-cm. .822 8. t2. ...-8.: 8. L2. ...-3.2 8.1.8. 2%
<2 $2 an <2 $2 an <2 ~m< an <2 Nm< an <2 ~m< an .082 2523.
gum/P 2.228% 233.35 .22-22mg DEB--35
38H 82283—0 “588832 m<m§

 

.02-222% 52.2252 825 M2222222a~ 2232.222 2.222.22w2222QK2 2222222M2- 2323:
Q 255-

90

the MEAP, F (8, 273) = 3.80, p < .0005. In the final model, sex and grade were the only
significant predictors of CTAS-Total scores on the MEAP, with sex recording a higher
beta value ([3 = .24, p < .0005) than Grade4 (B = .16, p = .02) or Grade5 (B = -.02, p =
.74) variables.

The second regression analysis of BASC-2-TA as the dependent variable
indicated that sex explained 5.2% of the variance in BASC-Z-TA scores in relation to the
MEAP, F (1, 284) = 15.51, p < .0005. After grade was entered in the model, the total
variance explained was 8.3%, R squared change = .031, F change (2, 278) = 4.77, p =
.009. Student ethnicity, educational verification, and SES did not add significantly to the
model. The final model was a significant predictor of BASC-Z-TA scores in relation to
the MEAP, F (8, 277) = 4.04, p < .0005. In the final model, sex and grade were the only
significant predictors of BASC-Z-TA scores in relation to the MEAP, with sex recording
a higher beta value ([3 = .24, p < .0005) than Grade4 (B = .20, p = .002) or Grade5 (B =
.11, p = .08) variables.

Next, three regression analyses examining the value of demographic variables in
predicting test anxiety measure scores in relation to classroom testing were conducted
using the same five-step model. The first regression analysis of CTAS-Total as the
dependent variable indicated that sex explained 4.1% of the variance in CTAS-Total
scores in relation to classroom testing, F (1 , 291) = 12.29, p = .001. Student grade,
ethnicity, and educational verification did not add significantly to the model. After SES
was entered as the final predictor in the model, the total variance explained was 8.6%, R
squared change = .035, F change (3, 284) = 3.60, p = .014. The final model was a

significant predictor of CTAS-Total scores in relation to classroom testing, F (8, 284) =

91

3.52, p = .001. In the final model, sex and SES were the only significant predictors of
CTAS-Total scores in relation to classroom testing, with sex recording a higher beta
value (B = .19, p = .001) than the SES-II (B = .01, p = .89), SES-III (B = .19, p = .008), or
SES-IV—V (B = .09, p = .18) variables.

The second regression analysis of the BASC-Z-TA as the dependent variable
indicated that sex explained 6.6% of the variance in BASC-2-TA scores in relation to
classroom testing, F (l, 294) = 20.74, p < .0005. Student grade, ethnicity, educational
verification, and SES did not add significantly to the model. The final model was a
significant predictor of BASC-Z-TA scores in relation to classroom testing, F (8, 288) =
3.04, p = .003. In the final model, sex was the only significant predictor of BASC-2-TA
scores in relation to classroom testing, beta value (B = .26, p < .0005).

The third regression analysis of the TASC as the dependent variable indicated that
sex explained 7.0% of the variance in TASC scores in relation to classroom testing, F (1 ,
294) = 22.10, p < .0005. Student grade, ethnicity, educational verification, and SES did
not add significantly to the model. The final model was a significant predictor of TASC
scores in relation to classroom testing, F (8, 287) = 3.84, p < .0005. In the final model,
sex was the only significant predictor of TASC scores in relation to classroom testing,
beta value (B = .26, p < .0005).

Question Five
Aim 1

The relationship between all measures of test anxiety were investigated using

two-tailed Spearman rho correlation coefficient analyses for each testing condition. First,

the relationship between the CTAS and the BASC-Z-TA scales in relation to the MEAP

92

was examined and the findings are summarized in Table 13. All measures were

significantly related to one another at the p < .0005 level. There was a strong, positive
correlation between the CTAS-Total scale and BASC-2-TA subscale, rs = .71, n = 305, p
< .0005. Additionally, there were strong, positive correlations between the CTAS-

Thoughts and CTAS-Autonomic subscales and the BASC-2-TA subscale (rs = .72 and rs

= .64, respectively). The CTAS Off-task subscale was only moderately related to the

BASC-Z-TA subscale (rs = .41).

As expected, while the three subscales of the CTAS were significantly correlated
with one another, correlations between subscales were less strong than the strength of the
relationship between the subscales and the CTAS-Total scale as the subscales were

designed to measure different types of test anxiety symptoms. There was a moderate,
positive correlation between the CTAS-Thought and CTAS-Off-task subscales, rs = .45,
n = 306, p < .0005. There was a moderate, positive correlation between the CTAS-
Autonomic and CTAS-Off-task subscales, rs = .46, n = 305, p < .0005. There was a
strong, positive relationship between the CTAS-Thought and CTAS-Autonomic
subscales, rs = .70, n = 306, p < .0005.

Next, the relationship between the CTAS, BASC-Z-TA, and TASC scales in
relation to classroom testing was examined and the findings are summarized in Table 14.

All measures were significantly related to one another at the p < .0005 level. There was a
strong, positive correlation between the CTAS-Total scale and BASC-2-TA scale, rs =
.73, n = 317, p < .0005. There was also a strong, positive correlation between the CTAS-

Total scale and TASC scale, rs = .77, n = 316, p < .0005. Finally, there was a strong,

93

Table 13

Correlations between Measures of Test Anxiety in Relation to the MEA Pa

 

 

Measure 1 2 3 4 5
l. CTAS-Total 1 .90** .73** .84M .71**
2. CTAS-Thoughts 1 .45* * .70* * .72* *
3. CTAS-Off-task l .46* * .41 * *
4. CTAS-Autonomic l .64* *
5. BASC-Z-TA l

 

aMEAP: Michigan Educational Assessment Program; CTAS: Children’s Test Anxiety

Scale; BASC-2-TA: Behavioral Assessment Scale for Children, Second Edition, Test
Anxiety Subscale

** p< .0005

94

Table 14

Correlations between Measures of Test Anxiety in Relation to Classroom T estinga

 

 

Measure 1 2 3 4 5 6
1. CTAS-Total l .90** .81** .84** .73** .77**
2. CTAS-Thoughts 1 .55M .70” .70** .74**
3. CTAS-Off-task 1 .58M .49** .50**
4. CTAS-Autonomic 1 .69" .74* *
5. BASC-2-TA 1 .78* *
6. TASC 1

 

aMEAP: Michigan Educational Assessment Program; CTAS: Children’s Test Anxiety

Scale; BASC-2-TA: Behavioral Assessment Scale for Children, Second Edition, Test
Anxiety Subscale; TASC: Test Anxiety Scale for Children

** p< .0005

95

positive correlation between the BASC-Z-TA subscale and the TASC scale, rs = .78, n =

318, p < .0005. Additionally, there were strong positive correlations between the CTAS-

Thoughts and CTAS-Autonomic subscales and the BASC-Z-TA scale (rs = .70 and r, =

.69, respectively). There were also strong positive correlations between the CTAS-
Thoughts and CTAS-Autonomic subscales and the TASC scale (rs = .74 and rs = .74,

respectively). There was only a moderate relationship between the CTAS-Off-task
subscale and the BASC-2-TA and TASC scales (rs = .49 and rs = .50, respectively).

As expected, while the three subscales of the CTAS were significantly correlated
with one another, correlations between subscales were less strong than the strength of the

relationship between the subscales and the CTAS-Total scale as the subscales were

designed to measure different types of test anxiety symptoms. There was a moderate,

positive correlation between the CTAS-Thought and CTAS-Off-task subscales, r3 = .55,

n = 319, p < .0005. There was a moderate, positive correlation between the CTAS-
Autonomic and CTAS-Off-task subscales, rs = .58, n = 321, p < .0005. There was a

strong, positive relationship between the CTAS-Thought and CTAS-Autonomic
subscales, rs = .70, n = 318, p < .0005.

Finally, a Kappa Measure of Agreement was calculated for children classified as
high test-anxious versus moderately or low test-anxious using the TASC (Definition 2)
and the CTAS-Total (Definition 1) classification systems. The Kappa Measure of
agreement indicated moderate agreement between the two classification systems, K = .45,

p < .0005. Out of the 65 cases classified as high test-anxious by the TASC, 26 were also

classified as high test-anxious on the CTAS indicating a sensitivity of 40%.

96

Aim 2

The relationship between test anxiety in relation to classroom testing and the
MEAP was investigated using two—tailed Spearman rho correlation analyses for the two
test anxiety measures repeated across conditions. These findings are summarized in Table
15. As hypothesized, all measures were significantly related to one another across testing
conditions at the p < .0005 level. The strongest relationship between reported test anxiety

across the MEAP and classroom testing conditions was found for the CTAS-Total scale,

rs = .73, n = 292, p < .0005. Students’ scores on the CTAS-Total scale in relation to

classroom testing explained 53% of the variance in students’ scores on the CTAS-Total

scale in relation to the MEAP. The correlations between the three subscales of the CTAS

in relation to classroom testing and the MEAP ranged from r, = .66 to .69. The weakest

relationship between reported test anxiety across the MEAP and classroom testing

conditions was found for the BASC-2-TA subscale, r5 = .61, n = 298, p < .0005.

Students’ scores on the BASC-2-TA subscale in relation to classroom testing explained
only 37% of the variance in students’ scores on the BASC-2-TA subscale in relation to

the MEAP.

97

Table 15

Correlation between Measures Assessing Test Anxiety in Relation to

the MEAP and Classroom Testing“

 

 

Measure n rs P
CTAS-Total 292 .73 <.0005
CTAS-Thoughts 294 .66 <.0005
CTAS-Off-task 295 .69 <.0005
CTAS-Autonomic 295 .67 <.0005
BASC-2-TA 298 .61 <.0005

 

alMEAP: Michigan Educational Assessment Program; CTAS:

Children’s Test Anxiety Scale; BASC-TA: Behavioral Assessment

Scale for Children, Second Edition, Test Anxiety Subscale

98

CHAPTER 5
DISCUSSION

The purpose of this study was to critically examine how children in elementary
school perceive high-stakes and classroom testing situations in order to enhance our
limited understanding of the impact of test anxiety in high-stakes testing situations and to
advance our understanding of test anxiety in young children in general. To this end, this
study examined differences in test anxiety reported by children in grades three through
five in relation to classroom testing and high-stakes testing. Additionally, teacher
perceptions of student anxiety, teacher anxiety, and student performance were
investigated. This aspect of the study enhances our understanding of test anxiety by
enabling a comparison between teachers’ perceptions of student test anxiety and students’
self-reported test anxiety. Further, this study examined differences in how children were
classified as low, moderate, or high test—anxious using three different classification
systems. Next, differences in student performance on the high-stakes assessment were
examined between students classified as being low, moderate, or high test—anxious.

In addition, the ability of student demographic characteristics to predict student
test anxiety was examined. Similarly, the ability of student grades, test anxiety, and
demographic characteristics to predict high-stakes test performance was examined.
Finally, the relationship between the three scales used to measure test anxiety was
examined. The current study offers multiple unique contributions to the literature by l)
by directly examining high-stakes test anxiety among young school children, 2)

examining differences in test anxiety by test condition, 3) examining differences in test

99

performance by test anxiety level, and 4) by examining differences in test anxiety
classifications across different measurement tools.
Student Self-Report of Test Anxiety

This study examined differences between students’ self-report of test anxiety in
relation to high-stakes and classroom testing situations. Previous test anxiety literature
suggests that test anxiety is associated with impaired test performance and impaired
knowledge acquisition in academic skill areas (Sarason et al., 1960; Zeidner, 1998). In
the current study, students as a whole reported significantly more test anxiety in relation
to the MEAP assessment than in relation to classroom testing across both measures of test
anxiety repeated for the two testing conditions. There were small, but significant
differences between students’ self-report of test anxiety in relation to the MEAP and
classroom testing for both the CTAS-Total scale and the BASC-Z-TA scale, r = -.21 and
r = -.10, respectively. These results are consistent with the hypothesis that as a whole,
students perceive high-stakes testing situations as more stressful and anxiety-provoking
than typical testing situations that occur as a regular part of the curriculum. Similarly,
students reported significantly more cognitive and physiological symptoms of test anxiety
in relation to the MEAP than in relation to classroom testing. When considered within the
context of previous research that suggests worry is more strongly related to impaired test
performance than emotionality in both laboratory and applied-testing situations
(Hembree, 1988, Holroyd et al., 1978, Liebert & Morris, 1967), this finding indicates that
the significantly greater cognitive and physiological test anxiety symptoms reported by
students in relation to the high-stakes assessment may be associated with performance

impairments.

100

The finding that on the whole young children experienced greater test anxiety in
the high-stakes testing condition as opposed to the classroom testing condition is first of
its kind in the literature. No previous study has directly compared test anxiety across
both high-stakes and typical testing conditions. The only study that offers some data for
comparison is a study comparing a mock high-stakes examination taken as part of test
preparation for an actual high-stakes terminal high-school examination (Putwain, 2008).
However, this study differs significantly from the current study in that students taking the
mock examination were taking the practice “high-stakes” examination for the first time.
In contrast to the current study findings, Putwain found that students taking the mock or
low-stakes examination experienced greater test anxiety than students taking the hi gh-
stakes terminal examination. One possible explanation for the difference in the findings
may be that students in the current study had a depth of experience with classroom testing
situations in comparison to the relatively novel experience of the high-stakes assessment
whereas the mock examination in the Putwain study was a completely novel experience
for the participants. Alternatively, students taking the mock examination in the Putwain
study may have perceived the practice test as having significant stakes associated with it.
While it is not possible to identify a causal mechanism for the differences in reported test
anxiety in either the current study or the Putwain study, the finding in the current study
that as a whole, young children reported significantly more test anxiety about the high-
stakes condition than the classroom testing condition suggests that students perceived the
two testing conditions significantly differently.

Interestingly, students’ average test anxiety scores across both the high-stakes

testing condition and the classroom testing condition fell within one standard deviation of

101

the normative samples for the BASC-Z-TA and the CTAS (Reynolds & Kamphaus, 2004;
Wren, personal communication, March 14, 2008). This finding would suggest that
although there were small and significant increases in test anxiety in relation to high-
stakes testing, these differences did not result in clinically significant increases in test
anxiety for the overall sample. Similarly, students’ average test anxiety scores on the
cognitive and physiological subscales of the CTAS also fell within one standard deviation
of the normative samples (Wren), suggesting that there was no clinically significant
increase in cognitive and physiological symptoms of test anxiety for the overall sample.
In light of the significant difference in overall test anxiety as well as in cognitive and
physiological aspects of test anxiety across the two testing conditions, differences in rates
of low, moderate, and high test anxiety among students are examined following
discussion of teachers’ perceptions of test anxiety among their students. As discussed
above, if the increase in reported test anxiety results in differences in rates of students
who are classified as low, moderate, and high test-anxious across test conditions, this
increase would be significant given the established relationship between test anxiety and
impaired task performance.
Teacher Perceptions of Testing

In addition to examining student self-reported test anxiety, this study examined
teacher perceptions of student test anxiety, enabling direct comparisons to be made
between teachers’ perceptions of student test anxiety and students’ self-reported test
anxiety. This aspect of the current study adds to the existing literature on teacher
perceptions of test anxiety, which is limited by a lack of direct student perception data. In

the current study, teachers reported that they believed students experienced significantly

102

more anticipatory anxiety regarding the high-stakes assessment than classroom testing.
This finding is consistent with multiple studies examining teachers’ perceptions of the
impact of high-stakes testing on children which indicate teachers report heightened
anxiety, stress, pressure, and worry among students due to high-stakes testing programs
(Abrams et al., 2003; Barksdale-Ladd & Thomas, 2000; Jones & Egley, 2004; Jones et
al., 1999). In the current study, this finding is also consistent with the significantly greater
test anxiety reported by students in relation to the high-stakes testing condition versus the
classroom testing condition. Hence, this study indicates that teachers’ perceptions of
students’ response to the different testing conditions were accurate.

In addition, teachers reported that they had significantly more anxiety about how
well students would perform on the high-stakes assessment than classroom testing.
Again, this finding is consistent with multiple studies examining teachers’ psychological
response to high-stakes testing programs which indicate that teachers experience
increased anxiety, stress, and pressure due to high-stakes testing programs and that
teachers change their instructional patterns to focus on test preparation as a result
(Abrams et al., 2003; Barksdale-Ladd & Thomas, 2000; Jones & Egley, 2004; Jones et
al., 1999). The teacher questionnaire also indicated that there was a trend toward teachers
believing that students’ performance on the high-stakes assessment was less
representative of their true ability than their performance on classroom tests. Future
studies examining whether or not teachers’ anxious response to testing affects student
self-reported test anxiety or student performance are warranted.

Test Anxiety Classifications and Rates across Measures

This study also examined differences in test anxiety classifications and prevalence

103

rates of students identified as low, moderate, and high test anxious in the high-stakes
testing and classroom testing conditions. Within testing conditions, there were significant
differences in how the different measures classified students’ test anxiety levels. The
largest difference in classification was found between the CTAS-Total and the TASC
classification systems in relation to the classroom testing condition. More students were
identified as low (53% versus 45%) and high (21% versus 11%) test anxious and fewer
children identified as moderate (26% versus 44%) test anxious using the TASC
classification system than the CTAS-Total classification system, respectively.

This finding demonstrates that test anxiety classifications for the same students
differed significantly depending upon the measure used to assess test anxiety. Using
students’ CTAS-Total score to determine test anxiety levels resulted in 11% of students
being classified as high test-anxious in relation to classroom tests, whereas 21% of
students were classified as high test—anxious using students’ TASC scores. This finding
offers some explanation for how different researchers have identified such disparate rates
of test anxiety in previous research. Previous estimates of high test anxiety rates have
varied widely based on the different methods researchers have used to define clinically
significant levels of test anxiety (Zeidner, 1998). For example, King and Ollendick
(1989) reported that the prevalence of test anxiety among school-aged children may range
from as little as 10% to as much as 30%, with Hill and Wigfield (1984) estimating that
only 10% of children experience impairments in test performance as a result of test
anxiety. Alternatively, Turner and colleagues (1993) reported that the rate of test anxiety
might be as high as 41% among African American children. Since these rates vary so

widely, it is difficult for a researcher determine whether or not these studies are

104

measuring the same test anxiety construct.

This is the first study that directly examines differences in classification systems
for test anxiety. It is clear that prevalence rate estimates of test anxiety cannot be directly
compared across studies using different measures because the established classification
systems for the different measures result in significantly different classifications for the
same subjects. In particular, the TASC classification system established by Beidel and
Turner (1988) appears to stratify significantly more students into the extreme high or low
ranges of test anxiety, with only 26% of students falling outside of these classifications.
Alternatively, 44% of students fell in the moderate, or “average” range using the CTAS-
Total to classify student test anxiety on classroom tests. In the current study, the CTAS-
Total classification system aligns closely with the suggested 10% prevalence rate of
children who experience clinically significant impairment as a result of test anxiety (Hill
& Wigfield, 1984; King & Ollendick, 1989).

Similarly, the classification system that used both CTAS-Total and TASC scores
to classify students’ test anxiety levels was more conservative in identifying children as
either high or low test-anxious than either the CTAS-Total or TASC classification
systems alone and may underestimate the number of students who report low and high
levels test anxiety. In the high-stakes test condition, fewer students were identified as low
(28% versus 32%) and high (4% versus 9%) test anxious and more students identified as
moderate (67% versus 59%) test-anxious with the combined classification system than
CTAS-Total classification system. Similarly, on classroom test condition, fewer students
were identified as low (41% versus 45%) and high (9% versus 11%) test-anxious and

more children identified as moderate (50% versus 44%) test-anxious with the combined

105

classification system than the CTAS-Total classification system.

Given the significant differences in how each classification system identified
students, the need for a standard classification system for identifying impairing levels of
test anxiety is clear. The high test anxiety classification on the TASC was established as
valid by examining students’ scores on the TASC in relation to clinical diagnoses of
anxiety disorders (Beidel & Turner, 1988). Alternatively, the high test anxiety
classification on the CTAS-Total measure was established by determining if students’
scores were significantly greater than the mean of the normative sample (Wren, personal
communication, March 14, 2008). In neither case was the test anxiety classification
made based on impaired student test performance.

In light of previous research indicating that test anxiety is associated with
impaired task performance (Hembree, 1988; Holroyd et al., 1978; Liebert & Morris,
1967), we propose that using impaired student performance would be an appropriate and
theoretically sound criteria for developing a standard classification system for identifying
students as having clinically significant test anxiety. Altemately, the Diagnostic and
Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM-IV-TR)
defines a mental disorder as a clinically significant behavioral or psychological pattern
that is associated with distress or disability (American Psychiatric Association, 2000).
Using this as the criteria for clinical impairment could result in also classifying a student
as clinically test-anxious if they report significant distress in relation to test anxiety
symptoms. These two classification systems may be appropriate for distinguishing
between students who experience psychological distress and impaired testing

performance from those who experience psychological distress but do not experience

106

 

performance impairments. Development and use of one or both of these proposed
classification systems may facilitate screening and decision-making about which students
are at most risk for educational and/or psychological impairment associated with test
anxiety.
Test Anxiety Classifications across Conditions

This study also examined differences in how students were classified as low,
moderate, and high test-anxious across the high-stakes and classroom testing conditions.
This aspect of the study is unique in that it examined how all students responded to
testing using low, moderate, and high classifications, rather than only examining
differences between low and high test-anxious students. In the current study, when
students were classified using their CTAS-Total scores, rates of high test anxiety ranged
from 9 to 11 percent across testing conditions, suggesting that there was no significant
difference in the rate of children identified as highly test-anxious across testing
conditions. However, there were significant differences in the proportions of students
classified as having low, moderate, or high test anxiety. Differences in classification
proportions resulted from more students reporting moderate test anxiety (59% versus
44%) and fewer students reporting low test anxiety (32% versus 45%), in relation to the
high-stakes testing condition than the classroom testing condition. Similarly, when
students were classified on the basis of stable scores across the CTAS-Total and TASC
scales, differences in classification proportions resulted from students more frequently
reporting moderate test anxiety (67% versus 50%) and fewer children reporting low test
anxiety (28% versus 41%) in relation to the high-stakes testing condition than the

classroom testing condition.

107

This is the first study to specifically examine shifts in student test anxiety
classifications across high-stakes and classroom testing conditions. Figure 4 shows
differences in the rates of students identified with the low, moderate, and high test
anxiety across testing conditions using the CTAS-Total classification system.
Additionally, Figure 4 illustrates shifts in how children were reclassified across
conditions by identifying which students reported low, moderate, and high test anxiety in
the high-stakes testing condition according to their test anxiety classification in the
classroom testing condition. As can be seen in Figure 4, 56 students, or 42% of the
students who were classified as low test-anxious during classroom testing situations
reported significantly heightened anxiety in high-stakes testing situations and were
reclassified as moderately test anxious in the high-stakes testing situation. Only 4
students, or 3% of the students who were classified as low test-anxious during classroom
testing were reclassified as highly test-anxious in the high-stakes testing situation.
Similarly, while 71% of the students classified as moderately test-anxious in the
classroom testing condition remained classified as moderately test-anxious in the high-
stakes condition, 14 (11%) were reclassified as high test-anxious and 23 (18%) were
reclassified as low test-anxious. Finally, the least stability was found among the students
who reported high test-anxiety in the classroom testing condition, with 21 (66%) being
reclassified as moderately test-anxious and 1 (3%) being reclassified as low test-anxious.

In sum, although there was some reclassification of students into each of the three
test anxiety levels across conditions, the largest shifi occurred in the number of students
(n = 56) who reported low test anxiety levels in the classroom testing condition and were

then reclassified as moderately test-anxious in the high-stakes testing condition. Overall,

108

£836.50 mime-r $2.52 mcouaocfimfio beta/<- Hmoh .macovam 5 £25 “2 ouswfi

 

 

 

 

 

 

 

 

 

cozmuEmmmU Em_xc< 2mm... a<mz 22385820 32222 amok mczmmh Eco-.mmflu
:9: BEE-.62 26.. :9: mum-6222 26..
. h 2 . 2 F 0
cm - mm
mm
mummk . - om
co <2 :3 D
H
Emu-r co <.r nm S
mum-6222 8: - mm m
n w
mumwh W
:o E :9: l -2: s
- mNH
. - omH

 

 

109

25% of students reported significant increases in test anxiety symptoms in relation to
high-stakes testing and were reclassified into higher test anxiety levels. Of note, however,
is the finding that a subset (15%) of students reported decreases in test anxiety in relation
to the high-stakes testing condition. It is also interesting that although the same
proportion of students reported high test anxiety across testing conditions, the individual
students who made up these groups was different across conditions.

There are several possible implications and interpretations of the finding that on
the whole, students’ reported significant increases in overall test anxiety in relation to
high-stakes testing situations as compared to classroom testing situations. This increase in
test anxiety was primarily accounted for by students who moved fi'om low test-anxious
classifications in the classroom testing situation to moderate test-anxious classifications
in the high-stakes testing situation. Moderate test anxiety was defined as falling within
one standard deviation of the mean of the standardization sample on the CTAS-Total
scale (Wren, personal communication, March 14, 2008) and moderate test anxiety on the
TASC was defined as falling below the clinical cut-point where children were identified
with comorbid anxiety disorders (Beidel & Turner, 1988). Given these classification
definitions, it would be reasonable to hypothesize that the reclassification from low to
moderate test anxiety would not have clinical or educational implications since students
with moderate test anxiety fall in the “average” range. Alternatively, it is possible that
increased test anxiety in the form of more cognitive worries and physiological arousal in
high-stakes testing situations had adverse effects on these students’ performance during
the testing situation and on students’ psychological well-being.

In order to understand the educational implications of students’ heightened test

110

anxiety in relation to high-stakes testing, it is necessary to directly compare students’
performance on the high-stakes test with their reported test anxiety. While previous
research among elementary school children suggests that high test-anxious children have
significantly impaired school achievement and school grades in comparison to low-
anxious children (Hill & Sarason, 1966), achievement differences between moderately
test-anxious children and low and high test-anxious children have not been previously
studied. A unique aspect of the current study is the ability to examine the relationship
between students’ performance on the high-stakes test and students’ test anxiety
classification. By including students classified as low, moderate, and high test-anxious, it
is possible to examine performance differences between moderately test-anxious children
and low and high test-anxious children.
Test Performance Differences across Test Anxiety Levels

As hypothesized, there was a significant association between student performance
on the MEAP and students’ test anxiety classification. Figure 5 illustrates students’
average English Language Arts and Math MEAP performance based on students’ test
anxiety classifications. When students’ CTAS-Total scale scores were used to classify
students as low, moderate, or highly test-anxious, students classified as low test-anxious
outperformed students classified as moderately test-anxious on both the English
Language Arts and Math sections of the MEAP. Low test-anxious students also
outperformed students classified as highly test-anxious on the Math section of the MEAP.
Similarly, when students’ TASC scale scores were used to classify students as low,
moderate, or highly test-anxious, students classified as low test-anxious outperformed

Students classified as moderately and highly test-anxious on both the English Language

111

Low

 

 

 

 

Advanced:
+ English
Language
High ‘ Arts
Proficient + Math
i
Middle ' r '
Proficient Low TA Mod TA High TA
CTAS-Total Classification
Low
Advanced
+ English
Language
~ Arts
High + Math
Proficient .
Middle Low TA Mod TA High TA
Proficient

TASC Classification

Figure 5: Test Performance Levels by Test Anxiety Classification

112

Arts and Math sections of the MEAP. There were no differences found between
moderately and highly test-anxious students’ performance on either the Math and English
Language Arts sections of the MEAP using both classification systems.

These findings have practical significance because they show that students’ test
performance is associated with heightened test anxiety. Perhaps most significant is the
finding that students’ classified as moderately and highly test-anxious performed
significantly less well on the MEAP than students classified as low test-anxious. This
finding highlights the importance of including moderately test-anxious students in the
analyses as these students experienced performance impairments. Combined with the
finding that significantly more students reported moderate test anxiety in relation to high-
stakes testing than classroom testing, this study suggests that up to 68% of students
experience heightened test anxiety and that may be associated with impaired performance
on high-stakes tests in comparison to their low test-anxious peers.

The findings of the current study are supported by previous research examining
the relationship between performance and test anxiety. In a meta-analysis of students
from third grade to postsecondary school, Hembree (1988) found that reading and
mathematics achievement was significantly and inversely related to test anxiety (r = -O.24
and r = -0.22, respectively). Similarly, among a large sample of elementary school
students, Hill and Sarason (1966) found that the strength of the negative correlation
between test anxiety and achievement increased over the elementary school years.
Alternatively, the findings are only partially supported by the only other study in the
literature that specifically examines differences in student performance in relation to test

anxiety classification. Putwain (2008) found that when students were dichotomized into

113

low and high test-anxious groups, high test-anxious students performed significantly
lower than their low test-anxious peers on low and medium-stakes examinations, while
performing slightly better on a high-stakes examination. However, differences in how
low and high test-anxious students performed on three examinations in the Putwain study
may be related to how students perceived the actual stakes of the examinations, making it
difficult to compare these findings directly with the current study findings. Ultimately,
the current study suggests that moderate and high test anxiety was associated with
significantly impaired test performance on the high-stakes assessment, indicating that
reclassification from low to moderate or high test anxiety may have significant
educational implications for students. In light of this finding, it is necessary to understand
how influential test anxiety was in predicting test performance on the high-stakes
assessment.

Association between Test Performance, Test Anxiety, and Student Characteristics
Using multiple regression analyses, the current study examined the unique
influence that test anxiety, student grades, and student demographic characteristics had in

predicting student test performance on the different content areas of the MEAP. The
unique variance in English Language Arts scores explained by student test anxiety ranged
from 1.3% to 3.7% across grades three through five. The unique variance in Math scores
explained by student test anxiety ranged from 3.2% to 10.7% across grades three through
five, suggesting that students’ math performance was more strongly associated with test
anxiety. Finally, the unique variance in Science scores explained by student test anxiety
was 1.2% for students in grade five. When student scholastic achievement was controlled

for, the unique contribution of test anxiety in predicting test performance was significant

114

for third and fifth grade students’ Math scores and fourth grade students’ English
Language Arts scores. Test anxiety was also marginally significant in predicting fourth
grade students’ math scores and fifth grade students’ English Language Arts scores.

Therefore, the hypothesis that test anxiety would be a significant predictor of test
performance even when controlling for student scholastic achievement was partially
supported. Across all analyses, test anxiety accounted for between 1.2 and 10.7% of the
variance in students’ test performance, suggesting that although test anxiety was a
significant predictor of some MEAP outcome measures across grades, the overall impact
that test anxiety had on students’ test performance was small. The finding that test
anxiety significantly contributed to the model of MEAP test performance for some
outcome measures differs from a previous study indicating that pressure and test anxiety
did not significantly contribute to test performance (Mulvenon et al., 2005).
Alternatively, Mulvenon and colleagues’ finding that test anxiety explained only a
relatively small amount of the variance in test performance was supported. In sum,
although the unique contribution of test anxiety to test performance was small in the
current study, test anxiety significantly contributed to students’ test performance.

In the current study, test anxiety contributed significantly to grade three students’
Math MEAP scores and grade three and four students’ English Language Arts MEAP
scores, even after student achievement, student sex, ethnicity, SES, and educational
verification were controlled for. This finding provides further evidence that the impact of
test anxiety on student test performance cannot be explained away by student scholastic
achievement or student demographic variables. When examining the contribution of test

anxiety in predicting test performance across grade levels, it is of interest that as student

115

grade increased, the relationship between student test anxiety and student test
performance decreased. This finding differs from a longitudinal study that found the
strength of the negative correlation between test anxiety and achievement increased over
the elementary school years, with achievement becoming significantly more impaired by
test anxiety as students advanced in elementary school (Hill & Sarason, 1966). In the
current study, it is possible that as students become more experienced with high-stakes
testing in the later elementary school years, the interference experienced as a result of test
anxiety decreases. Further studies examining the developmental nature of the relationship
between test anxiety, performance, and student grade level are needed.

Association between Test Anxiety and Student Characteristics

This study also examined the relationship between test anxiety and student
demographic characteristics. Previous studies of test anxiety suggest that there is a
significant relationship between test anxiety and child characteristics, including age, sex,
race, and socioeconomic status (Hembee, 1988). The current study found that there were
significant differences in test anxiety in relation to the high stakes assessment when
students were classified on the basis of their sex and grade. Student test anxiety on
classroom tests was also significantly different when students were classified by sex but
not by grade.

Contrary to the hypothesis that student characteristics, including ethnicity,
educational verification, and socioeconomic status would significantly predict test
anxiety on the high-stakes assessment since testing scores are reported using these
categories, the current study found that these variables did not significantly predict test

anxiety in relation to the high-stakes testing situation. The finding that there was no

116

relationship between student test anxiety and student ethnicity differs from previous
studies that suggest ethnic minority students report significantly greater test anxiety than
Caucasian students on high-stakes tests (Putwain, 2007) and testing in general (Turner et
al., 1993; Wren & Benson, 2004). Additionally, a meta-analysis examining the effect of
ethnicity and age on test anxiety suggests that African-American children report
significantly more test anxiety than Caucasian children, with greater differences reported
in the elementary years, grades 2 - 4: Effect size (E8) = 0.52 and grades 5 - 8: ES = 0.21
(Hembree, 1988). Similarly, the finding that there was no relationship between student
test anxiety and student SES differs from the Putwain (2007) finding found that teenagers
from low socioeconomic backgrounds reported significantly more test anxiety on hi gh-
stakes tests than students from high socioeconomic backgrounds.

The current study finding of sex differences in reported test anxiety is
substantiated by numerous research studies examining differences in females and males’
reports of test anxiety in relation to testing in general (Hembree, 1988) and high-stakes
testing (Putwain, 2007). A meta-analysis of the literature suggests that gender differences
in test anxiety are also influenced by child age, with moderate differences in grades 5-10
(ES = 0.43) and small differences in lower and upper grades, (grades 1 and 2: ES = 0.14;
grades 3 and 4: ES = 0.28; and grades 11, 12, and postsecondary: ES = 0.27; Hembree,
1988). In the current study, the effect size of the difference in test anxiety reported by
males and females was similar across both high-stakes and classroom testing situations
and all measures, ranging from rs = -.21 to -.27. These effect sizes closely replicate the
findings of the 26 studies of students in grades three and four that were examined in the

meta-analysis conducted by Hembree.

117

Interestingly, in the current study, student grade was associated with differences
in test anxiety in relation to high-stakes testing but not to classroom testing situations.
The current study findings are partially supported by previous research findings on the
effect of student grade on test anxiety, which have examined students’ responses to
testing situations in general rather than to high-stakes testing situations in particular.
Studies of the effect of grade on test anxiety suggest that test anxiety increases over the
course of elementary school (Hembree, 1988, Hill & Sarason, 1966; McDonald, 2001;
Sarason et al., 1960). According to a meta-analysis examining differences in test anxiety
between students in adjacent grades, test anxiety increased through fifth grade before
remaining relatively constant through the end of high school (Hembree). Hembree found
that there while there were increases in anxiety between adjacent grades 3 to 4 and 4 to 5,
the increase was only significant in grade 3 to 4 (ES = -.24, p < .01). Thus, the curr/bnt
study finding that grade was significantly related to test anxiety on the high-stakes
assessment, with students in grade four reporting greater test anxiety than students in
grade three on both the CTAS-Total and BASC-2-TA scales, is supported by the meta-
analytic findings. Alternatively, the finding that grade four students were also more
anxious than students in grade five on the CTAS-Total scale is not supported by prior
research, or the findings of no difference in test anxiety as measured by the BASC-Z-TA
scale between students in these grades.

The finding in the current study that student grade was not associated with test
anxiety in relation to classroom testing differs from the meta-analysis findings. This
difference is of interest in light of the significant differences found between grade and

test anxiety on the high stakes assessment. Previous researchers have suggested that

118

increases in test anxiety over elementary school may be related to increased exposure to
testing over time and to conditioning that occurs following repeated concern about test
failure, pressure to perform well, or feedback about poor performance (King &
Ollendick, 1989). While students in the current study had frequent experience classroom
testing, including weekly spelling tests, tests on math facts, and tests on curriculum
content, they were participating in hi gh-stakes testing for the first, second, or third time,
depending upon their grade level. Thus, the increase in test anxiety reported from grade
three to four may have resulted from the MEAP having more salience for fourth grade
students following the experience of taking the assessment for the first time in third
grade. Further research is needed to clarify the relationship between student grade and
test anxiety between typical and high-stakes testing situations as well as the
developmental trajectory of test anxiety over the course of children’s schooling.

Finally, regression analyses that examined the unique contribution of
demographic variables in predicting student test anxiety suggest that although student
sex, grade, and SES contributed significantly to students’ test anxiety scores on some
measures of classroom tests and high-stakes tests, the total variance in test anxiety
explained by any one of these demographic variables was quite small, ranging from 1 to
7%. This finding suggests that student demographic variables alone are not good
predictors of which students will report high levels of test anxiety in relation to classroom
testing or high-stakes testing. Thus, student characteristics cannot be used alone to
identify students at risk for experiencing significant test anxiety, and direct evaluation of

student test anxiety is warranted.

119

Measures of Test Anxiety
Correlational analyses between the three different measures used to assess test

anxiety in relation to high-stakes and classroom testing conditions indicated that there

was high convergence between all measures, all r, > .70. The high convergence between

all three measures suggests that all are acceptable measures of the psychological
construct of test anxiety. As the gold standard measure of test anxiety, the strong
relationships identified between the TASC scale and the CTAS-Total scale and the
BASC-2-TA subscale supports the concurrent validity of the CTAS-Total scale and the
BASC-2-TA subscale as measures of test anxiety. It is important to note, however, that
the concurrent validity of these scales as measures of test anxiety does not provide any
indication about the validity of the classification systems used to sort students into low,
moderate, or high test-anxious groups based on TASC and CTAS-Total scores.

The biopsychosocial model of test anxiety hypothesizes that test anxiety includes
components of cognitive interference, psychological arousal, and behavioral interference
(Lowe et al., 2008; Zeidner, 1997). The CTAS scale is the only publicly available
measure designed to assess test anxiety in young children that specifically examined all
of these components of test anxiety and provides subscale scores in each of these areas.
Examination of the correlations between these subscales and the BASC-2-TA subscale
and the TASC scale provide an indication of how well these measures assess these three
components of test anxiety. There were strong correlations between the CTAS subscales

assessing cognitive and physiological aspects of test anxiety and the BASC-Z-TA as well

as the TASC scale, .64 5 rs s .74. However, the relationship between the CTAS subscale

measuring behavioral interference and the BASC-2-TA as well as the TASC scale was

120

weaker, .41 S rs S .50. These results suggest that the BASC-2-TA and TASC scales are

better at assessing cognitive and physiological aspects of test anxiety than behavioral
aspects of test anxiety. Thus, these measures are more reflective of the two component,
worry and emotionality conceptualization of test anxiety advanced by Liebert and Morris
in 1967.

Additional consideration of the correlation between the CTAS-Total and the
BASC-2-TA scales across each testing condition is warranted. One potentially important
use of a test anxiety scale could be screening student test anxiety in relation to both
typical classroom testing and high-stakes testing to identify students who might benefit
from interventions designed to decrease test anxiety. While two assessments could be
conducted in an applied educational setting, the need for repeated testing could be
eliminated by identifying a test anxiety scale that assesses typical classroom testing and is
also highly correlated with test anxiety associated with high-stakes testing. In the current
study, while both the CTAS-Total and the BASC-Z-TA measures were significantly
correlated across the testing conditions, the strongest relationship between testing
conditions was on the CTAS-Total scale. Therefore, the best predictor of students’ test
anxiety in relation to the high-stakes assessment was students’ CTAS-Total scale score in
relation to classroom testing.

Thus, in light of current theoretical conceptualizations of test anxiety and the
findings of the current study, the CTAS appears to be the most appropriate measure of the
three used in the current study for assessing biopsychosocial aspects of test anxiety in
young children in relation to both classroom testing in general and high-stakes testing.

While the TASC has been the most extensively used measure of test anxiety in the

121

literature, the current study suggests that the TASC is only moderately correlated with a
measure of behavioral interference. Coupled with concerns about the length and reading
level of TASC items (Wren & Benson, 2004), use of the CTAS scale to assess test
anxiety appears most appropriate. Alternatively, when the goal of assessment is screening
overall test anxiety efficiently and universally within an educational setting, the brevity
of the 7-item BASC-2-TA, coupled with the strong correlation found between the BASC-
2-TA and the CTAS-Total scale make the BASC-Z-TA an attractive and appropriate
option.
Limitations

There were several limitations related to the student sample that confine the
generalizability of this study’s findings. First, this study design involved the examination
of how students from three schools in one Michigan school district responded to
classroom testing and high-stakes testing conditions. As such, the results of this study
may not generalize to schools in other districts in Michigan. Within states, districts and
schools vary in how they approach test preparation and administration, which may affect
students’ responses to high-stakes testing. Additionally, since high-stakes assessments
designed to meet NCLB requirements vary from state to state in multiple ways,
including, but not limited to the mode of administration, the length, and the consequences
attached to the assessments, caution must be exercised in using the current study findings
to draw conclusions about the how students respond to high-stakes testing in other states.
Similarly, the overall sample size of this study is extremely small in comparison to the
number of children enrolled in public schools in the United States who take high-stakes

assessments to comply with NCLB requirements.

122

Although the power to detect small to moderate effect size differences between
student sex, grade, ethnicity, SES, and educational verification, test anxiety, and test
performance was sufficient, further research with a larger, more diverse population of
students is needed to clarify the relationship between these demographic variables, test
anxiety, and test performance in high and low stakes testing situations in young children.
Similarly, further research with a larger sample of students at each grade level is needed
to clarify the relationship between test anxiety, scholastic achievement, demographic
characteristics, and test performance on high-stakes assessments. Additionally, a
limitation of this study was the reliance on parent-report to measure student
socioeconomic status and educational verification. This aspect of the study design may
have resulted in somewhat less accurate reports of student demographic characteristics.
Using school data to identify students’ Free and Reduced Lunch and educational
verification status in future studies may result in more accurate demographic information
about students.

Finally, there were methodological limitations with the teacher perception data
and caution is warranted in interpreting these findings. In addition to lower teacher
participation rates during the second survey, the teacher data were not coded in a way that
allowed for direct comparisons between teachers’ perceptions of classroom testing and
high-stakes testing. Therefore, it was necessary to analyze the data using independent t-
tests rather than repeated measures t-tests. Future studies of perception data should
include paired data. Additionally, by pairing teacher perception data with student data, it
would be possible to examine whether or not individual classroom teachers’ anxious

responses to testing or perceptions of students’ test anxiety affect the performance of

123

their students or students’ self-reported test anxiety. In sum, this study provided a number
of unique additions to the literature on test anxiety among young children in relation to
high-stakes testing and classroom testing; however, caution must be used in interpreting
this study’s findings given the aforementioned methodological limitations.
Clinical and Educational Implications

Several clinical and educational implications related to identifying children at risk
for impairing levels of test anxiety can be gleaned from this study. Perhaps the most
important finding in this study was the significant difference in test anxiety reported by
students between hi gh—stakes and classroom testing. The results of this study suggest that
as a whole, students report significantly more overall test anxiety and more cognitive and
physiological symptoms of test anxiety in response to high-stakes testing than classroom
testing. This increase was associated with more students reporting moderate levels of test
anxiety in the high-stakes condition than the classroom testing condition. This study was
the first to specifically examine differences in students’ test anxiety responses to different
testing conditions, and it adds substantially to the test anxiety literature by demonstrating
that students perceive these two types of testing differently. This difference has
implications for teachers and educators involved in student preparation for high-stakes
tests. By understanding that students may experience different test anxiety symptoms
across different testing conditions, teachers and educators can more effectively prepare
students for these different types of tests.

Additionally, the current study found that students who reported moderate and
high test anxiety in relation to high-stakes testing had significantly lower test

performance in comparison to low test-anxious students. This finding suggests that

124

performance impairments are associated with moderate and high test anxiety and that
there are no differences in performance between moderately and highly test-anxious
students. Hence, the current study provides evidence supporting concerns reported by
teachers that high stakes testing programs result in increased stress, worry, and negative
outcomes for students (Abrams et al., 2003; Barksdale-Ladd & Thomas, 2000; Jones &
Egley, 2004; 2006; Jones et al., 1999).

There is a need for teachers and educators to consider test anxiety during test
preparations given that students report more test anxiety in response to high-stakes testing
and that heightened test anxiety is associated with impaired testing performance.
Educators need to be knowledgeable about intervention programs shown to effectively
decrease students’ test anxiety (e. g., Beidel, Turner, & Taylor-Ferreira, 1999; Ergene,
2003; Tryson, 1980). The finding that students report more test anxiety in relation to
high-stakes testing provides a rationale for universally teaching students some coping
strategies to reduce test anxiety symptoms prior to taking high-stakes tests. Interventions
to reduce test anxiety may be appropriate for the 60 - 70% of the elementary school
student population who are classified as moderately to highly test anxious in relation to
high-stakes testing.

The findings of the current study also have implications for educators interested in
developing efficient systems for identifying which students would most benefit from
more intensive interventions to reduce test anxiety. While the CTAS is a more
theoretically sound instrument that assesses biopsychosocial aspects of student test
anxiety, the BASC-2-TA subscale is highly correlated with this measure and it is

Significantly more brief (7 versus 30 items). The brevity of the BASC-2-TA subscale

125

may make universal test anxiety screening more feasible and acceptable in schools.
Additionally, given the movement towards conducting universal mental health
assessments and interventions in schools (Salmon & Kirby, 2008), the use of the
complete BASC scale could provide educators with a measure of test anxiety as well as
indications of student functioning across internalizing and externalizing areas as well.

As experts in data-based decision-making, school psychologists are uniquely
prepared to organize, execute, and interpret universal screening programs in the school
setting. The response to intervention movement has increased school psychologists’
involvement in universal screenings of students’ academic skills and targeted progress
monitoring of students’ who are identified as having academic skill deficits. These skills
have been shown to transfer to effective universal social-emotional screening and
intervention programs (e.g., Cheney, Flower, & Templeton, 2008). Similarly, universal
screenings of test anxiety and progress monitoring the effectiveness of any programs
designed to decrease students’ test anxiety symptoms in response to different testing
conditions could be completed by school psychologists.

Future Research

The results of this study suggest that as a whole, elementary school children
experience significantly more test anxiety in relation to high-stakes assessments than
classroom testing. Examination of individual students’ test anxiety classifications across
testing conditions revealed that 60% of students reported stable test anxiety levels across
testing conditions, 25% reported significant increases in test anxiety level, and 15% of
Students reported significant decreases in test anxiety. These findings add nuance to

previous research, which have been limited by methodological weaknesses and report

126

equivocal findings on children’s responses to high-stakes testing. Thus, this study adds to
this limited research base and suggests that while as a whole, students report significantly
more test anxiety in relation to high-stakes assessments, individual students had differing
responses to the different testing conditions. Further research is needed to clarify which
students are most at risk for experiencing increased test anxiety in relation to high-stakes
testing and how these students differ from those who reported less test anxiety in relation
to high-stakes testing.

The finding that a large proportion of students (42%) who reported low test
anxiety in relation to classroom testing also reported moderate test anxiety in relation to
high-stakes testing takes on particular significance in light of the finding that moderately
test-anxious students performed significantly less well on the high-stakes assessment than
low test-anxious students. Future research studies that expand upon the current study
findings and specifically examine the effect of low, moderate, and high test anxiety on
test performance are needed. It will also be important for researchers to expand upon the
findings of the current study and existing literature to clarify the relationship between
increased test anxiety, test performance, and testing stakes. Further research that
specifically examines differences between students’ test anxiety responses to testing with
varying stakes is needed. Moreover, future research should examine whether or not the
impact that test anxiety has on testing performance differs depending upon the stakes of
the test. Similarly, replications and extensions of the current study are needed to clarify
whether or not increases in test anxiety on annual tests required by NCLB tests exist
across different states and more diverse samples of students. Similarly, the relationship

between test anxiety and test performance needs to be examined across more diverse and

127

larger samples of students. Future studies should involve students from different states,
districts with and without test preparation programs, and diverse socioeconomic and
ethnic backgrounds.

Additionally, the biospychosocial model of test anxiety supports the construct
validity of the Children’s Test Anxiety Scale (CTAS). However, as a recently developed
measure that has only been studied by the test developer, additional examination by an
independent researcher is needed to verify the psychometric properties of the CTAS.
Similarly, there is a need to develop a classification system for the CTAS that can be
used to identify children at risk for performance impairments or significant psychological
distress in relation to testing. The current suggested cutpoints for low, moderate, and
high test anxiety are based on the assumption that test anxiety is a normally distributed
construct; however, findings from the current study suggest that test anxiety is a
positively skewed construct. Therefore, identification of cutpoints based on verification
of impairment using a measure of test performance would be more appropriate than using
scores above or below one standard deviation of the mean of the normative sample to
determine test anxiety classification. Similarly, further examination of the feasibility,
acceptability, and utility of the CTAS and BASC—Z-TA as measures of student test
anxiety are needed.

Finally, test anxiety intervention research is needed to develop effective
interventions for test-anxious students that can be practically implemented into
educational settings. Future research examining the effectiveness, feasibility, and
acceptability of interventions designed to decrease students’ test anxiety and increase

Students’ coping skills are needed. Studies are needed that not only examine how

128

effective test anxiety interventions are in reducing test anxiety, but also examine how
effective test anxiety interventions are in decreasing performance impairments associated

with test anxiety.

129

APPENDICES

130

Appendix A:

Demographic Information Form

131

DEMOGRAPHIC INFORMATION FORM

 

 

 

 

Child Name: Child’s Date of birth:
Sex of Child: _ Male Child’s Age:
__ Female Child’s Grade:
Child’s ethnicity/race:
_ Caucasian _ African American
__ Mexican, Mexican-American _ Other Latino or Hispanic
_ American Indian _ Asian
_ Pacific Islander _ Other (describe)

 

Does your child receive special educational services for have any of the following?
(Please circle)

Learning problem Yes / No Vision or hearing problem Yes / No
Cognitive Delay Yes / No Language Delay Yes / No
ADHD Yes / No Physical handicap Yes / No

Emotional/behavioral problem Yes / No

Caregiver Information (i.e. mother, father, stepparent, adoptive parent, partner, etc. who
are responsible for caring for the child):

 

 

 

 

Caregiver #1

Highest level of education completed: Current Employment:

__ Grades 0-8 _ Yes, full time (describe)

_ Grades 9-11 _ Yes, part time(describe)

_ High School or GED _ Not working (receiving gov. assistance)
__ Some college (1 year or more) _ Not working by choice

_ College graduate

_ Post-college

Caregiver #2

Highest level of education completed: Current Employment:

_ Grades 0-8 _ Yes, full time (describe)

__ Grades 9-11 _ Yes, part time(describe)

_ High School or GED _ Not working (receiving gov.assistance)
_ Some college (1 year or more) _ Not working by choice

_ College graduate

_ Post-college

132

Appendix B:

Academic Performance Form

133

ACADEMIC PERFORMANCE FORM

Third-Grade Grades from Fall 2008

Reading: Level
Decoding
Comprehension
Fluency

 

 

 

 

Writing: Ideas
Organization
Word Choice
Conventions
Spelling

 

 

 

 

 

Mathematics: Numbers/Operations
Measurement
Geometry
Data/Problem-solving

 

 

 

 

Fourth and F ifth-Grade Grades from Fall 2008

Reading

Writing
Mathematics
Science (5th only)

 

 

 

 

MEAP Scores and Levels
Reading
Writing
ELA
Mathematics
Science (5“1 only)

 

 

 

 

 

DIBELS
Benchmark 1 ORF

 

134

Appendix C:

Teacher Hi gh-Stakes Testing Questionnaire

135

On this page there are a series of questions about your perceptions of MEAP testing.
Your responses to these questions are an indication of your consent for this information
to be used as part of a research study examining test anxiety among elementary school
children. You have the right to not participate in this study or to not answer any
individual questions. All information will be kept confidential and no identifying
information will be linked to this data.

School Name

 

 

 

 

 

 

 

Grade Taught
In general, NOT AT SOMEWHAT QUITE A VERY
ALL BIT
How anxious were your students about the MEAP
assessment before they took it? I 2 3 4
How anxious were your students about the MEAP
assessment during the assessment? 1 2 3 4
low anxious were 1% about how well your students
Id do on the MEAP assessments? l 2 3 4

 

 

 

 

 

 

 

Very much Somewhat At their Somewhat Very much

 

 

 

 

 

 

 

 

 

In general, below their below ability above their above their
ability their ability ability
ability
How well do you think your students
performed on the MEAP assessment? 1 2 3 4 5
Comments:

136

Appendix D:

Teacher Classroom Tests Questionnaire

137

On this page there are a series of questions about your perceptions of classroom testing
practices. Your responses to these questions are an indication of your consent for this
information to be used as part of a research study examining test anxiety among
elementary school children. You have the right to not participate in this study or to not
answer any individual questions. All information will be kept confidential and no
identifying information will be linked to this data.

School Name

 

Grade Taught

 

For the questions that follow, class tests are defined as any graded assessment of
scholastic skills that occurred during the past 3 weeks. This may include spelling tests,
quizzes or tests in content areas (such as math, science, social studies), or any other
formal assessment of student learning.

1. Please list the names of all class tests that have occurred in the last 3 weeks

¥

 

 

 

 
   
 
 
  
    

 

 

 

In general, NOT AT SOMEWHAT QUITE A VERY
ALL BIT
2- How anxious were your students about class
tests be fore they took them? 1 2 3 4
3- HOW anxious were your students about class
tests during the tests? I 2 3 4
' HOW anxious were ygu about how well your
students would do on classroom tests? 1 2 3 4

 

 

 

 

 

Very much Somewhat At their Somewhat Very much

 

 

 

 

 

 

 

In general, below their below their ability above above their
ability ability their ability
ability
' HOW well do you think your students
perfol‘l‘l'led on classroom tests? 1 2 3 4 5
comments:

138

 

 

Appendix E:

Parental Information Letter

139

 

COLLEGE OF
EDUCATION

Department of Counaellng,
Educational Psychology,
and Special Education

Natasha K. Segool, MA.
Michigan State University
4016 Erickson Hall

East Lansing, Michigan
48824-1034

segoolna@msu.edu

It IS! I is an affirmative-action.
equal opportunity institution.

MICHIGAN STATE

UNIVERSITY

 

September _, 2008

Dear Parent/Guardian:

We are writing to invite your child to participate in a research study
examining how students feel about test taking. We have partnered

 

with Elementary School and all of the third, fourth, and fifth
grade students at are being invited to participate in
this study.

We hope to learn how elementary school students feel about test
taking and in particular, how they feel about the MEAP. We also hope
to learn if students' feelings impact their performance on tests.

The study will involve your child completing two brief surveys about
his/her feelings about tests. The surveys will be completed during
non-academic time and your child will receive a small prize and snack
as a thank you. The surveys will be completed for research purposes
only and the results will be kept confidential.

We have enclosed a consent form for you to fill out in order for your
child to participate in this research study. Please complete the
consent and demographic forms and return them to your child's
classroom teacher by , 2008.

If you have any questions, please feel welcome to contact the
researchers, Natasha K. Segool, by phone: 517-231-9116 or email:
segoolna@msu.edu, or John S. Carlson, by phone: 517432-0843;
email: carlsoj@msu.edu.

Thank you,

Natasha K. Segool, MA. John S. Carlson, PhD.

 

 

140

Appendix F:

Consent Form

141

Research Partlcigant Information and Consent Form

Your child is being asked to participate in a research project. Researchers are required to
provide a consent form to inform you about the study, to convey that participation is
voluntary, to explain risks and benefits of participation, and to empower you to make an
informed decision. You should feel free to ask you the researchers any questions may have.

Study Title: Test Anxiety Associated with High-Stakes Testing among Elementary School
Children: Prevalence, Predictors, and Relationship to Student Performance

Researcher and Title: John Carlson, Ph.D., Associate Professor
Natasha Segool, MA, Doctoral Candidate

Department and Institution: Department of Counseling, Educational Psychology, and Special
Education, Michigan State University

PURPOSE OF RESEARCH:

Your child is being asked to participate in a research study examining how students feel
about test taking and whether or not these feelings affect their performance on tests.
Students are being asked how they feel about taking classroom tests and the Michigan
Educational Assessment Program (MEAP) tests. Your child has been selected as a possible
participant in this study because he/she is taking the MEAP test this fall. All of the third,
fourth, and fifth grade students at your child’s school are being invited to participate in this
study. In the entire study, approximately 500 students are being asked to participate in the
study. Your child’s participation in this study will take about a total of forty minutes. From this -
study, the researchers hope to learn how elementary school students feel about test taking
and in particular, how they feel about the MEAP. We also hope to learn if students’ worries
impact their performance on tests.

WHAT YOU AND YOUR CHILD WILL DO:

You will be asked to fill out a brief demographic form about your child. Your child will
complete two surveys about his/her feelings about tests. The first survey will be completed
after your child finishes taking the MEAP. The second survey will be completed about three
weeks later. Each survey will take fifteen to twenty minutes to complete. The surveys will be
read aloud to students in the classroom and students will complete the surveys individually.
The MEAP is a routine part of your child’s academic work. The researchers will collect data
on your child’s performance on this assessment as well as your child’s academic
performance in the prior academic year. The surveys will be completed for research
purposes only and the results will not be shared with you, your child, the teacher, or the
school.

POTENTIAL BENEFITS:

Your child will not directly benefit from your participation in this study, with the possible
exception that answering the questions could be a positive experience for you child. Your
child’s participation in this study may contribute to the understanding of how children feel
about testing. This research, along with future research, may increase our knowledge of test
anxiety and its relationship to school achievement, thus potentially benefiting children with
test anxiety in the future. Following the completion of the study, you will receive information
on the result of the study if you so wish. These will be aggregated results, and not results of
your child as an individual.

142

POTENTIAL RISKS:

This study poses no more than minimal risk for your child, although there is the potential for
psychological discomfort. We will ask your child to complete surveys about his/her thoughts,
feelings, and behaviors before, during, and after tests. Answering some of the questions may
cause your child to experience discomfort or distress. The researchers will be available to
answer children’s questions during and after the surveys.

PRIVACY AND CONFIDENTIALITY:

The data for this project will be kept confidential to the greatest extent allowable by law. It is
important for you to know, however, that the researchers are required to report evidence of
child abuse. Although the researchers will know your child’s identify and your child’s name
will be on some documentation during the study, all identifying information will be removed
as soon as the study is complete. At that time, an identification number will be assigned to
your data and your child’s name will be removed from all paperwork. Completed surveys will
be kept in a locked file cabinet in the office of one of the researchers. All documents will be
destroyed ten years after study completion. The results of this study may be published or
presented at professional meetings, but the identities of all research participants will remain
anonymous. It will not be possible for readers to know who participated in the study.

YOUR RIGHTS TO PARTICIPATE, SAY NO, OR WITHDRAW

Participation in this research project is completely voluntary. You and your child have the
right to say no. You or your child may change your minds at any time and withdraw from the
study. You or your child may also choose not to answer specific questions or to stop
participating at any time. Whether or not you or child chooses to participate will have no
effect on your child’s grade.

COSTS AND COMPENSATION FOR BEING IN THE STUDY:
It does not cost anything to participate in this study. Your child will receive a small
prize/snack following the completion of each survey.

CONTACT INFORMATION FOR QUESTIONS AND CONCERNS

If you have any questions about this study, such as scientific issues, how to do any part of it,
or to report an injury, please contact the researchers, Natasha Segool, by phone: 517-231-
9116; email: segoolna@msu.edu or John S. Carlson, by phone: 517-432-0843; email:
carlsoj@msu.edu or regular mail: 431 Erickson Hall, East Lansing, MI 48824. If you have
questions or concerns about your role and rights as a research participant, or would like to
register a complaint about this study, you may contact, anonymously if you wish, the
Michigan State University's Human Research Protection Program at 517-355-2180, Fax 517-
432-4503, or e-mail irb@msu.edu or regular mail at 202 Olds Hall, MSU, East Lansing, MI
48824

DOCUMENTATION OF INFORMED CONSENT.
Please select a box, fill in your child’s name, and sign below.

 

 

 

 

Cl Yes, my child may participate in this research study
[:1 No, my child may not participate in this research
study.

Parent Signature Date

143

Appendix G:

Survey One

144

Survey One Code #

Student Name

 

School

 

Teacher Name

 

Grade

 

We are from Michigan State University. We would like to learn more about children’s
feelings about tests. To do this, we are asking you and other children to take part in a
research study. In this study, we will ,be asking you some questions about your feelings
and behaviors.

If you agree to be part of the study, you will be asked to answer some questions that will
take about 20 minutes. You do not have to answer any of the questions if you do not want
to or if they make you feel uncomfortable.

No one will know the information belongs to you and the results of the research will not
be able to be connected to you. Only information that you tell us that we think is
dangerous to your health or others will be shared with your parents or others who can
help keep you or others safe.

Name Date

 

I45

Survey One

How I Feel about the MEAP

Code #

Directions: Listen to each question as it is read aloud. Circle one answer for

each question.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Whi'9 ' am taking the MEAP ' ' ' ALMOST 08.91MHEE Ofiql'sI-ITE ALMOST
NEVER TIME TIME ALWAYS

1. lwonder if I will pass. 1 2 3 4

2. My heart beats fast. . 1 2 3 4

3. I look around the room. 1 2 3 4

4. lfeel nervous. 1 2 3 4

5. I think I am going to get a bad score. 1 2 3 4

While I am taking the MEAP . . . ALMOST 319;“; ofififla ALMOST
NEVER TIME TIME ALWAYS

6. It is hard for me to remember the 1 2 3 4

answers.

7. I play with my pencil. . 1 2 3 4

8. My face feels hot. 1 2 3 4

9. Iworry about falling. . 1 2 3 4

10. My belly feels funny. 1 2 3 4

While I am taking the MEAP ' ' ' ALMOST $91qu5 Olgql'sHl-E ALMOST
NEVER TIME TIME ALWAYS

11. I worry about doing something wrong. 1 2 3 4

12. I check the time. . 1 2 3 4

13. I think about what my score will be. . 1 2 3 4

14. I find it hard to sit still. 1 2 3 4

15. I wonder if my answers are right. 1 2 3 4

 

146

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Survey One Code # __
Whi'e ' am taking th° MEAP ' ' ' ALMOST 53$: 0'43ng ALMOST
NEVER TIME TIME ALWAYS
16. I think that I should have studied 1 2 3 4
more.
17. My head hurts. 1 2 3 4
18. I look at other people. 1 2 3 4
19. I think most of my answers are wrong. 1 2 3 4
20. lfeel warm. 1 2 3 4
While I am taking the MEAP . . . ALMOST 519;“; $15,175 ALMOST
NEVER TIME TIME ALWAYS
21. I worry about how hard the MEAP is. 1 2 3 4
22. I try to finish up fast. 1 2 3 4
23. My hand Shakes. 1 2 3 4
24. I think about what will happen if I fail. 1 2 3 4
25. I have to go to the bathroom. 1 2 3 4
While I am taking the MEAP . . . ALMOST 3%; £315,175 ALMOST
NEVER TIME TIME ALWAYS
26. ltap my feet. . 1 2 3 4
27. I think about how poorly I am doing. . 1 2 3 4
28. lfeel scared. . 1 2 3 4
29. I worry about what my parents will
say. . 1 2 3 4
30. I stare. 1 2 3 4

 

 

 

 

 

147

 

Survey One Code #

TRUE FALSE

 

 

 

 

 

 

 

 

 

 

 

 

 

31. I hate taklng the MEAP. .1. F
32. I worry about the MEAP more
than my classmates do. . . . T F
33. NO matter how much I study for T F
the MEAP, I am afraid I will fail. .
ALMOST

NEVER SOMETIMES OFTEN ALWAYS
34. Even when I try hard, I fall. N S O A
[3‘5éiget nervous when I take the N S O A
36. When I take the MEAP, I can’t
think. N S O A
37. I have trouble sleeping the night
before the MEAP. N S 0 A

 

 

 

 

 

148

 

Appendix H:

Survey Two

149

Survey Two Code #

Student Name

 

School

 

Teacher Name

 

Grade

 

We are from Michigan State University. We would like to learn more about children’s
feelings about tests. To do this, we are asking you and other children to take part in a
research study. In this study, we will be asking you some questions about your feelings
and behaviors.

If you agree to be part of the study, you will be asked to answer some questions that will
take about 20 minutes. You do not have to answer any of the questions if you do not want
to or if they make you feel uncomfortable.

No one will know the information belongs to you and the results of the research will not
be able to be connected to you. Only information that you tell us that we think is
dangerous to your health or others will be shared with your parents or others who can
help keep you or others safe.

Name Date

 

150

Survey Two

How I Feel about Tests

Code #

Directions: Listen to each question as it is read aloud. Circle one answer for

each question.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

While I am taking tests . . . SOME MOST
ALMOST OF THE OF THE ALMOST
NEVER TIME TIME ALWAYS
1. I wonder if I will pass. 1 2 3 4
2. My heart beats fast. . 1 2 3 4
3. I look around the room. 1 2 3 4
4. I feel nervous. 1 2 3 4
5. I think I am going to get a bad grade. 1 2 3 4
While I am taking tests . . . SOME MOST
ALMOST OF THE OF THE ALMOST
NEVER TIME TIME ALWAYS
6. It is hard for me to remember the 1 2 3 4
answers. . . . . .
7. I play with my pencil. . 1 2 3 4
8. My face feels hot. 1 2 3 4
9. I worry about failing. . 1 2 3 4
10. My belly feels funny. 1 2 3 4
While I am taking tests . . . SOME MOST
ALMOST OF THE OF THE ALMOST
NEVER TIME TIME ALWAYS
11. I worry about doing something wrong. 1 2 3 4
12. I check the time. . 1 2 3 4
T3. I think about what my grade will be. . 1 2 3 4
T4. I find it hard to Sit still. 1 2 3 4
E. I wonder if my answers are right. 1 2 3 4

\

 

 

 

 

 

151

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Survey Two Code #
While I am taking tests . . . SOME MOST
ALMOST OF THE OF THE ALMOST
NEVER TIME TIME ALWAYS
16. I think that I should have studied more.
1 2 3 4
17. My head hurts. 1 2 3 4
18. I look at other people. 1 2 3 4
19. I think most Of my answers are wrong. 1 2 3 4
20. lfeel warm. 1 2 3 4
While I am taking tests . . . SOME MOST
ALMOST OF THE OF THE ALMOST
NEVER TIME TIME ALWAYS
21. I worry about how hard the test is. 1 2 3 4
22. I try to finish up fast. 1 2 3 4
23. My hand Shakes. 1 2 3 4
24. I think about what will happen if I fail. 1 2 3 4
25. I have to go to the bathroom. 1 2 3 4
While I am taking tests . . . SOME MOST
ALMOST OF THE OF THE ALMOST
NEVER TIME TIME ALWAYS
26. I tap my feet. . 1 2 3 4
27. I think about how poorly I am doing. 1 2 3 4
28. lfeel scared. . 1 2 3 4
29. I worry about what my parents will say. 1 2 3 4
30. I stars. 1 2 3 4

 

 

 

 

 

 

152

 

Survey Two

 

 

 

 

a test, I am afraid I will fail.

 

 

TRUE FALSE
31. I hate taking tests. T F
32. I worry about tests more than
my classmates do. T F
33. No matter how much I study for T F

 

 

Code # __

ALMOST

NEVER SOMETIMES OFTEN ALWAYS

 

34. Even when I try hard, I fail.

 

 

 

night before a big test.

 

 

 

 

 

N S O A
35. I get nervous when I take tests. N S O A
36. When I take tests, I can’t think. N S O A
37. l have trouble sleeping the N S O A

 

153

 

Survey Two Code #

 

38. Do you worry when the teacher says that she is going to ask

 

 

 

 

 

 

 

 

 

 

 

 

 

 

you questions to find out how much you know? Yes NO
39. Do you worry about being promoted, that is, passing from the
_ grade to the _ grade at the end of the year? Yes No
40. When the teacher asks you to get in front of the class and read
aloud, are you afraid that you are going to make some bad Yes NO
mistakes?
41. When the teacher says that she is going to call upon some
boys and girls in the class to do arithmetic problems, do you hope Yes No
that she will call upon someone else and not on you?
42. Do you sometimes dream at night that you are in school and
cannot answer the teacher’s questions? Yes NO
43. When the teacher says that she is going to find out how much
you have learned, does your heart begin to beat faster? Yes NO
44. When the teacher is teaching you about arithmetic, do you feel
that other children in the class understand her better than you? Yes NO
45. When you are in bed at night, do you sometimes worry about
how you are going to do in class the next day? Yes NO
46. When the teacher asks you to write on the blackboard in front
of the class, does the hand you write with sometimes shake a Yes NO
little?

147. When the teacher is teaching you about reading, do you feel
that other children in the class understand her better than you? Yes NO

' 9

48. Do you worry more about school than other children. Yes N O
49. When you are at home and you are thinking about your
arithmetic lesson for the next day, do you become afraid that you Yes NO
will get the answers wrong when the teacher calls on you?
50. If you are sick and miss school, do you worry that you will do
more poorly in your schoolwork than other children when you Yes NO
return to school?
51. Do you sometimes dream at night that other boys and girls in
your class can do things you cannot? Yes NO

 

 

154

 

 

Mou are home and you are thinking about your readifigde # __

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

lesson for the next day, do you worry that you will do poorly on the Yes NO

lesson?

53. When the teacher says that she is going to find out how much you

have learned, do you get a funny feeling in your stomach? Yes NO

54. If you did very poorly when the teacher called on you, would you

probably feel like crying even though you would try not to cry? Yes No

55. Do you sometimes dream at night that the teacher is angry

because you do not know your lessons? Yes No
- '7

56. Are you afraid of school tests. Yes N O

9

57. Do you worry 3 lot before you take a test. Yes N o

58. Do you worry a lot while you are taking a test? Yes N o

59. After you have taken a test do you worry about how well you did

on the test. Yes NO

60. Do you sometimes dream at night that you did poorly on a test

you had in school that day? Yes No

61. When you are taking a test, does the hand you write with shake a

little? Yes No

62. When the teacher says that she is going to give the class a test,

do you become afraid that you will do poorly? Yes NO

63. When you are taking a hard test, do you forget some things you

knew very well before you started taking the test? Yes NO

64. Do you with a lot of times that you didn’t worry so much about

tests? Yes No

65. When the teacher says that she is going to give the class a test,

do you get a nervous or funny feeling? Yes NO

66. While you are taking a test do you usually think you are doing

poorly? Yes No

67. While you are on your way to school, do you sometimes worry

that the teacher may give the class a test? Yes NO

 

155

 

Appendix 1:

Survey Directions

156

Directions read by examiner prior to survey administration*

 

My name is . I’m here from Michigan State University
because I am interested in learning about children’s feelings about tests. To do this, we
are asking you and other children to take part in a research study. In this study, we will be
asking you some questions about your feelings and behaviors.

If you agree to be part of the study, you will be asked to answer some questions
that will take about 20 minutes and then we will give you as a thank you!
You do not have to answer any of the questions if you do not want to or if they make you
feel uncomfortable.

No one will know the information belongs to you and the results of the research
will not be able to be connected to you. Only information that you tell us that we think is
dangerous to your health or others will be shared with your parents or others who can
help keep you or others safe.

If you are interested in helping us learn how students feel about tests today, please
write your name at the bottom of the page and write the date (write the date on the
Mboard for students to copy). Also, please write the name of your school, teacher’s
name, and grade on this page (write info on chalkbofiard for students to conv). Give
students time to finish reading the assent form you just summarized and to fill out
the first page.

The questions I’m going to ask you are different from the questions you are
normally asked in school because there are no right or wrong answers. The questions are
about how you think and feel, and therefore, they have no right or wrong answers. People

think and feel differently, so the person sitting next to you may put down one answer and

157

you might put down another answer. For example, if I asked you this question: “Do you
like to play ball?” some of you would answer “yes” and some of you would answer “no.”
Your answer depends on how you think and feel.

These questions are about how you thought and felt while you were taking [n_ang
test condition]. For these questions, think about how felt while you took [name test
condition]. You should listen to each question as I read it aloud and then put a circle
around your answer. Remember, listen carefully to each question and answer it by
deciding how you think and feel. If you don’t understand a question, raise your hand and

ask me about it.

Adapt order of next paragraphs based on the order of questionnaires
BASC-Z-TA

For the first few questions, I will say a statement and you have to decide if you
agree or disagree. If you agree, circle T for True. If you disagree, circle F for False. Read
Questions 1-3.

For the next four questions, I will say a statement and you have to decide if you
feel that way never, sometimes, often, or almost always. Circle 1 if you want to answer
never, 2 if you want to answer sometimes, 3 if you want to answer often, and 4 if you
want to answer almost always. Read Questions 4-7.

CTAS

For the next set of questions, 1 will say a statement and you have to decide if you

feel that way almost never, some of the time, most of the time, or almost always. Circle 1

if you want to answer almost never, 2 if you want to answer some of the time, 3 if you

158

want to answer most of the time, and 4 if you want to answer almost always. Now, turn

the page to here’s the first statement. Question 1: Read

 

Questions I -30 and add the “While I am taking [test condition] ” starter to each
statement. Remind studena to turn the page after #15.
TASC

For the last set of questions, I will say a statement and you have to decide if you
agree or disagree. If you agree, circle Yes. If you disagree, circle No. Read Questions I-

30.

*Directions adapted from Sarason et al., 1960.

159

 

REFERENCES

160

Abrams, L. M., Pedulla, J. J ., & Madaus, G. F. (2003). Views from the classroom:
Teachers’ opinions of statewide testing programs. Theory into Practice, 42, 18-
29.

American Educational Research Association. (2000). AERA position statement
concerning high-stakes testing in PreK-12 education. Retrieved November 18,
2007 from http://www.aera.net/about/po[icy/stakes.htm.

American Psychiatric Association. (2000). Diagnostic and statistical manual of mental
disorders, 4th edition, Text revision. Washington, DC: American Psychological
Association.

Barksdale-Ladd, M. A., & Thomas, K. F. (2000). What’s at stake in high-stakes testing:
Teachers and parents speak out. Journal of Teacher Education, 51, 384-397.

Bednarek, D. M. (1985). A study of test anxiety and helplessness-mastery orientation
under experimental conditions of success and failure (Doctoral dissertation,
University of Pittsburgh, 1984). Dissertation Abstracts International, 46,
DA8525835.

Beidel, D. C., Turner, M. W., & Trager, K. N. (1994). Test anxiety and childhood anxiety
disorders in African American and White school children. Journal of Anxiety
Disorders, 8, 169-179.

Beidel, D. C., & Turner, S. M. (1988). Comorbidity of test anxiety and other anxiety
disorders in children. Journal of A bnormal Child Psychology, 16, 275-287.

Beidel, D. C., Turner, S. M., & Taylor-Feneira, J. C. (1999). Teaching study skills and
test-taking strategies to elementary school students. Behavior Modification, 23,
630-646.

Bodas, J ., & Ollendick, T. H. (2005). Test anxiety: A cross-cultural perspective. Clinical
Child and Family Psychology Review, 8, 65-88.

Brown, D., Galassi, J. P., & Akos, P. (2004). School counselors’ perceptions of the
impact of high-stakes testing. Professional School Counseling, 8, 31-39.

Cheney, D., Flower, A., & Templeton, T. (2008). Applying Response to Intevention
metrics in the social domain for students at risk of developing emotional or
behavioral disorders. The Journal of Special Education, 42, 108-126.

C izek, G. J. (2001) More unintended consequences of high-stakes testing. Educational
Measurement: Issues and Practice, 20, 19-27.

Cizek, G. J ., & Burg, S. S. (2006). Addressing test anxiety in a high-stakes environment:
Strategies for classrooms and schools. Thousand Oaks, CA: Corwin Press.

161

Ergene, T. (2003). Effective interventions on test anxiety reduction. School Psychology
International, 24, 313-328.

Fielding, C. (2004). Low performance on high-stakes test drives special education
referrals: A Texas survey. The Education Forum, 68, 126-132.

Griffiths, J ., & Peterman, S. (2007). MEAP Reporting. Retrieved February 25, 2008
from http://www.michigan.gov/documents/mde/23_MEAP_Reports_
218829_7.ppt.

Hembree, R. (1988). Correlates, causes, effects, and treatment of test anxiety. Review of
Educational Research, 58, 47-77.

Hill, K. T. & Sarason, S. B. (1966). The relation of test anxiety and defensiveness to test
and school performance over the elementary-school years: A further longitudinal
study. Monographs of the Society for Research in Child Development, 31(2), 1-
76.

Hill, K. T. & Wigfield, A. (1984). Test anxiety: A major educational problem and what
can be done about it. The Elementary School Journal, 85, 105-126.

Hollingshead, A. B. (1975). Four factor index of social status. New Haven (Connecticut):
Department of Sociology, Yale University.

Holroyd, K., Westbrook, T., Wolf, M,, & Badhom, E. (1978). Performance, cognition,
and physiological responding in test anxiety. Journal of A bnormal Psychology,
87, 442-451.

Hopkins, K. D., & Gullickson, A. R. (1992). Response rates in survey research: A meta-
analysis of the effects of monetary gratuities. Journal of Experimental Education,
61, 52-62.

Jones, B. D., & Egley, R. J. (2004, August 9). Voices from the frontlines: Teachers’
perceptions of high-stakes testing. Education Policy Analysis Archives, 12(39).
Retrieved November 12, 2007 from http://epaa.asu.edu/epaa/vl 2n39/.

Jones, B. D., & Egley, R. J. (2006). Looking through different lenses: Teachers’ and
administrators’ views of accountability. Phi Delta Kappan, 87, 767-771.

Jones, M. G., Jones, B. D., Hardin, B., Chapman, L., Yarbrough, T., & Davis, M. (1999).

The impact of high-stakes testing on teachers and students in North Carolina. Phi
Delta Kappan, 81, 199-203.

162

King, N. J ., & Ollendick, T. H. (1989). Children’s anxiety and phobic disorders in school
settings: Classification, assessment, and intervention issues. Review of
Educational Research, 59, 431-470.

Liebert, R. M., & Morris, L. W. (1967). Cognitive and emotional components of test
anxiety: A distinction and some initial data. Psychological Reports, 20, 975-978.

Lowe, P. A., Lee, S. W., Witteborg, K. M., Prichard, K. W., Luhr, M. E., Cullinan, C. M.,
et a1. (2008). The Test Anxiety Inventory for Children and Adolescents (TAICA):
Examination of the psychometric properties of a new multidimensional measure
of test anxiety among elementary and secondary school students. Journal of
Psychoeducational Assessment, 26, 215-230.

Matters, G., & Burnett, P. C. (2003). Psychological predictors of the propensity to omit
short-response items on a high-stakes achievement test. Educational and
Psychological Measurement, 63, 239-256.

McDonald, A. S. (2001). The prevalence and effects of test anxiety in school children.
Educational Psychology, 21 , 89-101.

Meijer, J. (2001). Learning potential and anxious tendency: Test anxiety as a bias factor
in educational testing. Anxiety, Stress, and Coping, 14, 337-362.

Mulvenon, S. W., Connors, J .V., & Lenares, D. (2001, November). Impact of
accountability and school testing on students: Is there evidence of anxiety? Paper
presentation at the 30th Annual Meeting of the Mid-South Educational Research
Association, Little Rock, AR.

Mulvenon, S., Stegman, C., & Ritter, G. (2005). Test Anxiety: A multifaceted study on
the perceptions of teachers, principals, counselors, students, and parents.
International Journal of Testing, 5, 37-61.

National Association of School Psychologists. (2003). Position statement on using large
scale assessment. Retrieved November 18, 2007 from http://www.nasponline.org

/about_nasp/ pospaper_largescale.aspx.

National Commission on Excellence in Education. (1983). A nation at risk: The
imperative for education reform. Washington, DC: US. Department of Education.

Pallant, J. (2007). SPSS Survival Manual: A step by step guide to data analysis using
SPSS for Windows, third edition. Berkshire, England: Open University Press.

Putwain, D. W. (2007). Test anxiety in UK schoolchildren: Prevalence and demographic
patterns. British Journal of Educational Psychology, 77, 579-593.

163

I’: {JP q

Putwain, D. W. (2008). Do examinations stakes moderate the test anxiety-examination
performance relationship? Educational Psychologi, 28, 109-118.

Reynolds, C. R., & Kamphaus, R. W. (2004). BASC-2: Behavior assessment system for
children, second edition manual. Circle Pines, MN: AGS Publishing.

Ryan, K. E., & Ryan, A. M. (2005). Psychological processes underlying stereotype threat
and standardized math test performance. Educational Psychologist, 40, 53-63.

Salmon, G., & Kirby, A. (2008). Schools: Central to providing comprehensive CAMH
services in the future. Child and Adolescent Mental Health, 13, 107-114.

Sarason, S. B. (1959). What research says about test anxiety in elementary school
children. NEA Journal, 48, 26-27.

Sarason, S. B., Davidson, K. S., Lighthall, F .F ., Waite, R., & Ruebush, B. K. (1960).
Anxiety in elementary school children. New York: Wiley.

Sarason S. 3., Hill, K. T., & Zimbardo, P, G. (1964). A longitudinal study of the relation
of test anxiety to performance on intelligence and achievement tests. Monographs
of the Society for Research in Child Development, 29(7), 1-51.

Seipp, B. (1991). Anxiety and academic performance: A meta—analysis of findings.
Anxiety Research, 4, 27-41.

Smith, T. W., Ingram, R. E., & Brehm, S. S. (1983). Social anxiety, anxious
preoccupation and recall of self-relevant information. Journal of Personality and
Social Psychology, 44 , 1276-1283.

Spielberger, C. D. (1966). The effects of anxiety on complex learning and academic
achievement. In. C. D. Spielberger (Ed.). Anxiety and behavior (pp. 261-398).
New York: Academic Press.

Spielberger, C. D., Anton, W. D., & Bedell, J. (1976). The nature and treatment of test
anxiety. In M. Zuckerman & C. D. Spielberger (Eds). Emotions and anxiety: New
Concepts, methods, and applications (pp. 317-344). Hillsdale, NJ: Erlbaum.

State of Michigan. (2004a). Design and validity of the MEAP test. Retrieved February
25, 2008 from http://www.michigan.gov/mde/0,1607,7-140-22709_3l168-94522-
-,00.html.

State of Michigan. (2004b). Requirements for schools not making AYP. Retrieved

February 25, 2008 from ‘http://www.michigan.gov/mde/0,l607,7-140-
22709_22875-85932--,00.html.

164

State of Michigan. (2007). MEAP Fall 2007 Guide to Reports Grades 3-9. Retrieved
April 27, 2008 from http://www.michigan.gov/documents/mde/8_Fall_2007_
Guide_to_Reports_230806__7.pdf.

Steele, C. M. & Aronson, J. (1995). Stereotype threat and the intellectual test

performance of African-Americans. Journal of Personality and Social Psychology
69(5), 797-811.

Stdber, J ., & Pekrun, R. (2004). Advances in test anxiety research. Anxiety, Stress, and
Coping, 17, 205-211.

Triplett, C. F ., & Barksdale, M. A. (2005). Third through sixth graders’ perceptions of
high-stakes testing. Journal of Literacy Research, 3 7, 23 7-260.

Tryon, G. S. (1980). The measurement and treatment of test anxiety. Review of
Educational Research, 50, 343-3 72.

Turner, B. G., Beidel, D. C., Hughes, 8., & Turner, M. W. (1993). Test anxiety in African
American school children. School Psychology Quarterly, 8, 140-152.

US. Congress. (1994a). Public Law 103-227. Goals 2000: Educate America Act.
Washington: Author.

US. Congress. (1994b). Public Law 103-3 82. Improving America ’s Schools Act of 1994.
Washington: Author.

US. Congress. (2002). Public Law 107-110. No Child Left Behind Act of 2001 .
Washington: Author.

Wigfleld, A., & Eccles, J. S. (1989). Test anxiety in elementary and secondary school
students. Educational Psychologist, 24, 159-183.

Wine (1971). Test anxiety and direction of attention. Psychological Bulletin, 76, 92-104.
Wong, K. K., & Nicotera, A. (2007). Successful schools and educational accountability:
Concepts and skills to meet leadership challenges. Boston: Pearson Education,

Inc.

Wren, D. G., & Benson, J. (2004). Measuring test anxiety in children: Scale development
and internal construct validation. Anxiety, Stress, and Coping, 17, 227-240.

Yerkes, R. M., & Dodson, J. D. (1908). The relation of strength stimulus to rapidity of

habit-formation. Journal of Comparative and Neuropsychological Psychology,
18, 459-482.

165

Zeidner, M. (1997). Test anxiety: A transactional perspective. Occasional Paper.
Laboratory for Cross-Cultural Research in Personality and Individual Differences,
University of Haifa, Haifa, Israel.

Zeidner, M. (1998). Test Anxiety: The state of the art. New York: Plenum Press.

166