Beruflich Dokumente
Kultur Dokumente
www.elsevier.com/locate/jad
PII: S0165-0327(17)30368-3
DOI: https://doi.org/10.1016/j.jad.2017.11.048
Reference: JAD9362
To appear in: Journal of Affective Disorders
Received date: 17 February 2017
Revised date: 13 September 2017
Accepted date: 11 November 2017
Cite this article as: Nathaniel von der Embse, Dane Jester, Devlina Roy and
James Post, Test anxiety effects, predictors, and correlates: A 30-year meta-
analytic review, Journal of Affective Disorders,
https://doi.org/10.1016/j.jad.2017.11.048
This is a PDF file of an unedited manuscript that has been accepted for
publication. As a service to our customers we are providing this early version of
the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting galley proof before it is published in its final citable form.
Please note that during the production process errors may be discovered which
could affect the content, and all legal disclaimers that apply to the journal pertain.
RUNNING HEAD: TEST ANXIETY META-ANALYSIS
Temple University
ABSTRACT:
Background and objectives: Since the 1988 publication of Hembree’s seminal meta-analysis on
test anxiety, much has changed in the theoretical conceptualization of test anxiety, as well as the
influences on test performance. The present study examined the influence of test anxiety on a
variety of educational outcomes, in addition to demographic and intrapersonal correlates.
Design and methods: Results of 238 studies from 1988 to the present, were synthesized via a
meta-analytic framework to explicate predictors, correlates, and relationships with the test
anxiety construct. Pooled effect sizes were calculated.
Results: Results indicated that test anxiety was significantly and negatively related to a wide
range of educational performance outcomes, including standardized tests, university entrance
exams, and grade point average. Results were most pronounced at the middle grades level. Self-
esteem was a significant and strong predictor of test anxiety. Perceived difficulty of the test and
the high-stakes nature or consequences of the test was also related to higher test anxiety.
Conclusions: The magnitude of the relationship of test anxiety across a number of variables was
in the small to moderate range. Implications for researchers and educators are discussed
including the use of evidence-based assessment and interventions when warranted.
TEST ANXIETY META-ANALYSIS 3
The role of emotion in performance has long been subject to much interest and research
(see Pekrun, 2006). Yerkes and Dodson first proposed an optimal state of arousal with
performance in 1908, and subsequently changed the study of emotion and its related effects.
Anxiety has been the primary variable of interest in understanding the role of emotion with
negative physiological reactions (APA, 2013). Consistently, anxiety problems are reported to be
one of the most widespread mental health problems in children, with prevalence upwards of 41%
(Cartwright-Hatton, McNicol, & Doubleday, 2006). Heightened levels of anxiety are often
manifested when individuals encounter environments and scenarios that may provoke feelings of
fear or worry, such as a situation where an individual’s abilities are being tested. When faced
with an assessment of ability, there is often an implicit concern of negative consequences based
upon test performance. Test anxiety, exam stress, or test stress are often synonymous with the
Within educational settings, test anxiety is one form of “academic anxiety” (Cassady,
2010) that is characterized by (1) context specific stimuli (e.g., classroom instruction) and (2)
academic subject specific reactions (e.g., math anxiety; Hembree, 1990). Test anxiety is
differentiated from other forms of anxiety through its focus on evaluative (i.e., testing) situations.
Test anxiety has also been referred to as exam anxiety, exam stress or test stress. Recent
estimates have suggested that between 15 and 22% of students exhibit high levels of test anxiety
(Putwain & Daly, 2014; Thomas, Cassady, & Finch, 2017). Much research has demonstrated a
relationship between test anxiety and a host of negative variables including increased risk for
TEST ANXIETY META-ANALYSIS 4
subsequent anxiety and depression (Leadbeater, Thompson, & Grupposo, 2012), poor class
grades (Chapell, et al., 2005; Segool, Carlson, Goforth, von der Embse, & Barterian, 2013),
difficulties engaging in instructional content (Bedell & Marlowe, 1995), and low test
performance (Putwain, 2008; von der Embse & Witmer, 2014). Given these outcomes, research
has produced a variety of evidence-based interventions to mitigate test anxiety (Ergene, 2003;
von der Embse, Barterian, & Segool, 2013). However, to continue advancing evidence-based
treatment of test anxiety, it is necessary to better understand (1) the changing nature of the
construct, (2) the evolution and improvement in test anxiety measurement, and (3) the
Some of the earliest and most influential work on test anxiety was published in the
1950’s (Sarason & Mandler, 1952) and has since influenced research and advanced our
association with a range of negative outcomes, test anxiety research experienced a period of
decline following the 1980s (Zeidner, 1998). Database searches of Psychinfo and ERIC within
the present investigation returned five percent fewer results in the 1990s than the 1980s, with
another slight decline in the early 2000s. However, results from 2010-2017 indicate a 31%
increase in published research when compared to the 1980s. In addition, there has been
significant disagreement within the research literature regarding the relationship of test anxiety
with test performance (see Sommer & Arendasy, 2014; 2015). As the construct approaches 70
years of research, much has changed in the predictors, variables influenced, and theoretical
Theoretical models of test anxiety have evolved over the last five decades. Theorists such
as Alpert and Haber (1960), Liebert and Morris (1967) and Wine (1971) all supported an
interference model of test anxiety, which explains depressed performance by identifying factors
(e.g., emotionality and worry) that disturb the process of information recall and utilization during
testing situations. Shortly after the conceptualization of this model, other theorists proposed a
deficits model, which regards the occurrence and severity of test anxiety as being due to deficits
in knowledge and skills (e.g., study skills, self-efficacy, motivation, testing strategies) needed to
perform well in evaluative situations (Culler & Hollahan, 1980; Tobias, 1985). While both of
these models provide valid information on the complexities of test anxiety, current research
suggests that neither is capable of explaining the dynamic and highly variable nature of test
anxiety. Eysenck and colleagues’s (2007) work with the attentional control theory posited that
anxiety influences performance due to an adverse effect on attentional control, including shifting
and inhibiting. The transactional model of test anxiety (Spielberger & Vagg, 1995) integrated
prior models including both interference and deficit theory, while considering test anxiety as a
cyclical or recursive phenomenon with test performance. That is, test anxiety is associated with
both appraisals of threat and negative or positive evaluations with their related emotions. Zeidner
and Matthews (2005) extends the transactional model by integrating multiple influences of
anxiety on cognitive process with their self-referent executive processing model (S-REF). Within
the S-REF model, executive processing is shaped by declarative and procedural self-knowledge
and that self-referent processing occurs in response to threatening external stimuli (e.g., worry
over academic competence) eventually leading to the manifestation of test anxiety (Zeidner &
Matthews, 2005).
TEST ANXIETY META-ANALYSIS 6
More recently, new models of test anxiety have incorporated environmental influences.
Lowe et al (2008) proposed a biopsychosocial model of test anxiety that indicates a biological
and psychological basis of test anxiety that interacts with the social (i.e., educational) context.
Segool and von der Embse (2014) built upon this model, using advanced statistical modeling to
and prior learning experiences with demographic characteristics, social or educational context,
and environmental contingencies (i.e., educational expectations). It is clear that the construct of
test anxiety is still evolving within the research. Similarly, the measurement of test anxiety has
The Test Anxiety Scale for Children (TASC; Sarason, Davidson, Lighthall, Waite, &
Ruebush, 1960) was long considered the gold standard for test anxiety assessment. Theoretical
advances, namely the deficit and interference models, influenced the development of a two-
factor measurement model assesses autonomic arousal (e.g., emotionality) and cognitive (e.g.,
worry) symptoms that was later incorporated into assessments developed by Liebert and Morris
(1967). Additional instruments included the Test Anxiety Profile (Oetting & Cole, 1980),
Sarason’s Reaction to Test (RTT) measure as well as the revised version from Benson and
colleagues (1992; Revised Test Anxiety scale[RTA]), and the Test Anxiety Inventory (TAI;
Spielberger, 1980). A social dimension was later included in instruments such as the FRIEDBEN
Test Anxiety Scale (FTAS; Friedman & Bendas-Jacob, 1997). Measurement has continued to
evolve as consistent with theory and later, demands in the educational environment. For example,
Cassady developed the Cognitive Test Anxiety Scale (Cassady & Johnson, 2002) to identify
different cognitive aspects of test anxiety, Pekrun and colleagues (2004) created an assessment to
identify “test emotions” and Lowe created two assessments consistent with the biopsychosocial
TEST ANXIETY META-ANALYSIS 7
model, the Test Anxiety Inventory for Children and Adolescents (TAICA; Lowe et al., 2008) and
the Test Anxiety Scale for Elementary Students (TAS-E; Lowe, Grumbein, & Raad, 2011). As
schools have increasingly integrated tiered service delivery that utilizes brief screening measures
to facilitate early intervention and prevention efforts, test anxiety assessments progressed to
reflect these needs including screening (B-FTAS; von der Embse et al., 2013) and single-item
measures (DBR-A; von der Embse et al., 2015). Clearly, the measurement and understanding of
test anxiety has changed significantly, yet the last meta-analysis examining the effects of the
In 1988, Ray Hembree published a seminal meta-analysis of 562 studies on the correlates,
causes, effects, and treatment of test anxiety (Hembree, 1988). This work has had great influence
on the subsequent understanding of test anxiety differences (e.g., gender, ethnicity, age) as well
as its relations (e.g., consistently linked with lower academic performance). Many of Hembree’s
findings were later supported through Seipp’s (1991) meta-analysis. Hembree’s meta-analysis
had informed intervention development (e.g., Ergene, 2003; von der Embse et al., 2013) that was
targeted to groups at most risk for high levels of test anxiety. However, theoretical and
measurement advances have changed our understanding of the test anxiety construct (Sommer &
Arendasy, 2014). Research has now more clearly elucidated the processes inherent to the
manifestation of test anxiety (e.g., Zeidner & Matthews, 2005) while also calling into question
the causal link to performance (see Sommer & Arendasy, 2013). Moreover, the nature and use of
testing for academic progress and achievement in schools have changed dramatically, thus
raising again the importance of emotion in performance. For example, many countries have used
high-stakes exams to determine student academic gains, teacher efficacy, and school
effectiveness. Student test anxiety is higher on high-stakes exams when compared with typical
TEST ANXIETY META-ANALYSIS 8
classroom tests (Segool, Carlson, Goforth, von der Embse, & Barterian, 2013), underscoring the
potential influence this change has had on students. In addition, individual variability in the
self-regulation; Dull, Schliefer, & McMillan, 2015; Khalaila, 2015; Schnell, Ringeisen,
Raufelder, & Rohrmann, 2015), social influences (performance expectations and values,
achievement standards, and social support; von der Embse, Schultz, & Draughn, 2015; Ringeisen
& Raufelder, 2015; Sarason, 1981), and demographic variables (level of education, economic
status, and cultural background; Hembree, 1988; Putwain, 2008; Ziedner, 1990).
Given the change in the understanding and measurement of test anxiety over the last 50
years, there is a need to examine the potential related changes in the relationship of test anxiety
with a variety of domains. The purpose of the present meta-analysis is threefold. First, we
examine the relationship of test anxiety with a variety of performance outcomes, including
achievement tests, grade point average, and standardized exams. Second, intrapersonal variables
such as motivation, and their related association to test anxiety are assessed. Lastly, we examine
differences in the predictors of test anxiety with respect to various demographic variables.
Method
articles and doctoral dissertations) that address test anxiety. Methodological approaches adhered
to best practices identified by Hedges and Olkin (2014). Data collection was conducted through a
TEST ANXIETY META-ANALYSIS 9
multiple yield, narrowing criterion design. First, a coding manual was developed (Hunter &
Schmidt, 1990) that included relevant search terms and inclusion/exclusion criteria (see below).
Second, raters practiced searching with the relevant terms to compare initial yields and search
procedures, and discussion was used to confirm common methodology. Third, approximately 20%
of articles across all yields were examined by a second rater to determine inclusion/exclusion
criteria and calculate rater agreement. If disagreement existed, a third rater would examine the
Behavioral Sciences Collection, PsycINFO, Academic Search Premier, Medline and the
Educational Resource Information Center (ERIC), with articles published in English after 1988
(i.e., after the publication of Hembree’s meta-analysis). The initial search terms included “test
anxiety”, “test stress”, “exam anxiety”, and “exam stress.” The search yielded a total 1,511
Several inclusion and exclusion criteria were applied for a second yield of articles. These
include: (1) articles needed to identify test anxiety as the independent variable (or correlate), (2)
an indicator of test performance was included as a dependent variable, and (3) test anxiety was
examined within an educational setting (i.e., schools and universities are an educational setting,
hospitals or clinics were not considered an educational setting). Test anxiety is typically
measured and intervened within education settings and therefore the implications are most
directly relevant for educators, thus the inclusion of educational settings and the exclusion of
others (e.g., hospitals). Group studies, single-case designs, and both quasi-experimental (as well
as correlational) and experimental designs were included in the second yield if effect size was
provided or could be calculated. Article reviews, test development/validation studies, and other
TEST ANXIETY META-ANALYSIS 10
non-empirical articles were excluded based upon these criteria. The second yield resulted in 391
In order to be considered acceptable for use in the meta-analysis, all identified articles
were then individually screened and categorized by the authors for the following criteria: (1) year
status, and (4) academic performance (e.g., course grades, GPA). A final yield of 238 studies met
A correlational meta-analysis was employed with 238 studies that met inclusion criteria.
Each of these studies provided either an effect size or the requisite data to calculate an effect size.
Due to the differing methodologies of the identified studies, multiple types of effect sizes were
listed. While both continuous (Pearson’s r) and between-group (Cohen’s d) effect sizes were
identified, r was the primary measure of effect size. Multiple studies did not report effect sizes,
but provided sufficient information to calculate an effect size. In these cases, the following
[( Σ x2 )*( Σ y2 )]). Due to the interchangeable nature of effect sizes, d-values were converted to
coefficients through the Hedges-Olkin fixed-effects model (1985). A fixed-effects model was
preferred over a random-effects model due to the large amount of studies and the general
homogeneity of the studies within each analysis conducted (Hedges & Olkin, 2014). The
Hedges-Olkin method converts the correlation coefficients of each study to a standard metric,
TEST ANXIETY META-ANALYSIS 11
which in this case uses Fisher’s r-to-Z transformation. This transformation was made for each
effect size within each unit of analysis (i.e., study aims) in the meta-analysis. After these
transformations were completed for each group of studies, a weighted average of these scores
was calculated and applied to each study within this group to identify the mean effect size. The
z-score of the mean effect size was then calculated by dividing the mean effect size by its
standard error and converted back into r, or the pooled correlation. A chi-square test was then
conducted to determine whether the effect sizes were homogeneous (Johnson, 1993). The
significance level for the test for the homogeneity of the k values of r (df= k-1) was set at .01 due
to standard agreement within meta-analysis literature and the inherent variability of effect sizes
due to data obtained through self-report measures (Schmidt & Hunter, 2014). When homogeneity
of the sample was evaluated and considered acceptable, groups were considered complete and
confidence intervals were calculated to test the null hypothesis that the effect size is zero. If a
group of studies was found to be heterogeneous (Q > k-1), outliers were identified and examined
to determine the cause of their variability and whether they should be excluded from the analysis
(Hedges & Olkin, 2014). Outliers were identified through a close inspection of data obtained
upon calculation of the overall correlation coefficient for the area in question. For data that was
heterogeneous, the random/fixed 95% confidence interval was used as an indicator of potential
outliers. Each data point falling outside of this range was examined. Outliers were eliminated
when the methodology and/or analyses were considered to be insufficient or irrelevant to the
aims identified within the current study. Correlation coefficients were then recalculated after
identification and removal of identified outliers. Data points falling outside of the 95%
confidence interval were included when there was no clear cause of variability.
TEST ANXIETY META-ANALYSIS 12
The potential influence of publication bias, or the “File Drawer Effect”, which suggests
that studies with significant results are more likely to be published than those that report results
that are not statistically significant (Rosenthal, 1979) was examined. Orwin’s Fail-safe N (Orwin,
1983) to identify the number of potentially unpublished studies with a null effect size needed to
reduce the pooled correlated to a criterion effect size that would no longer be meaningful. The
criterion effect size was set at r = .10 or r = -.10, depending upon the directionality of the data, as
this represents the lowest threshold for a small effect sizes identified by Cohen (Cohen, 1988).
Results
Correlational Findings
The primary task of the current study was to identify significant correlates of test anxiety
with primary variables of interest (e.g., achievement). Most of the significant relationships fell
within the small to medium range. Results are compared and then discussed with respect to the
findings reported in Hembree’s (1988) seminal examination of test anxiety. Individual and
separate correlates were identified within each level of analysis. The aggregate effect of these
correlates is denoted with their respective tables. Additional information within the tables
detailing the relation with test anxiety include (1) the number of studies within the group of
analysis, (2) the total number of participants from the studies, (3) outlier values when included,
(4) the end values of r within the group of studies, (5) the pooled effects size or mean, and (6) an
indicator of significance when the mean was significantly different than zero. In several
instances, a single study included information that was utilized in multiple levels of analysis (e.g.,
The first aim of the current study (see summary within Table 1) was to examine the
relation between test anxiety and performance measures commonly identified in educational
settings. The coding procedure applied to these studies identified several performance categories,
including (1) achievement or typical classroom testing (e.g., quizzes, tests), (2) Grade Point
Average (GPA), (3) intelligence quotient (IQ), and (4) standardized exams (i.e., State Exams,
SAT, ACT). Subsequent analyses of several of these measures were conducted due to the depth
and range of research on test anxiety’s relation with differing measures of performance.
Additional analyses of these performance categories were conducted to align with the
differing components of test anxiety that are widely identified within the literature, including (1)
(4) social components. Not all studies differentiated between these components. Due to the
variability of scales administered within the studies examined, multiple scales were clustered
within differing test anxiety components dependent upon the individual scale’s operational
definition. The components and corresponding include: Cognitive (e.g, worry, test-irrelevant
tension, bodily symptoms, autonomic reactions), Behavioral (e.g., off-task behaviors), and Social
included measures of typical classroom or achievement tests (e.g., class test or quiz).
Performance measures included in this group were exam scores, midterm scores, and final exam
scores. Due to prior research and theories on the developmental trajectory of test anxiety (e.g.,
Dan, Bar Ilan, & Kurman, 2014; Hembree, 1998), this category of performance measures was
further delineated into school-grade clusters including elementary school (first to fifth grades),
TEST ANXIETY META-ANALYSIS 14
middle school (sixth to eighth grades), high school (nine to twelfth grades), and post-secondary
multiple clusters (e.g., sixth grade as elementary/primary in the United Kingdom in contrast to
middle school in the United States). The pooled correlates of these grade-clusters suggest that the
relationship of test anxiety with test performance increases from elementary (r = -.22) to middle
school (r = -.25), decreases in high school (r = -.16), and increases again in post-secondary
educational settings (r = -.24). Each cluster shared a significant inverse relationship with test
anxiety.
Grade Point Average (GPA). Thirty-six studies examined the relationship between test
anxiety and cumulative GPA. This category was distinguished from the achievement tests
category, as it is indicator of broad achievement. The pooled correlate was significant at r = -.17
indicating a negative relation. It should be noted that 20 out of the 36 studies identifying GPA as
contribute to the correlation between post-secondary achievement test performance and test
anxiety.
Intelligence (IQ). In total, 19 studies examined the correlation between test anxiety and
performance on IQ tests. Due to the structure of many of these studies, IQ was delineated into
three major clusters that align with current research on the construct of intelligence, including
nonverbal processing abilities (e.g., fluid reasoning, problem solving, visual processing),
cognitive proficiency abilities (e.g., memory, processing speed), and verbal abilities (e.g.,
verbal abilities (r = -.24), and nonverbal reasoning abilities (r = .10) all fell within the small
range.
TEST ANXIETY META-ANALYSIS 15
Standardized Exams. Within this study, standardized exams were defined as any test or
exam that was normed amongst a large population, so that individual scores could be compared
the relationship between test anxiety and performance on standardized exams. The overall
pooled correlate for this group of studies (r = -.26) was the highest of the measures of
during the coding of these studies. These sub-measures include Entrance Examinations
(SAT/ACT), State Exams, and Professional Exams. These were amongst the largest correlations
between test anxiety and performance measures. Performance on entrance exams shared the
second highest correlation with test anxiety (r = -.31). Due to significant variability within State
Exam correlation data, these studies were further delineated and examined according to
previously identified grade clusters. Pooled correlates for primary, intermediate, and secondary
grades indicated a similar relation between achievement and test anxiety as identified previously,
increasing from primary grades (r = -.16) to intermediate grades (r = -.40) before decreasing in
secondary grades (r = -.23). The highest correlation between performance and test anxiety was
identified among secondary students who had taken standardized state exams.
examined in relation to differing components of test anxiety, including the cognitive component,
the affective/physiological component, the behavioral component, and the social component.
Due to insufficient data, correlations between the behavioral and social components of test
anxiety and differing areas of performance (i.e., achievement testing, GPA, IQ, Standardized
Exams) could not be delineated and calculated. There was sufficient data to calculate correlations
for the cognitive and affective/physiological components of test anxiety. Results indicated all
TEST ANXIETY META-ANALYSIS 16
correlations fell within the small range and there were relatively higher correlations between
areas of performance and the worry component of test anxiety than the affective/physiological
= -.26; Affective/Physiological: r = -.13). The social component of test anxiety was significantly
related to performance (r = -.12), while the behavioral component was not (r = .04).
The second aim (see summary within Table 2) was to examine the relationship between
test anxiety and personal variables. Five categories were identified during the coding process to
Traits. These categories were established based upon the study goals and availability of data.
including Internal locus of/Personal control (i.e., i.e., one’s belief that have personal control of
academic tasks; r = -.32), Self-regulation (i.e., use of strategies, monitoring of learning outcomes;
r = -.21), Self-efficacy (i.e., personal belief in ability to engage in behaviors that facilitate
academic success; r = -.33), Self-esteem (i.e., judgment of one’s past successes or failures
relative to desired outcomes; r = -.43), and Academic Confidence (i.e., personal belief in
adequate performance of various required academic specific task demands; r = -.29). All sub-
categories shared a significantly inverse relationship with test anxiety, and were homogeneous
Motivation. Motivation was examined given its long association within the research
literature to academic achievement (Cerasoli, Nicklin, & Ford, 2014; Deci, 1972) Seventeen
studies examined motivation and were further divided into two categories, Intrinsic Motivation
and Extrinsic Motivation. A significant correlation was found for both correlates, but the
relationship differed. Test anxiety was negatively related to Intrinsic Motivation (r = -.08) but
Coping Skills. This category was divided into two sub-categories: Problem-
Focused/Active Coping and Avoidance Coping. Ender and Parker (1991) defined problem-
circumventing the stressor and avoidance coping as a strategy in which an individual engages in
non-relevant cognitions or behaviors in order to avoid the stressful situation. The pooled
correlate for approach coping was small and significantly negative (r = -.15), whereas the pooled
correlate for Avoidance Coping was moderate and significantly positive (r = .38), suggesting
that test anxious students are much more likely to ignore or escape stressors and stresses.
influence academic performance and educational attainment (Pintrich, 2000). Achievement goals
are cognitive representations of competency-related outcomes that are future oriented (Hulleman,
Shrager, Bodman and Harackiewicz, 2010). Elliot and McGregor (2001) created a 2 x 2
framework to distinguish mastery goals (i.e., desire to development of knowledge and mastering
new skills) and performance goals (i.e., desire to demonstrate knowledge relative to others) along
approach and avoidance dimensions. The goals were defined as the following: mastery-approach
performance-avoidance is the desire to not performance worse than the reference group (or
individual). The correlation between mastery approach goals and test anxiety was not significant.
Using Cohen’s criteria (.1 = small, .3 = medium, .5 = large), the relationship between test anxiety
and mastery-avoidance goals was in the medium range (r = .30). Similarly, a small yet
significant positive correlation was found for performance-approach goals (r = .09), and a
medium significant positive correlation was found for performance-avoid goals (r = .37).
Personality Traits. The “Big Five” factors of personality (Costa & McCrae, 1992) were
identified within many studies and have been shown to impact levels of test anxiety (e.g.,
Chamorro-Premuzic, Ahmetoglu, & Furnham, 2008). Test anxiety shared a significant positive
correlation with Neuroticism (i.e., attributes: anxious, depressed, inability to delay gratification,
and increased vulnerability to stressors in the environment; r = .46) and a significant negative
= -.18). There was a small significant negative correlation with Openness (i.e., Attributes:
and a non-significant negative correlation with both Extraversion (i.e., attributes: assertive,
positive, and sociable; r = -.05) and Agreeableness (i.e., attributes: trusting, empathetic, and
compliant in social situations; r = -.06). These correlations suggest that certain personality traits
The third aim (see summary within Table 3) was to examine demographic correlates of
test anxiety. As with the other study aims, these variables were coded and examined when there
were numerous studies that included a similar and distinct variable (e.g., gender, ethnicity, test
TEST ANXIETY META-ANALYSIS 19
conditions). There were variables of interest of which there was insufficient data (e.g., SES) and
Gender. In total, 49 studies included data comparing test anxiety levels between males
and females in education settings, subdivided into primary, intermediate, secondary, and post-
secondary educational settings. There was a significant relationship between gender and test
anxiety, with females reporting a significantly higher rate than males at all grade levels (r = .21).
This discrepancy between genders appears to increase from primary grades to secondary grades,
but then decreases slightly when students are enrolled in postsecondary educational settings.
Ethnicity. Many studies examined the role of ethnicity and its relationship with test
anxiety, but there was only enough data to examine test anxiety differences between
White/Caucasian and Black/African American students, and these differences have traditionally
been dichotomized as minority versus majority students. Seven studies examined the relationship
between test anxiety and ethnicity. Pooling data from across all grades, (insufficient data to
analyze individual grade levels), minority students reported significantly higher levels of test
Ability. Seven studies examined the relationship between student ability level (average
ability level v. high ability level), determined by teacher perception of performance in the
classroom, and test anxiety. For teacher ratings, perceived ability is a reflection of a student’s
discrepancies between intelligence and engagement (Christenson, Reschly, & Wylie, 2012). The
pooled correlate for this category shows a negative correlation between the two ability levels (r =
-.15), with high ability level students reporting lower levels of test anxiety than students
identified as having an average ability level. There were insufficient data on the effects between
TEST ANXIETY META-ANALYSIS 20
low and average ability level students, but prior research (e.g., Hembree, 1988) suggests a
negative relationship between ability level and test anxiety, with test anxiety decreasing as ability
level increases.
coding process. The first category examined students with and without Attention Deficit
Hyperactivity Disorder (ADHD) diagnoses. Students with diagnoses of ADHD showed higher
levels of test anxiety than students without (r = .12). This finding was similar for students who
were identified with a specific learning disability, who also displayed significantly higher levels
Test Conditions. Many schools and teachers differ in their testing methodology and form.
Three testing conditions were identified during the coding process. The first, perceived difficulty,
pooled effects from studies comparing test anxiety ratings of students in testing scenarios that
appeared simple or difficult. A large effect was identified for this condition (r = .31), with
students showing significantly increased levels of test anxiety in testing scenarios that they
perceived as difficult or challenging. The second test condition focused on test consequences.
Five studies presented data on the relationship between the evaluative nature of tests and test
anxiety. The pooled effect from these studies indicates that student report significantly higher
levels of test anxiety when they are presented with a test or evaluation that is clearly depicted as
being evaluative and having consequences (r = .21). Test anxiety levels were lower when tests
were presented as “activities” or “learning scenarios”. Third, experience with a test was
considered. That is, students who had previously taken a similar exam with regards to form (i.e.,
multiple choice questions) or content. Results indicated that students who had no prior
experience with a test reported higher levels of test anxiety (r = .16) than those with experience.
TEST ANXIETY META-ANALYSIS 21
Discussion
Since the publication of Hembree’s 1988 meta-analysis on the correlates and effects of
test anxiety, there has been significant changes in the theoretical understanding and measurement
of test anxiety. A primary goal of the present study was to employ a meta-analytic review to
empirical research concerning test anxiety over the last 30 years. The relationship of test anxiety
with performance, intrapersonal variables, and various demographic predictors was examined.
Relatedly, results are contrasted and discussed relative to Hembree’s original publication.
Importantly, this meta-analysis synthesizes correlations identified within the included studies.
Results do not imply causation and readers should exercise caution in interpretation and
generalization. Moreover, most of the identified correlations were within the small to moderate
range.
Results from the present study indicated a consistent pattern of relationships with higher
levels of test anxiety and lower levels of performance, across various testing formats. These
results were largely consistent with Hembree as effect sizes ranged from small (r =-.13) to
moderate (r= -.40), and may support the continued negative association of test anxiety with test
performance. However, our review allowed for a further specification across grade levels, and
results suggested that middle grades (i.e., 6th to 8th or ages 11 to 14) exhibited the largest
negative relationship whereas high school (i.e., 9th to 12th grades, or ages 15 to 18) the lowest.
Additionally, recent research has allowed for delineation of various types of standardized exams,
and results indicate the largest negative relationships of test anxiety and performance with
University entrance exams and Standardized State Exams within secondary grades. The use of
standardized, high-stakes testing in education has been gaining attention in recent years, most
TEST ANXIETY META-ANALYSIS 22
likely due to increases in test-based accountability practices associated with modern educational
legislation (Koretz & Hamilton, 2006). With this increase in attention has come an influx of
research on the relationship between test anxiety and student performance on these tests (von der
Embse & Witmer, 2014). These results are consistent with research demonstrating higher levels
of test anxiety on “high-stakes” exams than typical classroom tests (Segool et al., 2013). Thus,
educators may consider the exam format when targeting antecedent or symptom reduction
interventions (e.g., von der Embse et al., 2013). Test anxiety was also demonstrated to have
varying levels of association with IQ testing; for example, pooled effect sizes indicated a larger,
negative relationship to verbal and cognitive proficiency tasks than non-verbal tasks. Prior
research has also suggested that the cognitive component (i.e., worry) of test anxiety may
interfere with verbal demands (Markham & Darke, 1991; Lee, 1999). However, the debilitating
effects of test anxiety during IQ testing may complicate this relationship. The role of differing
test anxiety components was also examined. Results suggested that the cognitive component of
test anxiety consistently shares a higher negative relation with areas of achievement compared to
the other components of text anxiety (i.e., Affective/Physiological, Behavioral, Social). This
finding is consistent with extant research (e.g., Hembree, 1988; Chapell et al., 2005).
With regards to intrapersonal variables, results from the present manuscript were largely
similar to Hembree. That is, self-esteem and self-efficacy were consistently found to have the
strongest relationships with test anxiety. Recent research has demonstrated a clear relationship
between self-concept and test anxiety (Raufelder & Ringeiesen, 2016). Locus of control had a
weaker relationship with test anxiety (.04) than had previously been reported (.22), with strong
heterogeneity between studies, suggesting that students experiencing test anxiety may differ in
TEST ANXIETY META-ANALYSIS 23
their beliefs that their environment is related to their own individual behaviors. The relationship
between test anxiety and self-concept appears to be highly negative and relatively consistent,
depicting the profile of a student who may have difficulties due to negative beliefs of themselves
Results indicated a small, significant positive relationship of test anxiety with extrinsic
motivation. Thus, students who are motivated by external demands rather than internal interests
are more likely to exhibit higher test anxiety. Relatedly, students with high test anxiety are more
likely to endorse performance- and mastery-avoidance goals, and engage in avoidance coping
strategies. Taken together, external factors (e.g., high-stakes testing) may be an impetus for a
student with high test anxiety to avoid testing situations or use avoidance coping strategies.
Research has supported these relationships, as teachers that utilize negative motivational
strategies (i.e., fear appeal) that appeal to an external outcome (i.e., entrance to University) result
in higher student test anxiety (Putwain & Symes, 2011; von der Embse, Schultz, & Draughn,
2015).
Due to the availability of data, the relation between test anxiety and personality traits was
examined. Test anxiety shared its largest positive relationship with neuroticism and a small
negative relationship with conscientiousness. These findings align with prior research examining
the attributes of these traits and their relation to anxiety, as Neuroticism is considered to be
highly related to high levels of anxiety (Watson & Clark, 1984) and Conscientiousness as being
related to certain personal variables (e.g., intrinsic motivation, self-efficacy) that share a negative
relationship with test anxiety (e.g., Mount & Barrick, 1995). These findings suggest that certain
personality traits and their related attributes may be more predictive of higher or lower levels of
test anxiety.
TEST ANXIETY META-ANALYSIS 24
It is often stated that certain demographic groups exhibit higher rates of test anxiety that
may contribute to test achievement gaps (Cassady & Johnson, 2002; Steele & Aronson, 1995;
Wolf & Smith, 1995). Differences in the manifestation of test anxiety are then commonly used to
determine students most “at risk” and in need of intervention supports (Cizek & Burg, 2006).
Hembree’s work has frequently been cited as evidence of these differences, however much has
changed in the nature and consequences of testing. For example, results from the present
investigation indicated females are more likely to exhibit higher test anxiety—while similar in
direction (i.e., females higher), the strength of these relations are lower (.19 versus .43 for similar
grades) than past research. Similarly, our meta-analysis indicated small and significant
relationships (r = .11) with minority status that was again lower than the oft cited Hembree
finding of Black elementary students with much greater test anxiety (r = .52). Perhaps more
importantly, results indicated significant (and consistent with Hembree) differences in the
perceived difficulty (i.e., higher perceived difficulty = higher test anxiety) of the test as well as
test consequences (i.e., evaluative tests relate to higher test anxiety). Although significant
demographic differences remain, it seems prudent for educators to use psychometrically sound
tools to determine need for test anxiety intervention with certain testing conditions rather than
Limitations
Although this meta-analytic review presents several notable strengths, there are also
several limitations. First, though many significant correlations between test anxiety and other
factors have been identified, it should be noted that correlation does not equal causation, and
correlates identified within the studies included in this meta-analysis are not experimental in
TEST ANXIETY META-ANALYSIS 25
nature, and most correlations were within the small to moderate range. Thus, caution is
recommended when extrapolating implications from the pooled correlates reported within this
study. Some researchers have called into question the direction relationship of test anxiety with
test performance, instead suggesting that less competent test takers experience high test anxiety
accounting for the relationship to performance (Sommer & Arendasy, 2014). Additional
experimental research will be necessary to further delineate the relationship of test anxiety
throughout the assessment cycle (i.e., pre, during, post test; von der Embse et al., 2015).
Additionally, test anxiety theory and related measurement has changed significantly across the
last forty years. Therefore, the test anxiety may be variable across multiple studies which limits
potential comparisons.
Second, there were methodological limitations with regards to the search criteria; for
example, studies that were not written in the English language were excluded, despite much test
anxiety research taking place outside of the United States and English speaking countries
(Zeidner, 1998). Similarly, inclusion and exclusion criteria omitted studies which were article
reviews, non-empirical, and from non-educational settings. Given these exclusions, the results
from the present meta-analysis may not fully represent the test anxiety construct across settings.
Additionally, the quality of studies that were not published could not be controlled—thus
potentially leading to publication bias (i.e., file drawer effect) and there exists the possibility of
non-published test anxiety studies with null or contradictory results (Rosenthal, 1979). For
example, the Dissertations Abstracts International database was not searched as part of this meta-
analysis in a necessary attempt to restrict a vast literature base over nearly thirty years. Finally,
meta-analyses may not capture every published study on test anxiety and performance, as search
TEST ANXIETY META-ANALYSIS 26
terms and selected databases may vary, thus caution is urged when comparing meta-analyses
Third, there were limitations of the hypothesized correlates. For example, 20 out of 36
education. With such a large amount of the studies examining only post-secondary education,
results may skew relative to early education settings. Similarly, only seven studies examined
ethnicity and ability, which may not provide a representative sample of the aforementioned
demographic characteristics with the manifestation of test anxiety. Within ethnicity, only
Caucasian and African American students were examined, excluding all other ethnicities due to
insufficient data. In addition, differences amongst correlations were discussed relatively and
statistical differences were not assumed. Future research should critically evaluate the
hypothesized correlates, as well as conduct significance testing across groups to better inform
Conclusions
Test anxiety has continually been demonstrated to have a negative relationship with
important educational outcomes across two meta-analyses spanning nearly 70 years of research.
The relationship between higher test anxiety and lower test performance has been demonstrated
across many hundreds of studies and many thousands of participants. Importantly, testing has
taken a much more prominent role within a multitude of important educational decisions ranging
from grade promotion to University entrance to teacher evaluation. In this era of high-stakes
testing, it becomes even more crucial to understand the relationship of emotions to test
Educators have an important role to play in creating a facilitating rather that debilitating
testing environment for students (von der Embse & Putwain, 2015). This includes using positive
motivational strategies, and explaining the use of assessments to students. Researchers and
educators should consider evidence-based test anxiety interventions that are targeted to the most
variable with female students reported significantly higher levels of test anxiety than males.
However, other demographic groups (e.g., minority status) did not appear to have as strong as
relationships to test anxiety and others (e.g., SES) were not able to be analyzed in this review.
Research will be necessary to further delineate which students may be most at risk and thus the
need for continued development of responsive test anxiety assessment tools that can be used in
screening, decision-making, and progress monitoring situations (Cassady & Johnson, 2002;
Pekrun et al., 2004; Lowe et al., 2008; Regehr, Glancy, & Pitts, 2013). Although research has
offered guidelines for diagnostic criteria for depression and anxiety more generally (Gurley,
Cohen, Pine, & Brook, 1996), test anxiety cut score research is only now beginning to emerge
(Thomas, Cassady, & Finch, 2017). Educators are recommended to take a tiered intervention
approach that includes a universal or school-wide curriculum (e.g., Weems et al., 2015) followed
by more targeted and individualized supports (e.g., Ergene, 2003, Reiss et al., 2017) for students
exhibiting higher levels of test anxiety. Much has changed in the measurement, treatment, and
study of test anxiety, leading to a better understanding of the role of emotion in test performance
Contributors: Dr. Nate von der Embse designed the study and coordinating the writing of the
manuscript. Dane Jester managed the literature searches and analyses. James Post and Devlina
Roy undertook the literature search and inclusion of final sample, and Dr. Nate von der Embse
wrote the first draft of the manuscript. All authors contributed to and have approved the final
manuscript.
Role of the Funding Source: The authors declare no funding received for this manuscript or
study.
Acknowledgements: None.
Conflict of interest
References
American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders
Alpert, R., & Haber, R. (1960). Anxiety in academic achievement situations. Journal of
Bedell, J.R., & Marlowe, H.A. (1995). An evaluation of test anxiety scales: Convergent,
divergent, and predictive validity. In C.D. Spielberger & P.R. Vagg (Eds.), Test anxiety
theory, assessment, and treatment (pp. 35–45). Washington, DC: Taylor & Francis.
Benson, J., Moulin-Julian, M., Schwarzer, C., Seipp, B., & El-Zahhar, N. (1992). Cross
validation of a revised test anxiety scale using multi-national samples. In K. A. Hagvet &
T. B. Johnsen (Eds.), Advances in Test Anxiety Research Vol. 7 (pp. 62-83). Amsterdam:
Cartwright-Hatton, S., McNicol, K., & Doubleday, E. (2006). Anxiety in a neglected population:
Cassady, J. C. (2010). Anxiety in schools: The causes, consequences, and solutions for academic
Cassady, J. C., & Johnson, R. E. (2002). Cognitive test anxiety and academic performance.
Cerasoli, C. P., Nicklin, J. M., & Ford, M. T. (2014). Intrinsic motivation and extrinsic
Chamorro-Prezumic, T., Ahmetoglu, G., & Furnham, A. (2008). Little more than personality:
Dispositional determinates of test anxiety (big five, core self-evaluations, and self-
Chapell, M. S., Blanding, Z. B., Silverstein, M. E., Takahashi, M., Newman, B., Gubi, A., &
10.1037/0022-0663.97.2.268
Christenson, S. L., Reschly, A. L., & Wylie, C. (Eds.). (2012). Handbook of research on student
Cizek, G., & Burg, S. (2006). Addressing Test Anxiety in a High Stakes Environment. Thousand
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ:
Erlbaum
Costa, P. T., Jr. & McCrae, R. R. (1992). Revised NEO personality inventory (NEO-PI-R) and
Assessment Resources.
Culler, R. E., & Holahan, C. J. (1980). Test anxiety and academic performance: The effects of
10.1037/0022-0663.72.1.16
Dan, O., Bar Ilan, O., & Kurman, J. (2014). Attachment, self-esteem and test anxiety in
10.1080/01443410.2013.814191
TEST ANXIETY META-ANALYSIS 32
Deci, E. L. (1972). The effects of contingent and noncontingent rewards and controls on intrinsic
doi:10.1016/0030-5073(72)90047-5
Dull, R. B., Schleifer, L. L., & McMillan, J. J. (2015). Achievement goal theory: the relationship
doi: 10.1080/09639284.2015.1036892
Elliot, A. J., & McGregor, H. A. (1999). Test anxiety and the hierarchical model of approach and
10.1037/0022-3514.58.5.844
Eysenck, M. W., & Calvo, M. G. (1992). Anxiety and performance: The processing efficiency
Eysenck, M. W., Derakshan, N., Santos, R., & Calvo, M. G. (2007). Anxiety and cognitive
3542.7.2.336
Friedman, I. A., & Bendas-Jacob, O. (1997). Measuring perceived test anxiety in adolescents: A
doi: 10.1177/0013164497057006012.
TEST ANXIETY META-ANALYSIS 33
Gregor, A. (2005). Examination anxiety: Live with it, control it or make it work for you? School
Hancock, D. R. (2001). Effects of test anxiety and evaluative threat on students' achievement and
10.1080/00220670109598764
Hedges, L., & Olkin, I. (1985). Statistical methods for meta-analysis. Orlando, FL: Academic
Press.
Hedges, L. V., & Olkin, I. (2014). Statistical methods for meta-analysis. Academic press.
Hembree, R. (1988). Correlates, causes, effects, and treatment of test anxiety. Review of
Hembree, R. (1990). The nature, effects, and relief of mathematics anxiety. Journal for research
Hunter, J. E., & Schmidt, F. L. (1990). Methods of meta-analysis: Correcting error and bias in
Khalaila, R. (2015). The relationship between academic self-concept, intrinsic motivation, test
anxiety, and academic achievement among nursing students: Mediating and moderating
Koretz, D., & Hamilton, L. S. (2006). Testing for accountability in K-12. In R.L. Brennan (Ed.),
Educational Measurement (4th ed., pp. 531-578). Westport, CT: American Council on
Education/Praeger.
Lee, J. H. (1999). Test anxiety and working memory. The Journal of Experimental
Leadbeater, B., Thompson, K., & Gruppuso, V. (2012). Co-occurring trajectories of symptoms
adulthood. Journal of Clinical Child & Adolescent Psychology, 41(6), 719-730. doi:
10.1080/15374416.2012.694608
Liebert, R., & Morris, L. (1967). Cognitive and emotional components of test anxiety: A
distinction and some initial data. Psychological Reports, 20, 975-978. doi:
10.2466/pr0.1967.20.3.975
Lowe, P. A., Grumbein, M. J., & Raad, J. M. (2011). Examination of the psychometric properties
of the Test Anxiety Scale for Elementary Students (TAS-E) scores. Journal of
Lowe, P. A., Lee, S. W., Witteborg, K. M., Prichard, K. W., Luhr, M. E., Cullinan, C. M., Janik,
M. (2008). The Test Anxiety Inventory for Children and Adolescents (TAICA):
Markham, R., & Darke, S. (1991). The effects of anxiety on verbal and spatial task
Mount, M. K., & Barrick, M. R. (1995). Manual for the Personal Characteristics Inventory.
Oetting, E. R., & Cole, C. W. (1980). Test anxiety profile. Fort Collins, CO: Tri-Ethnic Center
Statistics, 8, 157-159.
TEST ANXIETY META-ANALYSIS 35
and implications for educational research and practice. Educational Psychology Review,
Pekrun, R., Goetz, T., Perry, R. P., Kramer, K., Hochstadt, M., & Molfenter, S. (2004). Beyond
test anxiety: Development and validation of the Test Emotions Questionnaire (TEQ).
Pintrich, P. R. (2000). Multiple goals, multiple pathways: The role of goal orientation in learning
0663.92.3.544
10.1348/000709906X161704
Putwain, D., & Daly, A. L. (2014). Test anxiety prevalence and gender differences in a sample of
doi: 10.1080/03055698.2014.953914
Raufelder, D., & Ringeisen, T. (2016). Self-perceived competence and test anxiety. Journal of
Regehr, C., Glancy, D., & Pitts, A. (2013). Interventions to reduce stress in university students:
10.1016/j.jad.2012.11.026
TEST ANXIETY META-ANALYSIS 36
Reiss, N., Warnecke, I., Tolgou, T., Krampen, D., Luka-Krausgrill, U., & Rohrmann, S. (2017).
Effects of cognitive behavioral therapy with relaxation vs. imagery rescripting on test
anxiety: A randomized controlled trial. Journal of Affective Disorders, 208, 483-489. doi:
10.1016/j.jad.2016.10.039
Ringeisen, T., & Raufelder, D. (2015). The interplay of parental support, parental pressure and
10.1016/j.adolescence.2015.08.018
Rosenthal, R. (1979). The file drawer problem and tolerance for null results. Psychological
Sarason, S. B., & Mandler, G. (1952). Some correlates of test anxiety. Journal of Abnormal
Sarason, S. B., Davidson, K. S., Lighthall, F.F., Waite, R., & Ruebush, B. K. (1960). Anxiety in
Sarason, I. G. (1981). Test anxiety, stress, and social support. Journal of Personality, 49(1), 101-
Sarason, I.G. (1984). Stress, anxiety and cognitive interference: Reactions to tests. Journal of
Schmidt, F. L., & Hunter, J. E. (2014). Methods of Meta-Analysis: Correcting Error and Bias in
Research Findings. Sage Publications.
Schnell, K., Ringeisen, T., Raufelder, D., & Rohrmann, S. (2015). The impact of adolescents'
gender and test anxiety matter? Learning and Individual Differences, 38, 90-98. doi:
10.1016/j.lindif.2014.12.008
TEST ANXIETY META-ANALYSIS 37
Segool, N., Carlson, J., Goforth, A., von der Embse, N. & Barterian, J. (2013). Heightened test
anxiety among young children: Elementary school students’ anxious responses to high-
Segool, N., von der Embse, N.P., Mata, A., & Gallant, J. (2014). Cognitive behavioral model of
test anxiety in a high-stakes context: An exploratory study. School Mental Health, 6, 50-
Sommer, M., & Arendasy, M. E. (2014). Comparing different explanations of the effect of test
10.1016/j.intell.2013.11.003
Sommer, M., & Arendasy, M. E. (2015). Further evidence for the deficit account of the test
Spielberger, C. D., & Vagg. R.P. (1995). Test Anxiety: A Transactional Process Model. In by C.
D. Speilberger and P. R. Vagg (Eds.) Test Anxiety: Theory, Assessment and Treatment,
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. doi:
10.1037/0022-3514.69.5.797
Thomas, C. L., Cassady, J. C., & Finch, W. H. (2017). Identifying severity standards on the
Cognitive Test Anxiety Scale: Cut score determination using latent class and cluster
Tobias, S. (1985). Test anxiety: Interference, defective skills, and cognitive capacity.
von der Embse, N., Barterian, J., & Segool, N. (2013). Test anxiety interventions for children
Thomas, C. L., Cassady, J. C., & Finch, W. H. (2017). Identifying Severity Standards on the
Cognitive Test Anxiety Scale: Cut Score Determination Using Latent Class and Cluster
von der Embse, N. & Hasson, R. (2012). Test anxiety and high-stakes tests: Implications for
10.1080/1045988X.2011.633285
von der Embse, N., Kilgus, S.P., Segool, N., & Putwain, D. (2013). Evaluation of a brief test
Test Anxiety Scale. International Journal of School and Educational Psychology, 1, 246-
von der Embse, N.P., Mata, A., Segool, N., & Scott, E.C. (2014). Latent profile analysis of test
10.1177/0734282913504541.
von der Embse, N.P., & Putwain, D. W. (2015). Examining the context of instruction to facilitate
doi:10.1177/0143034315612144
TEST ANXIETY META-ANALYSIS 39
von der Embse, N.P., Schultz, B., & Draughn, J. D. (2015). Readying students to test: The
impact of fear and efficacy appeals on student anxiety, motivation, and test performance.
von der Embse, N.P., Scott, E.C., & Kilgus, S.P. (2015). Sensitivity to change and concurrent
validity of Direct Behavior Ratings for Academic Anxiety. School Psychology Quarterly,
von der Embse, N.P. & Witmer, S. (2014). High-stakes accountability: Student anxiety and
10.1080/15377903.2014.888529.
Watson, D., & Clark, L. A. (1984). Negative affectivity: The disposition to experience aversive
Wine, J. (1971). Test anxiety and direction of attention. Psychological Bulletin, 76, 92-104. doi:
10.1037/h0031332
Wolf, L. F., & Smith, J. K. (1995). The consequence of consequence: Motivation, anxiety, and
10.1207/s15324818ame0803_3
Wren, D. G., & Benson, J. (2004). Measuring test anxiety in children: Scale development and
internal construct validation. Anxiety, Stress, and Coping, 17, 227–240. doi:
10.1080/10615800412331292606
Yerkes, R. M., & Dodson, J. D. (1908). The relation of strength stimulus to rapidity of habit-
doi: 10.1002/cne.920180503
Zeidner, M. (1998). Test Anxiety, The State of the Art. New York: Plenum Press.
TEST ANXIETY META-ANALYSIS 40
Zeidner, M., & Matthews, G. (2005). Evaluation anxiety. In A. J. Elliot & C. S. Dweck
(Eds.), Handbook of competence and motivation (pp. 141–163). London: Guilford Press.
Table 1
Test Anxiety with Performance Correlates
Correlati
on
Coeffici
Description of correlational group ent
(Fixed/
Random
Correlate of TA
)
Fail- 95% z
Numb
Samp Safe N CI (Fixed/ End
er of Outlie
le (Fixed/ (Fixed/ Q I2 valu
studie rs
Size Rando Rando Rando es
s
m) m) m)
Achievement
Testing,
by grades
Primary (Grade 1- -.29 to 0.00 -.34/
9 774 11 -6.2 5.00 None -.22*
5) -.15 % -.13
-.29 to
Intermediate -.22/ -13.63/ **23.4 40.16 -.19/
16 2,799 25 -.059 -.25*
(Grade 6-8) -..30 to -9.67 0 % -.111
-.20
-.17 to
Secondary 13,11 17/ -.14/ -18.16/ 75.12 -.42/ -.16*/
27 107.37 None
(Grade 9-12) 7 19 -.21 to -9.15 % -.04 -.17*
-.14
-.25 to
Post-Secondary
20,84 77/ -.23/ -34.86/ **282. 76.98 -.56/ -.24*/
(Undergraduate & 53 .00
9 95 -.29 to -16.71 40 % -.01 -.27*
Graduate)
-.24
-.19 to
-.69,
16,85 26/ -.16/ -22.25/ **51.4 39.70 -.33/ -.17*/
GPA 36 -
5 33 -.21 to -.15.53 1 % -.04 -.19*
.49, .0
-.16
IQ
TEST ANXIETY META-ANALYSIS 41
Nonverbal
-.14 to 4.42 -.29/
Reasoning 7 2,332 0 -.46 12.56 None -.10*
-.05 % .00
Abilities
Cognitive
-.25 to -.8.50/ 10.62 -.41/
Proficiency 9 1,711 10 11.19 None -.21*
-.16 -8.01 % -.05
Abilities
-.28 to
-
10/ -.19/ **17.6 48.97 -.025, -.31/ -.24*/
Verbal Abilities 7 1,771 .10.14/
9 -.29 to 4 % .00 -.02 -.23*
-6.82
-.16
Standardized -.28 to
Exams 44/ -.24/ -23.40/ 56.98 -.53/ -.26*/
26 7,748 92.97 None
47 -.30 to -14.78 % -.07 -.27*
-.24
-.34 to
-
SAT/ ACT/ -.28/ 48.14 -.53/
13 2,741 29 .16.64/ 28.93 None -.31*
Entrance Exam -.36 to % -.07
-11.36
-.26
State Exam,
by grades
-.26 to 0.00 -.18/
Primary 2 343 1 -2.89 0.42 None -.16*
-.05 % -.09
-.49 to 0.00 -.34/
Intermediate 2 350 7 -7.92 0.18 None -.40*
-.31 % -.39
-.26 to
7/ -.19/ -13.19/ 67.65 -.40/ -.23*/
Secondary 5 3,287 27.82 None
8 -.31 to -7.63 % -.12 -.25*
-.19
-.28 to
-.16/ -7.18/ 35.54 -.32/
Professional Exam 7 1,027 9 9.31 None -.22*
-.29 to -5.25 % -.09
-.14
Cognitive
Component
-.31 to
Achievement 10,24 42/ -.28/ -30.50/ **110. 60.94 -.53/ -.29*/
21 -.03
Testing 5 50 -.36 to -17.17 08 % -.11 -.32*
-.29
-.26 to
15/ -.22/ -20.63/ 49.55 -.37/ -.24*/
GPA 10 6,952 21.80 None
16 -.29 to -12.01 % -.11 -.25*
-.21
-.29 to
-.22/ -12.21/ 52.45 -.38/
IQ 5 2,245 8 25.23 None -.25*
-.31 to -8.21 % -.11
-.19
-.29 to
Standardized 17/ -.24/ -20.45/ 63.28 -.50/ -.26*/
10 5,785 68.09 None
Exams 18 -.31 to -11.82 % -.05 -.27*
-.23
Affective/Physiolo
gical Component
-.20 to
Achievement 12/ -17.01/ **99.0 67.68 -.47/ -.18*/
15 8,564 -.16/ .089
Testing 9 -6.95 1 % .09 -.16*
-.21 to
TEST ANXIETY META-ANALYSIS 42
-.12
-.10 to
0/ -.05/ -6.08/ 64.71 -.19/ -.08*/
GPA 7 6,112 19.84 None
0 -.11 to -2.21 % .07 -.06
-.01
-.26 to
-.17/ -9.84/ 56.31 -.33/ -.22*/
IQ 4 1,811 5 18.31 None
-.28 to -5.89 % -.05 -.21*
-.14
-.15 to
Standardized 3/ -.10/ -.9.33/ 56.71 -.38/ -.13*/
9 5,598 60.06 None
Exams 4 -.19 to -6.59 % -.02 -.14*
-.10
-.07
Behavioral 0/ to .14/ 74.64 -.18/ .04/
3 324 6.56 11.83 None
Component 0 -.15 % .32 .08
to .30
-.19 to
-.05/ -3.62/ 88.50 -.32/ -.12*/
Social Component 3 852 1 43.48 None
-.32 -1.32 % .09 -.13
to .06
* p < .01
**Outliers were examined for cause of variability
TEST ANXIETY META-ANALYSIS 43
Table 2
Test Anxiety with Intrapersonal Variables
Description of correlational group Correlation
Coefficient
(Fixed/Rand
om)
Correlate of
Fail- 95%
TA Numb z
Samp Safe N CI End
er of (Fixed/ Outlie
le (Fixed/ (Fixed/ Q I2 valu
studie Rando rs
Size Rando Rando es
s m)
m) m)
Self-Concept
-.02
0/
Locus of to .10/ 1.40/ 94.33 -.29/ .04/
6 1,082 0 88.22 None
Control -.25 0.00 % .35 .00
17
to .25
-.22 to
Self- 10/ -.15/ -9.56/ 24.70 -.27/ -.19*/
11 2,599 13.28 None
Regulation 9 -.23 to -7.96 % -.06 -.18*
-.14
-.33 to
72/ -.30/ -30.99/ 121.4 72.82 -.58/ -.31*/
Self-Efficacy 32 9,162 None
76 -.36 to -16.22 3 % -.14 -.32*
-.29
-.48 to
-.36/ -12.05/ 38.06 -.56/
Self-Esteem 5 744 18 6.46 None -.42*
-.50 to -9.49 % -.37
-.34
-.32 to
-
Academic 21/ -.25/ **24. 55.81 .43/ -.28*/
11 2,735 .15.11/ .50
Confidence 22 -.34 to 89 % -.14 -.29*
-9.89
-.23
Motivation
-.08 to
Intrinsic 0/ -.01/ -2.81/ 59.84 -.18/ -.05*/
13 3,886 29.88 None
Motivation 0 -.12 to -2.33 % .11 -.07
-.01
12
Extrinsic to .17/ 10.05/ 30.38 .04/
5 4,894 2 5.75 None .14*
Motivation .10 6.59 % .20
to .18
Self-Coping
Skills
.32
Avoidance 23/ to .41/ 13.79/ **10. 54.06 .28/ .36*/
8 1,323 .15
Coping 24 .30 9.35 88 % .47 .37*
to .44
-.20 to
Approach 4/ -.10/ -6.18/ 52.06 -.28/ -.15*/
8 1,732 14.60 None
Coping 5 -.23 to -4.48 % -.04 -.16*
-.09
Achievement
Goals
Mastery- 11 2,146 0/ -.00 1.82/ **19. 53.17 -.23 -.1/ .04/
TEST ANXIETY META-ANALYSIS 44
Table 3
Test Anxiety with Predictors
Correlati
on
Coeffici
Description of correlational group ent
(Fixed/
Random
Correlate of TA
)
Fail- 95% CI z
Numb
Sam Safe N (Fixed/ (Fixed End
er of Outlie
ple (Fixed/ Random) / Q I2 value
studie rs
Size Random Rando s
s
) m)
Gender, by grades
Primary (Grade 1- 1,13 .16 to .27/ 7.19/ 23.5 .03/
5 6 5.24 None .21*
5) 6 .13 to .28 5.43 9% .34
Intermediate 1,74 .11 to .20/ 6.53/ **7. 45.0 .04/
6 4 .03 .16*
(Grade 6-8) 1 .09 to .22 4.65 28 6% .27
Secondary 5,57 19/ .24 to .27/ 18.93/ **13 35.3 .04/ .25*/
12 .03
(Grade 9-12) 7 18 .20 to .27 12.73 .92 4% .3 .24*
Post-Secondary .0/
7,59 32/ .20 to .24/ 19.29/ **38 43.5 .06/ .22*/
(Undergraduate & 26 .51/
0 30 .18 to .25 13.13 .98 7% .35 .21*
Graduate) .64
Ethnicity (Cau vs. 1,34 .09/
7 1 .07 to .17 4.41 1.38 0.0% None .12*
AA) 0 .21
Ability (Low vs. -.19 to -
High) 2,14 4/ .11/ -7.11/ **10 51.1 -.34/ -.15*/
7 -.40
2 6 -.24 to - -5.11 .24 6% -.08 -.18*
.11
Disabilities
2,41 1/ .08 to .15/ 5.63/ 39.1 87.2 .0/ .12*/
ADHD 6 None
0 7 .08 to .32 3.21 7 4% .43 .21*
2,05 0/ .03 to .12/ 3.46/ 11.4 56.4 -.01/ .08*/
Learning Disability 6 None
0 1 .03 to .19 2.85 8 5% .17 .11*
Testing Conditions
Perceived
1,10 .25 to .36/ 10.49/ 10.8 63.1 .18/ .31*/
Difficulty (Low vs. 5 11 None
1 .20 to .39 5.96 5 2% .41 .30*
High)
Evaluative vs. .17/
5 881 6 .15 to .27 6.39 1.64 0.0% None .21*
Non-Evaluative .27
Prior Experience 3/ .09 to .23/ 4.69/ **4. 27.8 .09/ .16*/
5 843 .41
vs. No Experience 4 .09 to .25 4.05 16 5% .28 .17*
* p < .01
**Outliers were examined for cause of variability
TEST ANXIETY META-ANALYSIS 46
Highlights:
Test anxiety was significantly and negatively related to a wide range of educational
performance outcomes, including standardized tests, university entrance exams, and
grade point average.
Test anxiety effects were most pronounced at the middle grades level.
Self-esteem was a significant and strong predictor of test anxiety.
Perceived difficulty of the test and test consequences were also related to higher test
anxiety.