1 GENERAL ANXIETY AND ACADEMIC INDICA TORS AS PREDICTORS OF TEST ANXIETY IN ADOLESCENTS By ANNE LARMORE BRUEHL A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009
2 2009 Anne Larmore Bruehl
3 To my family
4 ACKNOWLEDGMENTS Many people have supported this project and helped to m ake its completion possible. I extend many thanks to my advisor and dissertat ion chair, Dr. Nancy Waldron, for her guidance and encouragement throughout this entire pro cess. Throughout my doctoral work, she has skillfully helped me to focus and refine my ideas and strengthen my professional identity. I have been extremely fortunate to have an excellent doctoral committee. I thank my committee membersDr. Diana Joyce, Dr. Tina Smith-B onahue, and Dr. Eric Storchwho have been tremendously supportive and helpful throughout my doc toral work. Each of these individuals has played an important role in my professional development as a prac titioner as well as a researcher. I could not have asked for a more knowledgeable and supportive committee. My friends and colleagues have also been a major positive influence in the completion of this project and this degree. I would like to thank the members of my dissertation support group for encouraging me to be assertive about the completion of my goals. The support, advice, and laughter offered by my fellow graduate students and dear friends have been essential to my success and my sanity throughout my doctoral work. Finally, I thank my wonderful family. My pare nts, Betty and Dave La rmore, have provided much love, wisdom, encouragement. My mom ha s been my biggest cheerleader and a great friend as I work to accomplish my goals. Leading by example, my dad has taught me that there is a difference between knowledge a nd wisdom, to not sweat the small stuff, and to always live with integrity. Finally, I thank my sweet husband, Jason Bruehl, for all his support, love, and patience. The completion of this degree and this project would not have been possible without the wonderful people supporting me.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ............................................................................................................... 4LIST OF TABLES ...........................................................................................................................7ABSTRACT ...................................................................................................................... ...............8 CHAP TER 1 REVIEW OF THE LITERATURE: TEST ANXIETY .......................................................... 10Introduction .................................................................................................................. ...........10Test Anxiety: Overview, Context, Pr evalence, and Associated Outcomes ............................ 11Context ....................................................................................................................... .....13Prevalence .................................................................................................................... ....14Associated Outcomes ...................................................................................................... 14Demographics and Group Differences in Risk for Test Anxiety ........................................... 16Gender ........................................................................................................................ .....16Race .......................................................................................................................... .......17Socioeconomic Status ...................................................................................................... 18Age ........................................................................................................................... .......19Test Anxiety Literature: Two Major Frameworks .................................................................. 20The Academic Approach: Models Linking Te st Anxiety to Academic Performance ..... 21Interference model .................................................................................................... 21Deficits model ..........................................................................................................22Information processing or hybridized model ........................................................ 23Applications of the information-processing model .................................................. 24Limitations of the Existing Models .................................................................................25Interference and Deficits: Understanding Te st Anxiety with Academic and Emotional Variables ..................................................................................................................... ........26Correlates and Predictors of Test Anxiety .......................................................................26Emotional Variables ........................................................................................................ 26Academic Variables .........................................................................................................31Integrating Academic and Emotiona l Contributors to Test Anxiety ...................................... 35The Academic Approach to Test Anxiety ................................................................... 36The Emotional Approach to Test Anxiety ................................................................... 37Existing Integrations of Academic and Emotional Factors and Test Anxiety ................ 38Hypotheses and Research Questions ...................................................................................... 382 METHOD ........................................................................................................................ .......40
6 Participants .................................................................................................................. ...........41Research Setting .............................................................................................................. 41Sample: Recruitment and Demographics ........................................................................42Design & Procedure ................................................................................................................42Data Collection Procedures .............................................................................................43Inclusion and Exclusion Criteria ..................................................................................... 44Instrumentation of Relevant Variables ................................................................................... 45Test Anxiety ....................................................................................................................45General Anxiety ...............................................................................................................47Perceived Academic Competence ................................................................................... 49Additional Variables of Interest ...................................................................................... 503 RESULTS ....................................................................................................................... ........54Descriptive Statistics across Demographics ...........................................................................54Inferential Statistics: Correlations and Multiple Regressions ................................................ 56Research Questions ......................................................................................................... 56Research question 1 ..................................................................................................56Research question 2 ..................................................................................................57Research question 3 ..................................................................................................58Research question 4 ..................................................................................................59Research question 5 ..................................................................................................60Follow-up Analyses ......................................................................................................... 604 DISCUSSION .................................................................................................................... .....66Demographics and Descriptive Findings ................................................................................67Gender ........................................................................................................................ .....67Grade level .......................................................................................................................69Race/Ethnicity ................................................................................................................ .69Correlates of Test Anxiety ......................................................................................................70Academic Indicators ........................................................................................................70Relationship among Various Academic Indicators ......................................................... 71Test Anxiety and Othe r Forms of Anxiety ......................................................................72Test Anxiety and Gender ....................................................................................................... .73Predicting Test Anxiety ..........................................................................................................74Predictors of Test Anxiety in Males and Females .................................................................. 76Implications of Findings ...................................................................................................... ...77Theoretical Implications .................................................................................................. 77Implications for Research ................................................................................................ 78Implications for Practice .................................................................................................. 78Study Limitations ............................................................................................................. .......79Summary & Future Directions ................................................................................................ 83REFERENCES .................................................................................................................... ..........85BIOGRAPHICAL SKETCH .........................................................................................................94
7 LIST OF TABLES Table page 2-1 Sample Demographics: Number (and %) of Male and Fem ale Participants for Each Grade and Race ..................................................................................................................533-1 Descriptive Statistics of Measures for Males, Females, and Total Sample ....................... 623-2 Matrix of Correlations for Entire Sample ..........................................................................623-3 Forced-Entry Regression of Test A nxiety regressed onto pr edictor variables .................. 623-4 Hierarchical Regression ................................................................................................... ..633-5 Matrix of Pearson Correlations Amo ng Major Variables for Males, Females .................. 643-6 Forced-Entry Regression Model for Males, Females ........................................................ 643-7 Correlations among MASC subscales and TAI ................................................................. 65
8 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy GENERAL ANXIETY AND ACADEMIC INDICA TORS AS PREDICTORS OF TEST ANXIETY IN ADOLESCENTS By Anne Larmore Bruehl August 2009 Chair: Nancy Waldron Major: School Psychology Despite several decades of research addressing test anxiet y, few studies have examined academic and emotional variables as they jointly relate to test anxiety. Two frameworks guide the existing test anxiety literature. The line of in terest most prominent in educational research focuses on test anxiety primarily as an academic problem relating to impairments in test-taking (e.g., Naveh-Benjamin, 1991). Another framework stem ming mostly from the clinical l iterature views test anxiety as an emotional problem relating to other broader forms of anxiety (e.g., Beidel, 1988). However, studies have not examined the extent to which various indicators of academic performance and general anxiety jointly pr edict test anxiety. Key research questions in this study examined the extent to which academic variables, general anxiety, and selected demographics predict test anxiety in male and female adolescents. Self-report rating scales addres sing test anxiety, general anxi ety, and students perceptions of academic skills and academic enabling behavi ors were obtained from 104 adolescents in grades seven through nine. In addition, particip ants grade point average and performance on a statewide achievement test were collected. Multip le regression analyses revealed that general anxiety and performance on a stat ewide achievement test were f ound to be significant predictors of test anxiety. A hierarchical regression analyses indicated that student perceptions of their
9 academic skills also contributed an additional small portion of the variance associated with test anxiety. Although some minor gender differences we re identified, the predic tive model including general anxiety and statewide achievement scores was applicable to both males and females. This study extends the existing informationprocessing model (Naveh-Benjamin, 1991) linking test anxiety with lower academic performance by incorporating the role of general anxiety in predicting test anxiety.
10 CHAPTER 1 REVIEW OF THE LITERATURE: TEST ANXIETY Introduction Test anxiety is a growing concern for both the academ ic and emotional well-being of children in public schools today. In recent years, the greater emphasis on educational accountability and high-stakes tes ting stemming from federal regula tions such as the No Child Left Behind Act means that many students are bei ng tested more than ever before. Although the concept of test anxiety has been documented by sc ientific and scholarly research with students across a wide range of age and grade levels (Cizek & Burg, 2007; Ergene, 2003), the processes operating in the experience of test anxiety have received minimal attention and remain not fully understood (Zeidner, 1998). Two distinct frameworks exist in the professional literature with regard to test anxiety. One of the major guiding theories evident in the litera ture places test anxiety centrally as a problem relating to academic performa nce (Naveh-Benjamin, 1991). At the same time, another smaller body of literature explains test anxiety as pr imarily representing an emotional-behavioral experience (Beidel & Turner, 1988). In accordance w ith these schools of thought, test anxiety has been linked to several academic and emotional variables. However, the relative role and influence of these various factors in explaining the experience of test anxiety in adolescents and predicting which students are at highest risk has not been sufficiently explored. Although previous research has revealed a consistent relationship between academic performance and emotional functioning (e.g., Swanson & Howell, 1996), the extent to which academic and emotional variables jointly predict to the expe rience of test-anxiety in adolescent students remains unclear. Furthermore, it is unknown whet her distinct patterns of academic versus emotional variables may best predict test anxiety.
11 The text that follows will review the literature on test anxi ety and associated variables. First, a general overview of the construct of test anxiety, its prevalence, and associated outcomes will be provided. Consideration will be given to demographic variables that predict groups that may be at increased risk for experiencing te st anxiety (Hembree, 1988; Osborne, 2001). Three models of test anxiety and its relationship with impaired academic performance will be introduced. Within the framework of these models of test anxiety, a discu ssion of correlates and predictors of test anxiety supporte d in the literature will follow. Emotional variables relevant to test anxiety will be discussed. Next, academica lly-related variables will be examined as they relate to test anxiety. Although pr evious studies regarding academi c and emotional variables that play a role in the experience of test anxiety consistently conc lude that test anxiety is a multifaceted phenomenon with numerous associated variables (e.g., Hembree, 1988), these studies fail to address how sets of variables may operate together to explain test anxiety. Test Anxiety: Overview, Context, P revalence, and Associated Outcomes Test anxiety refers to a set of physiological, behavioral, emotional, and cognitive responses to the perceived threat of fa ilure on a test or other evalua tive situation (e.g., Sapp, 1999). Among scholars, the most widely agreed-upon con ceptualization describe s test anxiety as a multidimensional unpleasant state. Test anxiet y involves an interac tion among physical and emotional tension and autonomic nervous system arousal, cognitive worry and doubt in ones abilities, and behavioral responses that interf ere with test preparedness (e.g., avoidance and procrastination of studying) (e.g., Ergene, 2003; Sarason & Sarason, 1990; Spielberger, 1972; Zeidner, 1998). Zeidner describe s test anxiety as the respon ses that accompany concern about possible negative consequences or failure on an exam or similar evaluative situation, (pp. 1718). A student experiencing test anxiety may present as easily distracted, forgetting previously known information, or freezing up when taking an exam (Emery & Krumboltz, 1967; Keogh,
12 Bond, French, Richards, & Davis, 2004). Other presentations of a test-anxious student may include negative emotions, negative thoughts abou t the self, and feelings of panic (Zeidner, 1998). Students with anxiety about tests may exhibit any combination of associated physiological, cognitive, emotional, or behavioral symptoms (Sarason, 1984). Across several decades of research e xploring test anxiety, a multidimensional conceptualization of test anxi ety has been widely agreed upon (Cizek & Burg, 2006; Zeidner, 1998). That is, test anxiety represents a comp lex phenomenon consisting of multiple symptoms that present somewhat differently across individu als. The first systematic study of test anxiety occurred in the 1950s, resulting in a portrayal of test anxiety as a multidimensional construct (Zeidner, 1998). Two major components of test anxiety have been em phasized, including both cognitive and emotional components, often referred to as worry and emotionality (Cizek & Burg, 2006; Liebert & Morris, 1967). Worry includes cognitive concerns about evaluation and consequences of failure, while emotionality consists of how an individual perceives physiological autonomic reactions to an evalua tive situation (Liebert & Morris, 1967). Worry or cognitive test anxietycan include a variety of thoughts a bout an evaluative situation. Specifically, these cognitions may occur in th e form of social comparisons about ones performance versus peers, negative outcomes asso ciated with failure, limited confidence in ones performance and preparedness, and fears of disa ppointing other individuals such as parents (Cassady & Johnson, 2002; Hembree, 1988). Specific to the affective dimension of test anxiety, physiological reactions can include sensations such as increased heart rate dizziness, nausea, or panic (Morris, Davis, & Hutchings, 1981). Emoti onality consists of an individuals subjective interpretation of such autonomic reactions (S chwarzer, 1984). Much research has supported the psychometric distinction between these two co mponents (Everson, Millsap, & Rodriguez, 1991;
13 Schwarzer,1984; Spielberger, 1980; Stober, 2004 ). In addition to cognitive and affective dimensions of test anxiety, a behavioral compone nt of test anxiety has also been acknowledged, pointing to behaviors such as procrastination and avoidance, as well as deficient skills related to studying and test-taking (Paulman & Kennelly, 1984; Zeidner, 1998). Context Em phasis on educational accountability and efforts to improve student achievement through high-stakes assessments has been highlighted in recent decades. The authorization of the national No Child Left Behind Act in 2002 mandated that each stat e department of education adhere to stricter accountability guidelinesinc luding annual testing in reading, mathematics, and writing of each student in grade three through eightin an effort to document and improve student achievement outcomes. The term high st akes testing has become familiar language in the national and local media. The emphasis on ac hievement on crucial tests has pervaded school and classroom environments. Test anxiety represents an unintended consequence of increased emphasis on test performance. Smyth (2008) asserted that due to test anxiety and other related negative side effects of high-stakes testing, subgroups of the populationin particular students from lower socioeconomic status backgrounds, minorities, English language learners, and special education studentsare especially negatively affected. The cost of high-stakes testing nationally was estimated between $20 and $50 billion annually, which is between 5.5 and 14 percent of every dollar spent for public schools (Center for Educational Policy, 2003). The state of Fl orida reports spending over $50 million annually on the development, norming, administration, scoring, and reporting of the Florida Comprehensive Academic Test (FCAT), which breaks down to approximately $19 per pupil to take the test in 2008 (Florida Department of Education [FDOE], 2008). Despite these costs, test anxiety may be
14 undermining the meaning of test results for so me students. Haladyna & Downing (2004) noted that test anxiety introduces conc erning amounts of construct-irre levant variance in high-stakes tests, thus representing a serious threat to the vali dity of test results. Prevalence W ithin the context of this increased na tional emphasis on high-stakes testing, many researchers have suggested that te st anxiety appears to have incr eased over time (Cizek & Berg, 2006; Wren & Benson, 2004). Test anxiety is a wide spread concern, with increasing numbers of students suffering from its detrimental eff ects (McDonald, 2001). Test anxiety has been examined in countries from numerous regions of the world, suggesti ng that this problem transcends national and cultural boundaries (Bodas & Ollendick, 2005; McDonald, 2001; Putwain, 2007). Previous estimates of students experiencing test anxiety range from 10% (Kondas, 1967) to 30% (e.g., Nottelmann & Hill, 1977) for school-age children. Early research estimated that two or three students per classroom suffered from severe levels of test anxiety (Hill & Wigfield, 1984). Somewhat more recently, prevalence estimates between 34% and 41% have been reported for children between the ages of 8 and 12 in the third through sixth grades (Beidel, 1991; Turner, et al., 1993). In light of increased frequency of te sting in schools, these figures are likely to be higher in the last de cade (Cizek & Burg, 2006; Zeidner, 1998; Whitaker Sena, Lowe, & Lee, 2007). This purported increase in recent decades has al so been explained as resulting from increased achievement pressures from parents and teachers, higher expectations for mastering more complex material at an ear lier age, and increased pressure to achieve on national or state-wide standardized testing (Locker & Cropley, 2004; Tu rner et al., 1993). Associated Outcomes Test anxiety has been linked with detrim enta l outcomes in regard to academic, social, and emotional functioning. Birenbaum and Nasser (1994) considered test anxiety to be one of the
15 most disruptive problems in schools. Academically, test anxiety can have a negative cognitive impact at all phases of learning new informa tion, studying and preparation, and taking tests (Cassady, 2004). Test-anxious children are at elev ated risk for receiving worse grades, being retained, and having lower test scores than children without test anxiety (Warren, Ollendick, & King, 1996). However, a distinction exists between facilitating and deb ilitating forms of anxiety, such that test anxiety does not impair functioning for all individuals. Modest amounts of anxiety about a test can be asso ciated with elevated performa nce for some students (Alpert & Haber, 1960). Although small or moderate levels of anxiety surrounding ones performance on a test often represent a norm al part of development excessive levels of te st anxiety can have detrimental effects on student s test-taking ability. In addition to its immediate academic effects, test anxiety can relate to lower motivation (Hancock, 2006), and lower levels of educati onal and occupational attainment (e.g., Zeidner, 1998). Given that some severely test-anxious students behavioral response is avoidance of the anxiety-provoking experience of te st-taking (e.g., McDonald, 2001), te st anxiety appears to place students at elevated risk fo r dropping out of school (Sc holze & Sapp, 2006). Scholze and Sapp suggested that dropout rates for students from culturally and li nguistically diverse backgrounds in particular may be affect ed by test anxiety (2006). Test anxiety has been associated with lower self-esteem, dependency, and passivity (Campbell, 1986). The experience of se vere test anxiety is thought to contribute to school refusal in some students (Ollendick & Meyer, 1984). At the social and emotional level, anxiety beyond the normal range can place students at risk for problems such as poor social relationships, chronic mental health problems, substance a buse, and suicidal tendencies (Merrell, 1999). Anxious children may avoid peer interaction and interact less co mpetently with peers (Barrett &
16 Heubeck, 2000). Cross-sectional stud ies have confirmed a relations hip between elevated anxiety and impairments in social functioning and peer relationships (Langle y, Bergman, McCracken, & Piacentini, 2004). Indeed, reducing students an xiety is associated with improved social functioning in addition to improvements in academic performance (Wood, 2006). This finding supports the idea of general anxiety playing a role in the academic impairment associated with test anxiety; however, it does not account for the ways in which academic variables may simultaneously operate in test anxiety. Test anxiety represents a significant problem with aversive effects on students academ ic and emotional experiences. Demographics and Group Differences in Ris k for Test Anxiety Although the processes surrounding test a nxiety are undoubtedly complicated and multidimensional, a body of literature indicate s support for certain non-causal demographic factors that may be useful in predic ting elevated risk for test anxiety. Gender From elementary school through medical school and across cultures, females typically report higher levels of test anxiety than males (Hembree, 1988; Putwain, 2007; Stober, 2004; Zeidner, 1998). Based on a meta-analytic review of 143 studies examining gender differences, Hembree (1988) found that females endorsed significantly higher levels of test anxiety than males from grades 1 through 12 and in colleg e. More recently, Wren and Benson (2004) found that from a sample of 261 children grades 3 through 6, girls test anxiety scores were significantly greater than boys. On surveys of general fear s, girls reported more evaluation and test-related anxiety than boys (e .g., King et al., 1989). Findings about higher reported test anxiety in females have been replicated using various specific test anxiety measures (Di Maria & Di Nuovo, 1990; Zeidner & Safir, 1989; Whitaker Sena Lowe, & Lee, 2007). Furthermore, this pattern of findings appears to extend beyond samples of elemen tary and secondary students.
17 Female college students have reported higher mean test anxiety scores than males, particularly with regard to the emotionality component of test anxiety (Corcoran, Macdougall, & Scarborough, 1985; McDonald, 2001). Despite these a pparent gender differen ces, when different cut-off scores are used to identify significant degrees of test anxiety in boys versus girls, the prevalence becomes more similar across sexe s (McDonald, 2001; Turner et al., 1993). Multiple theories have attempted to explain th e observed gender difference in test anxiety. One prominent explanation noted in the literature is that differences in socialization patterns play a role (Zeidner, 1998). This theory posits that from an early age wo men are socialized to express emotions whereas men learn to minimize emotional experiences (e.g., Maccoby & Jacklin, 1974, as cited in Lowe & Lee, 2008). Self-report data on general and specific ty pes of anxiety may in part reflect differences in reporting and willingn ess to admit anxiety (e.g.,Whitaker Sena, Lowe, & Lee, 2007). Other explanations of gender differen ces in test anxiety rela te to coping styles (e.g., Zeidner, 1998) and biological and hormonal factors (Altemus, 2006). The pervasiveness of higher anxiety in females across cultures and subt ypes of anxiety supports the role of biological processes (Altemus, 2006). Race In addition to gender differences, some eviden ce has suggested that demographic variables such as race and socioeconomic status may be re levant to the experience of test anxiety. For example, results from a meta-analysis reflecte d higher levels of test anxiety among African American and Hispanic students during elementa ry school, with the effects of race no longer evident in samples of high-school students (Hembree, 1988). However, other authors have concluded that test anxiety is roughly equivalent in Caucasian a nd African American children in grades 3 through 6 (Beidel et al., 1994). Asian-American students may also experience higher
18 levels of test anxiety; a small sample of Asian-American middle school students reported significantly higher levels of test anxiety than their Eur opean-American peers (Pang, 1991). More recently, it was found that African America n, Latino, and Native American students all had significantly higher anxiety related to test-taking compared to their White counterparts (Osborne, 2001). When achievement was covaried with anxi ety and race, African Americans and Latinos continued to have higher levels of anxiety than Whites; howev er, Native Americans no longer showed significantly higher levels of test anxiety (Osborne, 2001). Thus, although race and ethnicity have some relationship with test anxiety, other factors such as achievement may explain some of the variance in the experience of test anxiety. Socioeconomic Status Socioeconomic characteristics have also been linked with test anxi ety, with students from lower socioeconomic backgrounds experiencing greater levels of test anxi ety than students from middleand upper-class backgrounds across several cultures (Guida & Ludlow, 1989; Zeidner & Safir, 1989). Several explanations have been posed to account for this phenomenon, including differences in parenting and home variables as well as school and teacher variables (e.g., Cizek & Burg, 2006; Zeidner, 1998). An explanation that has received some empi rical support is the phenomenon of stereotype threat; which has been found to play a role in the existence of higher test anxiety in lower-income as well as minority students. In a recent study of stereotype threat and test anxiety in college students, Harrison and colleagues (2006) found that students from lower-income backgrounds exhibi ted higher levels of test an xiety and decreased academic performance specifically when they were informed prior to an academic test that the information obtained would be used for dia gnostic purposes to understand why lower income students have lower academic abilities. Stereotype threat has also been studied in racial minority groups and in
19 women. For example, African American student s performed worse than European American students when informed that a test would be used to measure their academic abilities but performed generally the same when not informed of the purpose of the assessment (Steel & Aronson, 1995; Osborne, 2001). Thus, the role of stereotype threat may represent a possible explanation for socioeconomic, racial, and gende r variations occurring in test anxiety and performance. Age Som e developmental differences in the expe rience of test anxiety have also been suggested in the literature (Hembree, 1988; Whitaker Sena, Lowe, & Lee, 2007). However, the pattern is somewhat inconsistent and must be interpreted with ca ution due to the cross-sectional designs often used in the research (Zeidner, 1998 ). Hembrees (1988) meta-analysis revealed that test anxiety increased from gr ades 1 through 5 and then stabil ized throughout middle school and high school. Other findings revealed an increase in test anxiety in middle school that stabilized in high school (Wigfield & Eccles, 1989). More recently, Whitake r Sena and colleagues (2007) found age-related differences within specific symp toms of test anxiety. Specifically, they found that older students experienced higher levels of cognitive symptoms and less physiological arousal, and hypothesized that perhaps physio logical symptoms are more common among children and cognitive symptoms are more apparent in adolescents and adults. Although demographic variables ar e of interest in understandi ng aspects of the processes underlying the experience of test anxiety, other variables (e.g., academic achievement) represent more prominent correlates (e.g., Osborne, 2001) Several theorists a nd researchers have attempted to explain this relationship between test anxiety and observed impairments in testtaking.
20 Test Anxiety Literature: Two Major Frameworks Several decades of test anxiet y research reveal a division in the general focus of studies identified in the literature. Th ese studies focus on either understa nding test anxiety as prim arily an academic problem or test anxiety as primarily an emoti onal problem. In educational research, test anxiety has been approached as a school-specific problem, with a focus on detrimental academic correlates (e.g., Cassady & Johnson, 2002). Drawing primarily from the clinical literature, a relationship has b een identified among test anxiety and other anxiety disorders (e.g., Beidel & Turner, 1988; Bodas & Ollendick, 2005). Thus, this line of research focuses on test anxiety largely as an individual emotional problem. Despite this somewhat dichotomous approach to understanding test anxiety, more theoretical and empirical work has been done examining the relationship of test anxiety wi th impaired academic functioning. Few empirical studies have examined academic and emotional vari ables together and their relationship to test anxiety. The next section of this review will examine both academic and emotional conceptualizations of test anxiety in the literatu re. First, three models of test anxiety and the relationship to academic functioni ng will be introduced. Although th ese three models explain the effects of test anxiety on academic performance, it is noted that emotional variables are likely to play a role in the processes suggested in each of the models discussed below. Further attention will be given to the third, information-processing model, as it allows a broader conceptualization of test-anxiety (e.g., Naveh-Benjamin, 1991) and is more compatible with the inclusion of emotional processes operating in test anxiety. Including both academic and emotional variables in existing models of test anxi ety is of central interest in this study, and these will be subsequently discussed in the text.
21 The Academic Approach: Models Linking Test Anxiety to Academic Performan ce Three major models have been proposed to explain the association between test anxiety and impairments in test-taking and academic ach ievement: interference model, deficits model, information processing model. Each of these mode ls has received some support in the literature and provides insights regardi ng the processes by which test anxiety may develop and be maintained (Bodas & Ollendick, 2005). Interference model Cognitive interference refers to interfering thoughts that impede the direction of ones total attention during learning or test -taking situations (Sarason & St oops, 1978). In the case of test anxiety, cognitive interference occurs in the form of thoughts and fears about ones performance and consequences of poor test outcomes (e.g., Sa rason, 1984). Thus, the negative influence of test anxiety on student academic performance has often been e xplained as resulting from the diversion of cognitive resources (e.g., distraction) from the academ ic task at hand that interfere with the processes of information encoding and retrieval (Morris, Davi s, & Hutchings, 1981; Wine, 1971). Task-irrelevant thoughts (e.g., worr ies about performance) during a test-taking situation hinder ones ability to learn and stor e new information as well as their ability to demonstrate previously learned information dur ing the testing situati on. Data indicate that students who report freezing up during a test experience cognitive interference during all stages of the learning-testing cycle, resulting in difficulties retrieving previously-learned information during the test in addition to probl ems in the initial proc essing and storage of information prior to a test (Naveh-Benjamin, 1991) Within the framework of this interference model of test anxiety, the cognitive component of test anxiety specifically (e.g., worry) has received much attention as being associated with detrimental academic outcomes, while the emotionality component of test anxiety does not appear to have the same link to negative
22 academic outcomes (e.g., Cassady & Johnson, 2002; Hembree, 1988; Kim & Rocklin, 1991; Morris & Liebert, 1970; Wine, 1971). The associ ation between worry and reduced academic achievement has been observed in both adolescents (Williams, 1991) and college students (Bandalos, Yates, & Thorndike-Christ, 1995). For example, a classic study by Sarason and Palola (1960) revealed that unde r high-pressure testing conditi ons, students experiencing test anxiety perform worse on a simple memory task. Replications of this work provided additional support that highly test-anxious college stude nts performed worse on a memory task and experienced greater cognitive interference than st udents reporting low levels of test anxiety (Sarason, 1984). Overall, this li ne of research supported the relationship among test anxiety, worry, cognitive interference, and reduced academic outco mes (Cassady & Johnson, 2002). However, these findings do not inform the extent to which previous academic experiences or existing skills deficits may play a role in the development of test anxi ety, nor do they provide information about trait-like individual charact eristics that may in fact predict cognitive interferences (e.g., the extent to which student s may exhibit worry and emotionality across situations). Furthermore, this model does not expl ain the experience of te st anxiety in students for whom academic performance is not significantly impaired; that is, students who exhibit test anxiety but continue to perf orm successfully on tests. Deficits model Theoretical weaknesses of the interference model of test anxiety were highlighted when certain intervention strategies suc cessfully reduced symptoms of te st anxiety but did not result in the expected academic improvements in many students (Tryon, 1980; Tobias, 1985). Thus, it was proposed that the connection between test-anxie ty and academic struggles may in fact reflect lower abilities and deficient skills related to studying and test-taki ng (Culler & Hollahan,1980; Tobias, 1985). A skills-deficit model has receive d empirical support as evidenced by students
23 responsiveness to interventions designed to e nhance study skills and te st-taking strategies (Beidel, Turner, & Taylor-Ferreira, 1999; Er gene, 2003; Naveh-Benjamin, 1991; Zeidner, 1998). Specifically, the Testbusters program includes 11 sessions to teach study skills and test-taking strategies to children in grad es 4 through 7 (Beidel et al., 1999). This program has been evaluated with children experiencing both academic difficulties and test anxiety, demonstrating promise as an intervention for both improving ac ademics and decreasing test anxiety (Beidel et al., 1999). Furthermore, this line of research de monstrates the existence of a group of students whose test anxiety may be likely to be most st rongly predicted by academic types of variables. However, these studies did not in clude variables reflecting general anxiety or trait anxiety in the sample of children with academic deficits and test anxiety, thus limiting the extent to which conclusions can be drawn regarding the relationship among emotional variables and academic variables in the process of devel oping test anxiety. In addition, re search on the effectiveness of academic-focused interventions (e.g., Testbusters) primarily focuses on a sample of students identified based on meeting criteria related to having both academic deficits and test anxiety. Tobias (1985) argued that the de ficits model of test anxiety does not account for highly testanxious students with strong academic and st udy skills. Finally, although general self-esteem within various domains was considered, the role of student beliefs and perceptions specific to their academic skills was not taken into account in the work of Beidel and colleagues. Information processing or hybridized model In addition to the interference and deficit models relating test anxiety to academic impairments, an information processing model ha s received support in the literature (Benjamin, McKeachie, Lin, & Holinger, 1981; Naveh-Benjam in, 1991). Essentially a combination of the other two theories, this model pos its that test-anxious students experience detrimental cognitive interference at all stages of the learning and test-taking proc ess as well as academic or study skill
24 deficits (Bodas & Ollendick, 2005) Highlighting theoretical wea knesses of the deficits-model linking test anxiety to poor academ ic performance, Tobias (1985) poi nted out that if test anxiety stems from a students knowledge about their lack of mastery of test-rel evant material, anxietyreduction intervention strategies without academic -improvement components (i.e., exclusively aimed at reducing anxiety) should not result in lowered test anxiety. However, if cognitive interference represents the central element im pairing academic performance, interventions targeting worry should result in improved in academic performance (Tryon, 1980). Thus, poor performance in test-anxious stude nts is attributed to cognitive de ficits relating to organizing and encoding information (e.g., learning and studying de ficits), as well as difficulties in retrieving academic information during test-taking scenarios due to cognitive interference (NavehBenjamin, 1991). Applications of the information-processing model Based on work with university students in Ge rmany and consideration of the test anxiety literature over several decades, Naveh-Benjamin (1991) posited th at two types of highly testanxious students exist within th e framework of the information-pr ocessing model. Specifically, in individuals with good study sk ills and no deficits in enc oding and organizing information, reduced academic performance stemmed from probl ems in retrieval associated with interfering worries and doubts in their abilities ; in other words, the classic interference model is thought to indeed apply for some students (Naveh-B enjamin, 1991; Sarason, 1984; Tobias, 1985). Interventions focused on reducing an xiety (e.g., desensitization in th is study) in this type of student resulted in a reduction in test anxiety as well as improvement in academic functioning (Naveh-Benjamin, 1991). However, this work di d not include any measurement of emotional characteristics of the student apart from test anxiety and it is unknown whether these variables
25 may have explained the experienced interferen ce. Furthermore, information about students perceptions of their academic skills was not included in this work. A second type of highly te st-anxious student, according to Naveh-Benjamin (1991), is characterized by problems with en coding and organizing information; thus, for these individuals test anxiety begins with a more academic type of problem. When these individuals are trained in study skills, thus facilitating succes sful encoding of information dur ing the learning process, test anxiety is reduced and academic performa nce improves (Naveh-Benjamin, 1991). Again, the measurement here reflects actual demonstrated academic skills rath er than student perceptions or beliefs about their academic skills. Limitations of the Existing Models Although offering viable explanations of test anxiety and its relationship with reduced academ ic performance, this area of the lite rature did not take in to account emotional characteristics that may explain the interf erence that can impair academic functioning. Furthermore, these three models do not attempt to explain the processe s contributing to the nonacademic, emotional sequelae associat ed with test anxiety. In light of individual differences, it is considered unlikely that any one-dimensional theory would explain how test anxiety operates in all students. An information-processing or hybrid model, which posits that both academic deficits and cognitive interferen ce explain the academic deficits a ssociated with test anxiety, may represent the most plausible framework for unders tanding the academic side of test anxiety (e.g., Cassady & Johnson, 2002). The emotional side of test anxiety may be best understood by incorporating findings from the clin ical literature examining the re lationship of test anxiety with other anxiety disorders and general or tr ait-level anxiety (e.g. Zeidner, 1998). The aforementioned models relating test anxi ety to academic deficits, along with the literature on emotional and academic correlates of test anxiety, will inform the understanding of
26 mechanisms operating in the experience of test anxiety in adolescents Corresponding to the information-processing (i.e., hybrid) model linki ng test anxiety to detrimental academic outcomes, both academic variables and emotional va riables will be examined as they relate to test anxiety and may be applicab le at various points in specified models. The following section will review in greater detail the specific academic and emotional variables of interest that may shape the development and continue d experience of test anxiety. Interference and Deficits: Understanding Te st Anxiety w ith Academic and Emotional Variables Correlates and Predicto rs of Test Anxiety One of the challenges and limitations in te st anxiety research is determining which variables tend to precede versus which variables tend to follow or co-occur with test anxiety. Researchers have described the re lationship among test anxiety a nd its correlates as a negative spiral (e.g., McDonald, 2001) in which unsuccessful test-taking experiences result in increased anxiety followed by more negati ve test-taking experiences, maki ng it difficult to disentangle where the cycle originated. It remains unclear ho w this cycle or developmental pattern of test anxiety may differ across individual s, and it is likely that some inter-individual variations in developmental sequences may occur. The following section will discuss specific variables that have a demonstrated relationship with test anxiety in pr evious research. Emotional Variables Em otional considerations are relevant to understanding how test anxiety relates to academic impairments (e.g., Swanson & Howell, 1996) as well as understanding test anxiety in individuals without reduced academic performance (Saras on, Sarason, & Bierce, 1990; Zeidner, 1998). Thus, emotional variables pert ain to test anxiety in studen ts with or without academic impairments. In students showing evidence of aca demic difficulties emotional characteristics
27 may play a role in the experi ence of detrimental cognitive in terference (e.g., worry and doubt resulting in diversion of cognitive resources) (Cassady & Johnson, 2002). In students experiencing primarily social-emotional effect s of test anxiety without obvious academic impairment, emotional variables may reveal mo re pervasive forces operating to drive the experience of test anxiety (e .g., Beidel and Turner, 1988). Differences in the breadth and pervasiveness of anxious symptoms experienced may facilitate the understanding of co re predictors of test anxiety in individual students. Somebut not allstudents experiencing test anxiety may ha ve a clinically signifi cant anxiety disorder (e.g., Biedel & Turner, 1988) or an intrinsically and persistently anxious personality (e.g., Biedel et al., 1994; King, Mietz, Tinne y, & Ollendick, 1995). Howeve r, other individuals may experience anxiety solely related to test-taki ng and academic evaluation with minimal anxiety related to non-evaluative situations. Emotional variables can inform the understa nding of test anxiety. For example, when reflecting upon test results, st udents with high levels of test anxiety make more external attributions and report higher feelings of helplessness (Cassady, 2004). The following emotional characteristics and experiences have been linked to the experience of test anxiety. Psychopathology. Although not specified as its own diagnosis by the Diagnostic and Statistical Manual of Mental Disorders (4th ed., text rev. [ DSM-IV-TR ]; American Psychiatric Association, 2000) test anxiety is sometimes considered a high ly specific type or feature associated with the broader group of anxiety disorders (McDonald, 2001). Beidel and Turner (1988) found that from a sample of test-anxi ous students between the ages of 8 and 12, approximately 60% met diagnostic criteria for an anxiety disorder. Indeed, many school-aged children who experience test a nxiety may have symptoms of ot her disorders, the most common
28 of which are what was previously termed ove ranxious disorder (OAD) (now referred to as Generalized Anxiety Disorder; GAD) or social phobia (SP) (Beidel, 1991; King, et al., 1995). Compared to the non-test-anxious peers, adolescents with higher levels of test anxiety also reported greater amounts of anxiety on self-report instruments of general anxiety such as the Revised Childrens Manifest Anxiety Scal e (RCMAS; C.R. Reynolds & Richmond, 1978) and on the trait subscale of the State-Trait Anxiety Inventory for Children (STAIC; Spielberger, 1973) (King et al., 1995). In addition, high-test-anxious adolescents repor ted higher levels of depression on the Childrens Depression Inve ntory (CDI; Kovacs, 1983) (King et al., 1995). King and colleagues concluded that highly test-anxious students ofte n experience a global state of emotional distress or neg ative affectivity, (p. 52). Indeed, children with diagnosable anxiety di sorders such as GAD and SP are more likely to experience worries during evaluative situ ations (Beidel, 1991). Children who exhibit impairing levels of general anxiety may be especi ally sensitive to pressures in the environment such as those sometimes propagated in evaluativ e and testing situations (e.g., Zeidner, 1998). A significant limitation here is that much of the literature examining the overlap between test anxiety and other anxiety disorders fails to acco unt for academic variables that may operate in the development of test anxiety. Trait anxiety. Spielberger (1966) referred to the no tion of some students having a higher baseline anxiety in his distin ction between trait-level anxiet ythat is, anxiety that remains somewhat stable similar to a personality traitand anxiety that exists as a more transitory experience in response to a threatening s ituation, or state anxiety. Test anxiety is often conceptualized as a situation-sp ecific personality trait, or a di sposition to react with excessive anxiety in evaluative situati ons (Sarason, 1984; Spielberger, 1975; Stober, 2004). Thus, children
29 with a general inclination to exhibit anxious re actions to stressful stimuli are naturally more inclined to have an anxious reac tion in test-taking situations. In elementary school children, testanxious students had significantly higher levels of trait anxiety than thos e without test anxiety (Beidel, et al., 1994). Interestingl y, some data has linked slightly elevated trait anxiety with higher performance outcomes in adults, while elev ated test anxiety alone resulted in impaired performance (Fletcher, Lovatt, & Baldry, 1997). Thes e authors interpreted this data as indicating that modest increases in trait anxiety can facilitate higher perfor mance across areas of functioning (Fletcher et al., 1997) A limitation of these studies is the failure to consider variables such as initial skill s and abilities (e.g., academic ability and experiences) which may reveal a different pattern of experiences contributing to test anxiety. In additional studies of persistent anxious qualities, highly test a nxious students score significantly higher on the Psychasthenia scale on the Minnesota Multiphasic Personality Inventory-Version 2 (MMPI-2) th an do non-test-anxious students (Lufi et al., 2004). According to Graham (1990) individuals who score high on this scale, tend to be very anxious, tense, and agitated. They worry a great deal, even over very small problems, and they are fearful and apprehensive. High strung and jumpy, they report difficulties in concentr ating and often receive anxiety disorder diagnoses, (p. 74). Thus, researchers have concluded that a trait-like component of anxiety seems to exist within test anxiety (Lufi et al., 2004). Once again, however, studies that acknowledge the emotional variable s operating in the experience of test anxiety appear to omit the consideration of academic abilities and experiences. Although a general inclination towards anxi ety may explain test anxiety for some individuals, it does not account fo r test anxiety in individuals without high levels of trait anxiety
30 or specific forms of anxiety disorders. Other emotional characteristics merit consideration as individual differences that may reveal valuable insights about the experien ce of test anxiety. Perfectionism Perfectionism can serve an adaptive function that motivates students to excel academically or a maladaptive function that is associated with test anxiety and fear of not meeting standards (Blatt, 1995). For the purposes of this discussion, maladaptive perfectionism will be the focus, as it is more closely linked with test anxiety. In general, perfectionism is characterized by excessively high standards fo r ones own performance (Stober & Joorman, 2001). In addition, maladaptive perfectionism incl udes overly critical evaluations of ones own actions, worries about making mistakes, and doubt in actions taken (Fro st, Marten, Lahart, & Rosenblate, 1990). This conceptualization of perfectionism was found to have three major dimensions, including worry about mistakes and doubt in ones actions, parental expectations and criticism, and elevated pe rsonal standards (Stober, 1998). Not surprisingly, individuals higher in perfectionism also worry significan tly more, which is thought to explain the relationship between perfectionism and test anxiety (Stober & Joorman, 2001). When worry is controlled for, the relationships of anxiety to perfectionism and to procrastination approach zero (Stober & Joorman, 2001). An additional concep tualization of perfec tionism distinguishes between self-oriented perfectionism, which relate s to ones expectations for self-perfection versus socially-prescribe d perfectionism, which entails the belief that others require or expect perfection (e.g., Hewitt & Flett, 1991). Although in adults an xiety relates more strongly to socially-prescribed perfectionism, anxiety in children relates to both self-oriented and sociallyprescribed forms of perfectionism, sugges ting a more complex interplay among social expectations and a childs own high standards fo r performance (Hewitt, Caelian, Flett, Sherry,
31 Collins, Flynn, 2002). Indeed, parental perfectionism and school-wide pressures may play a role in shaping childrens beliefs about their own needs for achievement. Academic Variables W ith regard to the deficit component of the information-processing model of test anxiety, academic variables are most relevant. Severa l factors relating to a students academic performance play a role in th e experience of test anxiety. Achievement. Although some forms of test anxiety may be experienced by students of all achievement and intellectual levels, academic skills appear to play a role in test anxiety for many students. Specifically, students with higher academic skills tend to experience less test anxiety than students with lower abilities (Hembree, 1988). Likewise, test anxi ous students earn lower grades than their less anxious peers of compar able ability (King, Ollendick, & Gullone, 1992). Various indicators of student achievement have b een utilized in the te st anxiety literature. In high school and college student s, a negative relationship between test anxiety and grade point average (GPA) has been identified such that high er-achieving students expe rience lower levels of test anxiety (Hembree, 1988; Naveh-Benjamin, 1991). However, GPA does not appear to represent a pure measure of academic skills. Pi ntrich and De Groot (1990) found that in a sample of seventh-grade students, test anxiety negatively pred icted to overall grades and performance on important tests but not performan ce on measures of daily performance or written essays that are also reflected in GPA. In a sample of 4,000 undergraduate and 1,414 graduate students, a small inverse relati onship was identified between test anxiety and cumulative GPA (Chapell et al., 2005). In addition to grades in school, studies have examined the connection between test anxiety and performance on high-stakes tests. Scholastic Aptitude Test (SAT) scores have been found to negatively correlate with test anxiety in both males and females (Cassady & Johnson, 2002;
32 Spielberger, 1980). Additional high st akes tests have been examined as they are affected by test anxiety. Drawing from a large sample of 1,908 a dolescents in Australia, Matters and Burnett (2003) found that test anxi ety predicted students tendency to omit or not respond to test items on a high stakes test, therefore resulting in lower scores. Grades in school and high-stakes test performance represent frequently used indicators of academic performance examined in conjunction w ith test anxiety. However, no studies have differentiated students actual performance on daily academic tasks (e.g., GPA), standardized test performance (e.g., SAT or other important te st), and individual belie fs and perceptions of academic skills as they relate to test anxiety. St udent perceptions of their academic skills may be a particularly relevant consider ation in understanding how academic skills relate to a cognitive and emotional experience such as test anxiety (e.g., McDonald, 2001). Although no known studies examined students perceptions of their own skills specifically, previous researchers have identified student perceptions of test difficulty as playing a role in te st anxiety and academic performance (Hong, 1999). Although it is unclear whether deficient academ ic skills precede the experience of test anxiety (e.g., Paulman & Kennelly, 1 984) or if test anxiety actually results in deficient academic performance (e.g., Eynsenck, 1985; Sarason, 1980), a negative relationship between overall academic skills and test anxiety has been consisten tly identified in the literature (e.g., Cizek & Burg, 2006; Hembree, 1988; Zeidner, 1998). Intellectual ability. A negative relationship between general intelligence ( g) and test anxiety is supported in the literature. In a meta-analytic review of 135 studies, Ackerman and Heggestad (1997) found a significant negative correlation between g and test anxiety ( r = -.33). Although lower-ability students may indeed experience higher levels of test anxiety in general,
33 the picture becomes more complicated when a dditional factors are cons idered. For example, students of above-average intelle ctual ability can experience test anxiety as well (Zeidner & Schleyer, 1999; Preckel, Zeidner, Goetz, & Schleyer, 2008). A considerable limitation of the current literature is that st udies of academic and intellectual variables tend to omit the consideration of emotional vari ables such as trait anxiety. A study of self-concept and test anxiety in gift ed adolescents (ages 12 to 15) demonstrated that IQ alone does not predict test anxiety; rather, achievement and academic experiences may play a more important role (Van Boxtel & Monks, 1992). Specifically, underachieving gifted students who still performed academ ically in the average to above -average range but below the expected level based on superior IQ scores often exhibited poorer self-concepts and higher levels of test anxiety (Van Boxtel & Monks, 1992). Thus, neither intellectual ability nor academic performance may predict test anxiety, but the pe rception of a discrepancy between expected and actual achievement. Studies linking test anxiety to learning disabi lities (LD) support this notion (Swanson & Howell, 1996; Whitaker Sena, Lowe, & Lee, 2007). In addition to general intellectual abilities, a relationship between test anxiety and specific areas of cognitive ability may exist. For example, from a sample of college students identified with LD, those experiencing test anxiety had, on av erage, significantly lower Verbal IQ scores than their non-test-anxious LD peers (Lufi, Okasha, & Cohen, 2004). Total IQ scores were not significantly different for test-a nxious LD and non-test-anxious LD individuals. These authors posited that this significant rela tionship between test anxiety a nd reduced Verbal IQ could be explained by (a) the reduced academic performance a ssociated with test anxiety or (b) the results of the emotional difficulties associated with te st anxiety interfering specifically with verbal abilities (Lufi et al., 2004). However, as this particular study did not consider academic
34 performance apart from the LD diagnosis, the exte nt to which conclusions about the relationship between specific areas of cognitive ability and test anxiety can be drawn is limited. Intellectual ability does not uniquely explain test anxiety for students; therefore, academic performance is thought to represent a more useful variable as it relates to test anxiety. Overall, it appears that academically-related variables play a central role in many students experience of test anxiety. Educational experiences and classification. Related to ability level, a students past educational experiences and identification for Exceptional Student Education (ESE) services may relate to the experience of test anxiet y. In college students id entified with learning disabilities (LD), test anxiety was reported as of primary c oncern, followed by concentration, frustration, and remembering information. The role of learning difficulties and ESE status has also been examined in children and adolescent s, although not as extensively as in college populations. Children with learning disabilities and behavior disord ers experience higher levels of test anxiety than their t ypically-developing peers (e.g., Bryan, Sonnefeld, & Grabowski, 1983; Whitaker Sena, Lowe, & Lee, 2007). This higher leve l of test anxiety in high school students identified for ESE services has been explained by two major factors: more cognitive interference (e.g., task-irrelevant thoughts) and lower st udy skills (Swanson & Howell, 1996). In addition, this study also supported a negative relationship between scores on standardized tests such as SAT and test anxiety in student s with learning disabilities or behavior disorders (Swanson & Howell, 1996). Indeed, students with high levels of test anxiety we re consistently outperformed by those with lower levels of test anxiety (Cassady & Johnson, 2002). A respectable body of literature from educational research links test anxiety with academic deficits and underachievement. Indeed, students with a pattern of unsuccessful test-taking experiences may experience test anxiety in evaluative situations. However, as highlighted in the
35 emotional variables discussed above, this may not be the case for all in dividuals. Test anxiety has also been approached as a more psychosoc ial problem, with a body of studies stemming from a more clinical standpoint. Very little work has examined academic and emotional variables together as they relate to test anxiety, and this study is de signed to address this limitation. Integrating Academic and Emotiona l Contributors to Test Anxiety Given the wide body of em pirical evidence rega rding the symptom domains in which test anxiety symptoms can be manifested (e.g., academ ic, cognitive, emotional, behavioral), along with numerous studies of academic and emotional factors related to test anxiety, it naturally follows that individual differences would exist with regard to how symptoms are experienced across each of these dimensions (Stober, 2004). Furthermore, it is likely that differences in each of these domains also shape the development of test anxiety. McDona ld (2001) noted that, Although all evaluative situations will be accomp anied by some emotional reactions, it is the individuals past e xperiences and beliefs, that have been shaped by a multitude of factors, that will mould their unique reactions to a test situation, (p. 92). Based on two major theoretical stances in the test anxiety literat urethat is, test anxiety as an academic problem and test anxiety as an emotional problemprocesses related to both academic and emotional variables likely play a central role in the experience of test anxiety in individuals. Furthermore, students perceptions and beliefs about th eir academic skills may repres ent an important link between academic and emotional considerations. Understanding which academic and emotional variables may predict test anxiety in adolescents is a central goal of this study. By adapting the aforementioned information-processing or hybrid model and incorporating emotional variables, student perceptions of academic skills, and vari ous indicators of academic performance, a more thorough understanding of the expe rience of test anxiety can be obtained. The following section
36 will review how the academic and emotional variable s discussed here can operate in light of the predominant existing conceptualizations. The Academic Approach to Test Anxiety Much of the literature fo cuses on the academic variables operating in the experience of test anxiety (e.g., Goetz, Preckel, Pekrum, & Hall, 2007). As discussed prev iously, two traditional models explain the reduced academic performance associated with test anxiety: the deficit model, indicating that deficits in learning information precede poor performance and test anxiety, and the interference model, which explains reduc ed academic functioning as a result of cognitive distraction associated with worry and doubt durin g testing and preparing for the test (Cassady & Johnson, 2002; Cassady, 2004; Sarason, Sarason, & Pierce, 1996). The information-processing approach essentially synthesizes these two explanations (Naveh-Benjamin, 1991). Much theoretical and empirical support for the information-processi ng model exists (e.g., Bodas & Ollendick, 2005; Naveh-Benjami n, 1991; Zeidner, 1999). The current literature supports that some students test anxiety relate s to a history of academic struggles or skills deficits specific to studying and test-taking. Studies li nking test anxiety to a history of learning difficulties supports the role of academic variables in the proce sses surrounding test anxiety (Fisher et al., 1996; Swanson & Howell, 1996). In addition, studies linki ng test anxiety to deficient study skills (e.g., Paulman & Kennelly, 1984) and demonstrating a decrease in test anxiety following an academically-oriented intervention (Beidel et al., 1999) both support the deficit-relevant academic variables operating in the experience of test anxiety. Unsuccessful test-taking experiences and self-perceptions of poor academic skills may represent additional predictors of test anxiety in adolescent st udents. Supporting the interference component of the informationprocessing model linking test anxiety to impa ired academic functioning, test-anxious students
37 perform worse in high-pressure te sting situations desp ite having sufficiently learned information previously (e.g., Cassady & Johnson, 2002; Sarason et al., 1996). No known studies have integrated the info rmation-processing model with a more emotional or clinical approach to test anxi ety to examine how both emotional and academic mechanisms operate. Furthermore, the role of student perceptions of academic skills represents a key variable that may be particularly relevant to the integration of academic and emotional variables (McDonald, 2001). The extent to which academic and emotional variables may explain test anxiety in adolescents at various grade levels me rits further investigation. The Emotional Approach to Test Anxiety The literature supports that em otional characteristics operate in the experience of test anxiety. Support for the existence of test-anxiou s students in whom emotional variables play a central role stems from literature regarding the re lationship of test anxiety with anxiety disorders, perfectionism, and trait anxiety. Ma ny students with test anxiety m eet diagnostic criteria for an additional anxiety disorder (e .g., Beidel & Turner, 1988; Saras on, 1975). In addition, students with high levels of perfectionism worry more and are more test anxious (Stober & Joorman, 2001), and test-anxious students ha ve significantly higher trait an xiety than do non-test-anxious students (Beidel et al., 1994). Eina t (2000) posited that severe test anxiety is caused primarily by high personal standards of persons who exp ect maximum success and are afraid that they cannot meet their own standards, (p. 177; as c ited in Lufi et al., 2004) Indeed, it logically follows that students who exhibit hi gher levels of anxiety as a tr ait-like disposition are also more likely to experience test anxi ety (e.g., Sarason, 1984). However, some otherwise non-anxious students exhibit significant test anxiety specific to academic or test-taking situations, and the literature including emotional mechanisms driv ing test anxiety larg ely does not account for
38 academic variables. The present study integr ates the academic focus of the informationprocessing model with emotional ch aracteristics to better explain the operation of test anxiety. Existing Integrations of Academic a nd Emotional Factors and Test Anxiety Som e recent work has addressed the dearth of research on students emotions during the learning and testing processes that occur in schools. Drawing from a sample of sixth-grade students in Germany, consistent re lationships were identified between students ability levels and the specific emotional states ex perienced during testing. Students of higher abili ties experienced more positive emotions such as enjoyment, a nd students of lower abil ities experienced more anger and anxiety (Goetz et al ., 2007). Although studies of this sort explain the emotional outcome of test anxiety, once again they rely on an academically-oriented variable to explain the emotional experiences, without full y considering the role of emo tional tendencies that also can facilitate an understanding of how test anxiety operates. Another area in which the literature has inte grated emotional and academic factors in test anxiety is in considering student s ways of coping with different components of test anxiety (Stober, 2004). For example, university students a ppear to cope differently with worry (i.e., the cognitive component of test anxi ety) than with emotional symptoms of test anxiety (Stober, 2004). Once again, however, this type of resear ch does not provide information about any underlying factors that may explain th e components of test anxiety. Hypotheses and Research Questions Test anxiety represents a specific academic and mental health concern that is relevant to multiple domains of functioning in the lives of adolescents today. Despite its timeliness as a genuine problem plaguing students in schools, much of the literature on test anxiety is dated and lacks a comprehensive view of both the academic and emotional variables that may be operating.
39 By clarifying the predictors of test anxiety, intervention procedures can potentially be tailored to the most relevant variable s for adolescent students. Despite evidence that both emotional and academic factors may play a role in the experience of test anxiety, no known study examines the roles of both emotional and academic sources of test anxiety in youth Furthermore, beliefs and percepti ons of academic skills have not been examined as it relates to academic and emotional variables. Based on this review of the extant literature, it is hypothe sized that both academic performance and general anxiety play distinct yet interrelated roles in the expression of test anxiety in adol escent students. The guiding research questions for this study include: 1. What is the relationship of ge neral anxiety and academic indi cators with test anxiety? 2. To what extent do general anxiety, academic indicators, and demographic variables predict test anxiety in adolescents? 3. Beyond the established academic and emotional pr edictors of test anxiety, does student perception of academic skills contribute significantly to test anxiety? 4. What is the relationship of ge neral anxiety and academic indi cators with test anxiety in males and female adolescents? 5. How do general anxiety and indicators of acad emic performance predict test anxiety in male and female adolescents?
40 CHAPTER 2 METHOD Test anxiety represents a multifaceted experien ce that intersects the social-emotional and academic domains of functioning. Extending the ex isting hybrid model of test anxiety (Cassady & Johnson, 2002; Naveh-Benjamin, 1991) to incl ude both emotional and academic variables should improve our understanding of the processes operating in test-anxious students. Previous studies have supported the role of both academic and emotional variables in the operation of test anxiety (e.g., Beidel & Turner, 1988; Cassady & Johnson, 2002). However, how these key variables may act together to affect test anxi ety and each other had not been investigated. In addition, recent research examining the pervasiven ess of test anxiety has been lacking (Cizek & Berg, 2006; McDonald, 2001). Furthermore, a lthough both emotional and academic variables have been established as having a significant re lationship to test anxiety, how these variables may operate in the experience of test anxiety for male and female adolescent students has not been addressed. The purpose of the present stud y was to examine the relationship among test anxiety, emotional variables, and academic variables in adolescent students. The guiding research questions for this study were: 1. What is the relationship of ge neral anxiety and academic indi cators with test anxiety? 2. To what extent do general anxiety, academic indicators, and demographic variables predict test anxiety in adolescents? 3. Beyond the established academic and emotional pr edictors of test anxiety, does student perception of academic skills contribute significantly to test anxiety? 4. What is the relationship of ge neral anxiety and academic indi cators with test anxiety in males and female adolescents? 5. How do general anxiety and indicators of acad emic performance predict test anxiety in male and female adolescents?
41 This chapter will describe the procedures followed for participant recruitment and selection, data collection procedures, and the inst rumentation of relevant variables. In addition, descriptive information regarding sample characterist ics on variables of intere st will be provided. Participants Research Setting Participants were recruited from P.K. Y onge Developmental Research School (PKY), a unit in UFs College of Education that was es tablished in 1934 as a center of educational innovation for K-12 education. PKYs primary goal relates to the development, evaluation, and dissemination of exemplary education programs. It is a public school district that serves the state of Florida by conducting research on teaching and learning, often in collaboration with UF researchers. Instructional pract ices are investigated through fo rmal studies, faculty directed action research, and graduate stude nt research projects. The school's primary research goal is to enhance instruction in reading, mathematics, a nd science, in programs supported by state of the art educationa l technology. Based on data from the No Child Left Behind School Public Accountability Reports (20072008), 1162 students attended P.K. Yonge duri ng the 2007-2008 academic year. The student body was comprised of 49.7 percent females and 50.3 percent males, with a racial/ethnic distribution as follows: 52.9 percent White, 24.6 percent Bl ack, 14.9 percent Hispanic, 2.5 percent Asian, 0.4 percent American Indian, and 4.6 percent Multiracial (Flo rida Department of Education [FLDOE], 2008). According to the 2008 data, 9.6 percent of the students were classified with a disability and 22.3 percent of the student body was considered economically disadvantaged.
42 Sample: Recruitment and Demographics The total sample consisted of 104 students in grades seven through nine at PKY. Following approval from the University of Florida Institutional Review Bo ard (IRB) and school adm inistration, seventhand eighth-grade studen ts were recruited th rough required elective classes in Physical Education (P.E.) and Art. Ninth grade st udents were re cruited through required English classes. Specific class selectio n was based primarily on teacher cooperation and willingness to allow student participation. Parent s of students in grades seven through nine in participating classes were provi ded with basic information about the nature of the study and given the opportunity to provide consent for thei r child to participate in the study. Consent forms described the goals of the study, measures that we re administered, and potential contributions of the study. Students with parental consent were eligible for participation in this study. Recruitment and data collection took place over several weeks in Spring 2008. No immediate incentives were provided to participants in this study. Of the participating students, 53.8% were fe male (n=56). The sample was comprised of 38.5% seventh-grade students, 31.5% eighth-graders and 29.8% ninth-grad ers. A comparison of the race of participants in the sample to the ch aracteristics of the schoo lwide population revealed similar percentages of participants in each category: White (52.9 % schoolwide, versus 61.2% in this sample), Black (24.6 % schoolwide, versus 23.3% in this samp le), Hispanic (14.9% schoolwide, 12.6% in this sample); Asian (2.5 % schoolwide, 1.9% in this sample). Table 2-1 provides a more complete description of the to tal number and percentage of male and female participants for each grade level and race/ethnicity. Design & Procedure This study used a cross-sectional survey desi gn to investigate dem ogr aphic differences in test anxiety and the relationship of test anxiety with emotional and academic variables. Data was
43 obtained in two phases. In the fi rst phase of data collection, students provided self-report data on rating scale measures. In the second phase of data collection, archival data was obtained from the school regarding participants scores on the Fl orida Comprehensive Assessment Test (FCAT) and GPA. Data Collection Procedures Prior to data collection, a fi eld test of com piled measures was administered to three adolescents to check for general comprehensi on and feasibility of le ngth. These individuals reported that the questionnaires were an appropr iate length and strai ghtforward in meaning; therefore, feedback from initia l trials suggested no need to adapt the questionnaires. Data was collected during the spring semester of 2008. Either before or after school, or during school hours with parent and teacher pe rmission, students completed self-re port measures consisting of scales measuring test anxiety (Test Anxiety Inventory; TAI), general an xiety (Multidimensional Anxiety Scale for Chidlren; MASC), and academic competence (Academic Competence Evaluation Scales; ACES). Students completed m easures in small groups with the researcher. Group composition was based on class assignment and when consent forms were received. No more than six or seven same-grade students we re included in each group. Students were taken to a separate room (e.g., school library) and each pr ovided with a packet of the three measures included. The order in which the three measures were presented was counter-balanced with each possible order of measures equally represented among participants. Students were spaced apart to help ensure the privacy of their responses on rating scales. In addition, they were asked to not talk about their answers during the administratio n. Once each student provided assent to continue their participation in the study, directions were read aloud to the group. If an individual student had a question about the directions, the research er was available to clarify procedures for marking their answers. If a student was unsure about how to answer a particular item, he or she
44 was encouraged to respond with what most described him or her. Students were reminded to not leave items blank. When the three measures were completed, students were instructed to place them in a blank, unidentifiable envelope and retu rn the entire packet to the researcher. The majority of participants t ook approximately 25-30 minutes to complete all scales. At the time of survey completion, each students name was matched with an identification number. Upon entering informa tion into the database, students names and identifying information were removed, and data from each individual were associated only by the assigned identification number. Archival da ta regarding each participants academic performance (2008 year-end GPA) and standardi zed testing performance (2008 FCAT scores) were also obtained from the school with studen t information matched using identification code numbers. Specifically, the school wa s provided with a list of part icipating students matched with identification numbers. School-based profession als then provided GPA and FCAT scores for each student using only identification numbers such that no names were entered in the database. Only a master list with identification numbers linked each student with collected data, and this information was destroyed once all data was r eceived from the school and entered in the database. Inclusion and Exclusion Criteria Assenting students in the targeted grade leve ls and participating classes with parental consent were eligible to participate in this study. Data reflecting any obvi ous, invalid patterns of responses (e.g., answerin g each question the same or following an obvious pattern of answers) or missing data points were not utilized. However, this was not an issue given the small group setting in which measures were administered. If students left items blank on the rating-scale measures, for example, clarificat ion was provided and students were asked to provide their best approximation. Given these procedures, missing data was not an issue.
45 Instrumentation of Relevant Variables Due to the nature of the research questions, the f ollowing variables were of interest in understanding how test anxiety oper ates in adolescents. This section will provide an overview of each of the variables of inte rest and describe instrumentation in the present study. Test Anxiety The Test Anxiety Inventory (TA I) is a norm -referenced, self-re port measure of test anxiety developed by Spielberger and colleagues (1980). It represents perhaps the most widely used and extensively evaluated measure of test anxiety in the literatu re and the only norm-referenced measure that has been used specific to test anxiety (e.g., Cizek & Berg, 2006; Zeidner, 1999). The TAI provides subscales for both worry and emo tionality components of test anxiety, as well as a total test anxiety scale (S pielberger, 1980). Confirmatory fact or analyses have established the two-factor structur e (e.g., emotionality and worry) of the TAI for both males and females (Everson, Millsap, & Rodriguez, 1991). In this study, data analyses will use the total test anxiety score provided by the TAI. This instrument has b een described as a trait measure scale that reflects an individuals proneness to experience anxiety in situat ions specific to test-taking (Anastasi, 1988; Sapp, 1999). Based on responses on twenty questions using a Likert-type format (e.g., almost never, sometimes, often, almost always), the TAI provi des information about the frequency of test anxiety symptoms occurring before, during, and af ter tests (Spielberger 1980). Scores on the TAI range from 20 to 80 which represent the mi nimum and maximum experi ence of test anxiety based on reported frequency of symptoms (S pielberger, 1980). Regarding its psychometric soundness, Spielberger (1980) obtained validity coe fficients of .82 for males and .83 for females. Test-retest reliability coefficients ranged from .80 to .81 for 3-week and one-month intervals, respectively (Spielberger, 1980). Unlike the data provided from many test anxiety measures, raw
46 scores from the TAI can be converted into t-scores, allowing for normative comparison (Sapp, 1999); however, given the dated natu re of the available normative da ta, this study did not utilize provided t-scores. A review of the litera ture regarding test a nxiety in adolescent popul ations indicates that numerous studies, both assessmentand intervention-oriented, have utilized the TAI to measure test anxiety in high school a nd college students (Cizek & Be rg, 2006; Lufi & Cohen, 2004). One major application of the TAI has been in cr oss-cultural test anxiety research. Several international adaptationsincluding German (Hodapp, 1995), Hebrew (Zeidner, Nevo, & Lipschitz, 1988), Chinese (Rocklin & Ren-Min, 1983), and Hindi (Sharma, Sud, & Spielberger, 1983), among othershave been developed and u tilized to examine test anxiety across and within various populations (P eleg-Popko, 2004; Bodas & Ollendick, 2005). The TAI has also been utilized as a preand post-intervention meas ure in practical research. For example, Orbach, Lindsay, and Grey (2007) recently used the TAI as a major indicator of change in evaluating an internet-based cognitive-behavio ral therapy (CBT) intervention. These researchers found the TAI to be more sensitive to longitudinal changes in test anxiety than their own, recently-developed test anxiety measure (Orbach et al., 2007). The TAI has been th e assessment tool of choice in numerous doctoral dissertations as well. Overall, the TAI has been regarded by researchers as an acceptable measure of test anxiety in adolescent populations. In this study, the TAI was used as the criterion or dependent variable for multiple regression analyses. Regarding its technical adequacy in this sample the TAI appeared to represent an adequate form of measurement. Cronbachs alpha for the to tal TAI score (TAI Total) in this study was .928. For the worry and emotionality s ubscales, Cronbachs alpha was .848 and .857, respectively. Despite previous researchers finding that Worry and Emotionality represented
47 distinct subscales in older adolescents and university st udents (e.g., Everson, Millsap, & Rodriguez, 1991; Stober, 2004), in this exclusiv ely younger sample, these subscales did not offer a valid indication of separate subtypes of test anxiety. Given the developmental changes in adolescents self-awareness and possible limitations in their ability to differentiate cognitive and physiological symptoms of test anxiety, the emot ionality and worry subscale scores were not used in addressing the guiding research questions in this study. For participants in the presen t study, the mean raw score fo r total test anxiety was 38.74 (SD=12.29). This is consistent with the mean obt ained in previous research using the same instrument in adolescent and young adult stude nts (e.g., Putwain, 2007). As normative data for students below high school grades is not available, the percentage of participants reporting total test anxiety of more than two standard deviations above the lo cal mean was calculated. Three participants (2.8%) responde d with total test anxiety scores at or above this level (raw score greater than 63). General Anxiety The Multidim ensional Anxiety Scale for Childre n (MASC) is a 39-item self-report anxiety measure used for children and adolescents ages 8 through 19. An empirically -derived instrument, the MASC yields scores across four domains of anxiety: Physic al Symptoms, Harm Avoidance, Social Anxiety, and Separation Anxiety/Panic (March, Parker, Sullivan, Stallings, & Conners, 1997). The Social Anxiety domain consists of tw o subscales assessing Humiliation/Rejection and Performance Fears, as well as a Total Social A nxiety scale. Each item on the MASC is scored along a four-point, Likert-type scale (0=never, 1=almost never, 2=sometimes, 3=often) (March et al., 1997). The MASC has demonstrated excellen t internal reliability for all factors and subfactors in both males and females (March et al., 1997), as well as excellent test-retest reliability over a three-week period (March et al., 1999).
48 In a review of rating scales for internaliz ing problems, Myers & Winters referred to the MASC as a preferred anxiety ra ting scale in both clin ical and research se ttings, (2002; p.648). The MASC has been used with both school-based and clinical samples, with slightly higher reliabilities obtained for school-based samples (March et al., 1999). School-based intervention research has used the MASC as a screening measure for anxiety disorders and as a measure of post-treatment outcomes (e.g., Bernstein, Layne, Egan, & Tennisan, 2005). Compared to other established self-report an xiety rating scales, the MASC has de monstrated excellent divergent and convergent validity and is considered superior to the RCMAS (Dierker et al., 2001). Notably, the MASC was determined to have good utility in di scriminating symptoms of anxiety from other forms of psychopathology, namely depressive symptoms, in both school-based and pediatric psychiatric samples (Dierker et al., 2001; R ynn et al., 2006). The MASC represents a highly regarded screening instrument for multidimensi onal symptoms of anxiety. In this study, the MASC was of central importance to the research question regardi ng the extent to which general anxiety, grades, student ra tings of academic skills, and statew ide achievement test scores predict test anxiety. In addition, the MASC was considered for the research questions related to gender differences. For the sample included in this study, the MASC appeared to have solid technical adequacy overall. Regarding internal consis tency for all 39 items, Cronbachs alpha was .927. This is highly consistent with data in previous studies (March et al., 1997; Muris, Merckelbach, Ollendick, King, & Bogie, 2002). The mean total raw score on the MASC for participants in this study was 38.23. This is highly consistent with prev ious findings with a n on-clinical sample of students this age (Muris et al ., 2002). For general fo rms of anxiety indi cated on the MASC, 6.38 percent of male participants repo rted general anxiety in the clini cally significant range (at least
49 two SD s above the normative sample mean for ma les), and 10.63 percent reported general anxiety at least one SD above the normative mean. For females, 5.35 percent of participants had MASC scores within the clini cally significant range, or 2 SD s above the normative sample mean for females, and 8.92 percent fell at least 1 SD above the mean. In a variable with a normal distribution, approximately 95% of participants would typica lly report scores within two standard deviations of the mean. For both male s and females in this sample, slightly more individuals reported the highe st levels of anxiety. Perceived Academic Competence The Academ ic Competence Evaluation Scale (ACES) represents a body of rating scales designed to measure students academic competence based on teacher, student, or parent report (DiPerna & Elliott, 2000). Specifically, ACES provides information about Academic Skills (including Mathematical, Readi ng, and Critical Thinking ski lls) and Academic Enablers (including Interpersonal Skills, Motivation, Engagement, and Study Skills) using a 5-point, Likert-type scale (DiPerna & Elli ot, 2000). The term academic enab lers refers to the behaviors and attitudes that influence the extent to wh ich a student participates and benefits from instruction in the classroom (DiPerna & Elliott, 2000). The ACES yields a total score for academic skills as well as a total score for academic enablers. The scale provides qualitative interpretive guidelines correspondin g to specified ranges for both Sk ills and Enablers, including Developing, Competent, and Advanced. All scales and subscales of the ACES have established strong validity and reliability for elementary through high school students (DiPerna & Elliott, 1999; DiPerna, Volpe, & Elliot, 2005). In addition, the authors found high internal consistency coefficients ranging from .92 to .98. Adequate test-retest reliability was obtained (.81 to .92) for all scales. The ACES has been found to correlate moderately-to-highly with la rge-scale achievement tests and measures of
50 social skills in known groups of students with and without disabili ties (DiPerna & Elliott, 2000). Since a students perception of his or her academic competence is thought to play the most central role in shaping his or her beliefs and expectations about te st-taking, this study utilized the student self-report form of the ACES. The ACES Academic Skills Total Score and Academic Enablers Total Score were utilized as an academic indicator of central interest in the third research question. In the sample for this study total scor es for ACES Skills ranged from 85 to 150 ( M = 119.60). The majority of students (68.3%) repo rted overall academic skills within the Competent range based on the more conservative cutoffs provided for grades 6-8. Based on the qualitative interpretive guidelines provided on th is scale, 23.1% of the total sample reported having total Academic Skills with in the Advanced range. In addition, 8.6% of the total sample reported Academic Skills within th e Developing (o r lower) range. The ACES Enablers total scores ranged from 67 to 190 in the sample for this study ( M = 151.39). Based on the interpretive guidelines for the Enablers scale, 17.3% of the sample reported total Academic Enablers in the Advanc ed range (based on ranges for grades 6-8), 70% within the Competent range, and 12.5% within the Developing range. Additional Variables of Interest Demographic information. Gender represented a central va riable of interest in the guiding research questions. Information about stude nts grade level was relevant for obtaining an adequate sample size across grades seven to nine. Data from stude nt standardized test scores provided information regarding pa rticipants racial/ethnic backgr ound which were used to further describe the sample. Grades in school. Participants 2008 year-end grade point average (GPA), based on school records, was used as an indicator of st udents academic performance in school. This
51 number reflects students grades based on th e PKY established grading scale (A= 4.0, B=3.0, C=2.0, D=1.0). School policy allows the option of calculating weighted GPAsin which advanced level courses yield more points at each letter gradefor high school students only. Unweighted GPAs were used in this study in order to ensure a comparable grading system across middle and high school students. In order to protect students identities and maintain confidentiality of school reco rds, this information was pr ovided by school personnel to the researcher using coded identifica tion numbers that were not linked to each students identity in the database. Standardized Test Performance. In addition to students GPA, performance on the Spring 2008 Florida Comprehensive Achievement Test (FCAT) was obtai ned as an additional indicator of academic skills and performance. The FCAT was developed by school district curriculum content committees in conjunction with commercial contractors commissioned by the Florida Department of Educa tion (Florida Department of Education, [FDOE] 2008). It is administered annually to students in grades 3-11 statewide as part of the efforts to increase student achievement through higher standard s and accountability (FDOE, 2008). The FCAT includes a criterion-referenced portion assessing reading that yields a Sunshine State Standard (SSS) score ranging from 100 to 500, which translates to levels one (lowest score) through five (highest score) (FDOE, 2008). For the purpose of th is study, SSS Reading scores were utilized as indicators of students academic skills and test performance. As with GPA, this information was provided by school professionals us ing coded identificati on numbers to protect student identities and maintain the confidentia lity of school records. Many prior studies have used student perfor mance on statewide high-stakes achievement tests as a dependent variable in research (e.g., Martindale, Pearson, Curda, & Pilcher, 2005). Few
52 previous studies have utilized students scores on standardized, statewide achievement tests as a predictor variable for intern alizing problems (Ackerman, Izard, Kobak, Brown, & Smith, 2007). FCAT data represents an academic variable in th e correlations and regression analyses used to answer each of the research questions. The mean FCAT Reading Scaled Score for al l participants in the present sample was 330.10. For the entire state of Florida, the mean FCAT Reading Scaled Scores were 315, 310, and 313 for grades 7, 8, and 9, respectively (Flori da Department of Edu cation, 2008). Therefore, the FCAT reading performance in this sample was slightly higher than the state average.
53 Table 2-1. Sample Demographics: Number (and %) of Male and Female Participants for Each Grade and Race Gender Male Female Total Grade 7 22 (55.0%)18 (45.0%) 40 Grade 8 17 (51.5%)16 (48.5%) 33 Grade 9 9 (29.0%)22 (71.0%) 31 White 29 (44.6%)36 (55.4%) 65 Black 10 (41.7%)14 (58.3%) 24 Hispanic 8 (61.5%) 5 (38.5%) 13 Asian 1 (50.0%) 1 (50.0%) 2
54 CHAPTER 3 RESULTS The purpose of this study was to investigate th e relationship of test anxiety with academ ic performance and general anxiety in a sample of adolescent students. In addition, this study considered the relationship of test anxiety with academic variables and general anxiety in male and female students. Students self-ratings of individual academic skills represented an additional variable of interest. Chapter 3 will present the results of this st udy. First, key variables of interest will be examined for differences across demographic cate gories including gender, race, and grade level. With regard to the research questions and corr esponding hypotheses, data a nd relevant statistical analyses will be described. Correlational findings will be presented followed by the results of multiple regressions. A general exploratory fo rced-entry multiple regression was utilized to examine the role of several included predictor va riables. Next, a hierarchical regression was conducted to test a specific hypothe sis about the role of student perceptions of academic skills in predicting test anxiety. Finally, th e results of separate regression equations for males and females are provided. Descriptive Statistics across Demographics The total sample included 104 participants: 56 fem ales and 48 males. Means and standard deviations of all major variables for males, female s, and the total sample are listed in Table 3-1. An independent samples t-test reve aled no significant differences in total test anxiety (TAI Total) across males and females, t (102) = -.807, p = .417. Similarly, no significant gender differences emerged from t-tests comparing TAI Emotionali ty and TAI Worry, t (102) = -.149 and t (102) = -.928, respectively (one-tailed). One-way analyses of variance (ANOVAs) also indicated no
55 differences in total test anxiety (TAI Total) across grade levels, F (2, 101) =.011, p = .989, or race/ethnicity, F (3, 100) = .896, p = .446. Independent samples t-tests were also c onducted to examine the significance of gender differences across other major variables and subscales. For the Multidimensional Anxiety Scale for Children (MASC) total score, t-tests revealed no significant differences in mean total anxiety score for males and females, t (102) = -1.15, p = .25. Additional independent samples t-tests conducted for subscales from the MASC, incl uding Physical Symptoms, Harm Avoidance, Social Anxiety, Separation/Panic, and the Anxiet y Disorder Index, all re vealed no significant gender differences. Possible gender differences on the Academ ic Competence Evaluation Scales (ACES) Student Form (DiPerna & Elliot, 2005) were al so examined. No significant differences were revealed on either of the two major subscales, Academic Skills Total Score, t (102) = -.516, p = .607, or Academic Enablers Total Score, t (102) -.950, p = .344. Subscales comprising the Academic Skills score on the ACES, including st udents self-ratings of skills specific to Reading, Math, and Critical Thinking, also revealed no significan t gender differences. Subscales pertaining to Academic Enablers, including In terpersonal Skills, Engagement, Motivation, and Study Skills, were also checked for potential di fferences between males and females. Females reported significantly higher Interpersonal Skil ls (M = 42.50, SD = 6.11) than males (M = 39.95, SD = 5.69), t (102) = -2.167, p =.033. For GPA, t-tests revealed no significant gender differences, t (102) = -.720, p = .473. Males in this sample had slightly higher FC AT reading scaled scores (M = 337.29, SD = 35.32) than females (M = 323.20, SD = 34.20), approaching statistical significan ce, t (102) = 1.956, p =
56 .053. This is equivalent to a small-to-medium e ffect size (r =.190) (Field, 2005; Rosenthal, 1991). Inferential Statistics: Correlatio ns and Multiple Regress ions Prior to conducting additional analyses, dist ributions of scores on each measure were examined for normality, skewness, kurtosis, and homogeneity of vari ances. Distributions of most variables approximated a normal di stribution. To address modest vi olations in assumptions, data transformations were conducted. For the MASC Tota l Score, a square root transformation best approximated normality. For the TAI Total Score, a natural log transformation was used to improve normality and reduce kurtosis and skewness prior to regression analyses. Throughout this chapter, data pertaining to MASC scores will reflect the s quare root transformation, and TAI Total score will reflect the natural log transformation. For all regression equations, re sidual plots were examined for patterns indicative of violations of homoscedasticity, linearity, and normality. No obvious shapes or patterns were visible upon close examination of residual plots. In addition, statistics and diagnostics checked for collinearity (Field, 2005). Variance influation factor (VIF) and tole rance statistics were calculated to check for problems regarding multicollinearity. For all regression analyses presented in this chapter, VI F values remained around 1.0 while tolerance values ranged from 0.653 to .964. Therefore, assumptions regarding no multicollinearity in the model were met (e.g., Myers, 1990). Research Questions The following research questions will gui de the presentation of results. Research question 1 What is the relationship of general an xiety (MASC Total Score) and academic indicators ( GPA, FCAT, and ACES Skills & Enablers), with test anxiety (TAI)? Based on
57 previous studies, scores reflecting academic performance and engagement including GPA, FCAT reading score, and ACES skills are expected to relate negatively to test anxiety. It was hypothesized that general anxiety would co rrelate positively with test anxiety. The relationships between Test Anxiety, MASC total score, GPA, FCAT Reading Scale Score, ACES Skills, and ACES Enablers were ca lculated using Pearson correlations. Table 3-2 presents correlations among variables of interest base d on the entire sample. Significant positive correlations were found betw een a number of variables. The strongest correlations were found be tween total scores on the TAI and the MASC ( r =.451, p <.01). In addition, the TAI total score had significant negative correlat ions with academic indicators including ACES Skills ( r = -.273, p <.01), FCAT Score (r = -.390, p <.01), and GPA ( r = -.264, p <.01). ACES Enablers scores were not signi ficantly correlated w ith TAI total scores. Research question 2 To w hat extent do general anxiety (MASC Total Score), academic indicators (GPA, FCAT, and ACES Skills & Enablers), and dem ographic variables (race, gender) predict test anxiety (TAI) in adolescents? A series of multiple regression analyses were conducted to examine the predictive utility of academic perf ormance variables along with general anxiety while statistically controlling fo r demographic differences (race and gender) and the relationship among variables. Due to the narrow range of grade levels included in the sample and the insignificant differences in test a nxiety across grade levels, this va riable was not included in the regression analyses. Race was included in the regression equation using dummy-coded variables (e.g., Field, 2005). An initial multiple linear regression analysis using forced-entry procedures examined the relationship of GPA, MASC, ACES Skills, AC ES Enablers, FCAT, gender, and race as predictors of test anxiety. A simultaneous regression was selected due to the exploratory nature
58 of including all of these variables jointly (Field, 2005). This m odel indicated that these predictor variables jointly accounted for 41.2% of the variance in test anxiety, R2 = .412, F (8, 95) = 8.313, p <.000. Table 3-3 presents regression coeffi cients, standard error, standardized weights, and tvalues and significance for each variable. Based on these initial regression analyses, FCAT and MASC scores were determined to be significant predictors in the equation when c ontrolling for GPA, ACES Skills, ACES Enablers, gender, and race. Therefore, other predicto rs included in the regression model did not additionally contribute to the variance in test anxiety. Subsequent hierarchical regres sion analyses built upon these tw o key variables to test an additional hypotheses based on the literatu re (e.g., McDonald, 2001) and theory. Research question 3 Beyond the established academic and emotion al predictors of test anxiety, does student perception of academic skills explain additional vari ance in test anxiety? It was hypothesized that beyond general an xiety and academic performance, students perceptions and beliefs about their academic skills would account fo r an additional portion of the variance in test anxiety. The initial simultaneous regression analyses used a for ced-entry procedure given the exploratory nature of including all of these variables jointly in th e prediction of te st anxiety. In order to test this specific hypot hesis about the importance of st udent perceptions of skills beyond actual test performance and general anxiety, a hi erarchical procedure was utilized. Given the hypothesis that the model would improve by in cluding student ratings of academic skills, a hierarchical multiple regression procedure was utili zed to examine the effect of this additional predictor in the model (Petrocelli, 2003).
59 A two-step hierarchical regr ession was run with MASC (squa re root transformation of MASC Total score) and FCAT Reading Scale Scor e forced into step 1 and ACES Skills entered at step 2. Previous analyses indicated that MA SC and FCAT scores accounted for the largest amount of variance in total test anxiety; therefore, these were entered into the first step (Field, 2005; Petrocelli, 2003). ACES Skills to tal was selected for Step 2. Table 3-4 presents the coefficients and standardized beta weights for hierarchical regression analyses. At the fi rst step of the regression equation, MASC and FCAT jointly accounted for 37.7% of the variance in total test anxiety. At step two, when ACES Skills entered the model, 41.1% of the variance in test anxi ety was accounted for. This was a significant improvement in the model ( R2= .411; R2= .035 F p =.017). Research question 4 What is the relationship of general anxiety an d academic indicators with total test anxiety in male adolescents and in female adolescents? It was hypothesized that general anxiety (MASC) would positively correlate to te st anxiety in both males and females. In addition, it was hypothesized that academic indicators (GPA, FCAT scores, ACES skills, ACES enablers) would negatively correlate to test a nxiety in males and females. To examine possible differences in the relationship of academic and em otional variables with test anxiety and its two components in males versus females, separate Pe arson correlations of va riables were calculated for males and females. Table 3-5 provides correlations between major variables and level of significance for both males and females separately. As expected, general anxiety (MASC total score) positively correlated to total test anxiety (TAI total) in both males and females. In male s, total test anxiety significantly negatively correlated to self-ratings of academ ic skill, academic enablers, and GPA. In females, total test
60 anxiety did not significantly correlate with self-ratings of academic skills, academic enablers, or GPA. FCAT Reading Scale had a significant negative relationship w ith total test anxiety in both males and females. Research question 5 How do academic indicators and general an xiety predict test anxiety in male and female adolescents? It was hypothesized that general anxiety and FCAT score would differentially predict test anxiet y in male and female adolescents Using the data-splitting feature of SPSS to separate the sample by gender, MASC and FCAT Reading Scaled Score were entered into a regression equation using forced-entry pr ocedure. For males, this model accounted for 37.8 percent of the variance in test anxiety; for fe males, it explained 36.7 percent of the variance in test anxiety. Table 3-6 presents coefficients and standardized beta weights for MASC and FCAT scores in both males and females. When comparing two populations for the same variable, unstandardized coefficients should be examined to best represent differences across groups (Algina, 2005). This comparison is possible b ecause the units are the same. Comparing the slopes for males versus females, the slope fo r FCAT did not change; however, the slope for MASC scores was slightly higher for females. Exam ining the standardized beta values within the model for males, FCAT Reading Scaled Score ap peared to have a sli ghtly higher predictive value for test anxiety than MASC score. Examin ing standardized beta values within the model for females, general anxiety had a slightly heavier weight in predic ting total test anxiety than did FCAT scores. Follow-up Analyses Although previous analyses indicated that test anxiety is significan tly correlated with general anxiety as reflected by the MASC total score, how subscales a nd subtypes of anxiety reflected on the MASC m ay relate to test anxiety was unknown. Previous research has identified
61 test anxiety as often associated with Generalized Anxiety Di sorder (GAD) and social phobia (SP) in children (Beidel, 1991; King, et al., 1995). Therefore, follow-up analyses were conducted to examine the relationship of subscales of the MA SC with test anxiety in this adolescent sample (see Table 3-7). Subscales specific to physical sy mptoms and social anxiety correlated most strongly with test anxiety.
62 Table 3-1. Descriptive statistic s of measures for males, females, and total sample Variable Males M SD Females M SD Total Sample M SD TAITotal 37.67 (13.21) 39.64 (11.48) 38.73 (12.29) TAIWorr y 13.87 (5.62) 15.85 (5.07) 15.40 (5.33) TAIEmotionalit y 15.35 (6.08) 15.51 (4.94) 15.44 (5.47) MASC 36.02 (18.49) 40.12 (17.80) 38.23 (18.15) ACESSkills 118.72 (15.79) 120.34 (15.84) 119.60 (15.76) ACES-Enablers 149.11 (22.52) 153.30 (22.16) 151.39 (22.31) GPA-Unweighted 3.05 (.63) 3.13 (.62) 3.098 (.62) FCAT-ReadScaled 337.29 (35.32) 323.93 (34.20) 330.10 (35.20) Table 3-2. Matrix of correlations for entire sample TAIT TAW TAE MASC ACESS ACESE FCAT GPA TAIT 1.00 TAW .934** 1.00 TAE .926** .798** 1.00 MASC .451** .428** .447** 1.00 ACESS -.273** -.289** -.159 -.069 1.00 ACESE -.150 -.202* -.062 -.145 .620** 1.00 FCAT -.390** -.414** -.326** -.025 .282** -.013 1.00 GPA -.264** -.305** -.188* -.060 .343** .269** .521** 1.00 *indicates significant at the .05 level (1-tailed) ** indicates significant at the .01 level (1-tailed) Table 3-3. Forced-entry regres sion of test anxiety regres sed onto predictor variables Predictor SE tp (Constant) 4.611 .357 12.928 .000 MASC** .097 .018.4425.509 .000 GPA -.009 .053-.018-.176 .860 FCATRS** -.003 .001-.358-3.376 .001 ACESskills -.004 .002-.196-1.805 .074 ACESenablers .000 .002.010.112 .911 Gender -.010 .054-.015-.130 .897 Race Dummy1a -.017 .073-.023-.240 .811 Dummy2 b .003 .081.003.032 .974 R2 =. 412, Adj. R2 = .362; F (8, 95) = 8.313, p < .0001. ** p < .001 aDummy Coded Variable fo r White versus Black; bDummy-coded variable for Caucasian versus Hispanic
63 Table 3-4. Hierarchical regression B SEBt p Step 1 (Constant) 4.266 .265 --16.113 .000 MASC Total .099 .017.4515.746 .000 FCAT RS -.004 .001-.410-5.219 .000 Step 2 (Constant) 4.593 .292 -15.742 .000 MASC Total .097 .017.4405.728 .000 FCAT RS -.003 .001-.356-4.456 .000 ACES Skills -.004 .002-.194-2.422 .017 Step 1: R2 = .377, Adj. R2=.364, p < .001; Step 2: R2 = .035, R2 = .411 Adj. R2=.393, F p = .017.
64 Table 3-5. Matrix of Pearson correlations among major variables for males, females TAIT TAW TAEMASC ACESSACESE FCAT GPA TAIT Male Female 1.00 TAW Male Female .953** .913** 1.00 TAE Male Female .926** .907** .852** .742** 1.00 MASC Male Female .451** .459** .474** .372** .435** .467** 1.00 ACESS Male Female -.350** -.209 -.341** -.254* -.267* -.051 -.102 -.052 1.00 ACESE Male Female -.247* -.071 -.303* -.125 -.132 .008 -.255* -.071 .576** .655** 1.00 FCAT Male Female -.434** -.332** -.445** -.367** -.406** -.249* -.144 .126 .302* .293* .057 -.041 1.00 8. GPA Male Female -.353** -.190 -.385** -.246* -.326* -.046 -.201 .051 .124 .525** .247* .280* .594** .504** 1.00 *indicates significant at the .05 level (1-tailed) ** indicates significant at th e .01 level (1-tailed) Table 3-6. Forced-entry regres sion model for males, females BSEB t P Male (Constant)** MASC* FCAT** 4.445 .092 -.004 .445 .027 .001 Female (Constant)** MASC** FCAT** 4.161 .107 -.004 .334 .023 .001 For Males, R2 = .378, Adj.R2 = .350, F =13.652, p < .001; For Females, R2 = .367, Adj.R2 = .343 for Females; F = 15.380, p<.001. p < .005, ** p < .001
65 Table 3-7. Correlations among MASC subscales and TAI TAIT MASCT PhysSa HarmAvd b SocAnxc SepAnx d ADIe TAIT 1.00 MASCT .451** 1.00 PhysSa .494** .880** 1.00 HarmAvd b .110 .635** .338** 1.00 SocAnxc .410** .877** .758**.366**1.00 SepAnx d .400** .801** .610**.465**.599**1.00 ADIe .443** .926** .797**.589**.838**.731** 1.00 *indicates significant at the .05 level (1-tailed) ** indicates significant at the .01 level (1-tailed); aPhysical Symptoms; bHarm Avoidance; cSocial Anxiety; dSeparation Anxiety; eAnxiety Disorder Index.
66 CHAPTER 4 DISCUSSION A m ultitude of findings from several decades of research have supported the role of academic variables and emotional variables as important correlates of test anxiety. Using a multifaceted conceptualization of test anxiety, this study integrated academic and emotional understandings of test anxiety in an adolescent sample. In the academic-related research pertaining to test anxiety, three major models have attempted to explain the connection between test anxiety and impairments in academic achievement (Bodas & Ollendick, 2005): the cognitive interference model, skill-deficits model, and an information-processing model, which re presents an integration of the others. The cognitive interference model proposes that the diversion of cognitive resources from the academic tasks (e.g., worrying and fears about test outcomes) interferes with encoding and retrieval during the learning and test-taking process (Morris, Davis & Hutchings, 1981; Sarason & Stoops, 1978; Wine, 1971). Within this particular framework, the worry component of test anxietybut not the emotionality componenthas been linked to detrimental academic achievement in both adolescents (Williams, 1991) and college students (B andalos, Yates, & Thorndike-Christ, 1995; Cassady & Johnson, 2002). The skill-deficits model proposes that the connection between test-anxiety an d academic struggles stems from underlying academic deficits and poor studying and test-taking skills (Beidel, Turner, & Ta ylor-Ferreira, 1999; Culler & Hollahan, 1980; Ergene, 2003). Finally, the inform ation-processing or h ybridized model posits that students with test anxiety experience both cognitive interference and academic or study skill deficits (Naveh-Benjamin, 1991). A major limitation of existing academically-ori ented models is the omission of emotional variables such as general or trai t-level anxiety that also may repr esent an influential variable and
67 an associated outcome. Indeed, emotional vari ables are relevant to understanding how test anxiety relates to academic impairments (e.g., Swanson & Howell, 1996) and the experience of test anxiety in students not exhibiting reduced academic performance (Sarason et al., 1990; Zeidner, 1998). Based on the lite rature, higher levels of test an xiety relate to more general anxiety as reported on general measures of anxiety such as the Revised Childrens Manifest Anxiety Scale (RCMAS; C.R. Reynolds & Rich mond, 1978). Students perceptions and beliefs about ones academic skills were identif ied as a variable of particular interest as it related to both academic and emotional considerations. This study examined the relationship among acad emic indicators and general anxiety with test anxiety. Unlike previous research, this study integrated academic and emotional understandings of test anxiety wh ile also including other potentially relevant variables. Another feature of the present study was consideration of students self-ratings of academic skills and academic enabling behaviors. In addition, demogr aphic variables including race and gender were examined. The following text will summarize majo r findings, as well as interpret and explain results related to the gu iding research questions. In additi on, how findings relate to previous research, limitations, and future directions will be discussed. Demographics and Descriptive Findings Gender In this study, m ales and females reported si milar levels of overall test anxiety and emotionality and worry components of test anxiet y on the Test Anxiety Inventory (TAI). Based on the literature, females were expected to repo rt more test anxiety th an males (e.g., Hembree, 1988; Stober, 2004). Males and females also reported similar levels of general anxiety. Subtypes of anxiety included in subscales of the Multidimensional Anxiety Scale for Children (MASC) also revealed no gender differences. This was une xpected given the large body of literature that
68 has documented that females exhibit higher levels of anxiety across all subtypes of anxiety (e.g., Somers, Goldner, Waraich, & Hsu, 2006). Despite th is well-documented gender difference, these authors noted that the magnitude of the obser ved gender difference varies quite a bit across studies (Somers et al., 2006). Furt hermore, van Gastel and Ferdin and (2008) recently indicated that although the MASC offers a valid indication of anxiety sy mptoms, it does not adequately screen for specific diagnosable anxiety disorders. Had this study used a clinical interview to obtain more detailed information about symptoms of specific anxiety disord er diagnostic criteria, perhaps gender differences would have emerged. Al so, as discussed in earl ier chapters, slightly more students than expected repor ted clinically significant levels of general anxiety as reflected by a total anxiety score more than two standard deviations above the normative sample mean. Thus, this sample may have included slightly more anxious youth than expected for both genders. Regarding males and females self-ratings of academic skills and academic enabling behaviors, as measured on the Academic Comp etence Evaluation Scales (ACES), no differences were found on the two major scales (Skills and Enablers). The only subscale within the Enablers scale that reflected gender differences was the Interpersonal Skills scale, on which females reported higher levels. Specific items on this subs cale relate to getting along well with teachers and peers, following rules, cooperating, and communi cating effectively. It is unclear whether this reflects a true gender difference in this sample or simply a reporting difference in which females were more inclined to report socially-pleasing behaviors. Elliot, Di Perna, Mroch, and Lane (2004) did not find statistically significant gende r differences pertaining to overall academic enabling behaviors based on student self-reports; however, a medium effect size was reported for the Interpersonal Skills subscal e with females reporting higher levels. In addition, teacher
69 ratings on the teacher version of the ACES sugg ested that female adolescents exhibit higher enabling behaviors overall, includ ing interpersonal skills (Ellio t et al., 2004). Considering the convergent support from both teacher-ratings and st udent self-ratings in previous research, the present findings may reflect more pervasive differences in the demonstration of interpersonal skills in female and male adolescents. Grade level Total test anxiety was found to be sim ilar for students across the three grade levels included in the study. However, as this study dr ew from a narrow range of grade levels (7-9) with a limited number of participants in each grade, the nonexistence of grade level differences should be interpreted with caution. No known studi es have examined anxiety specific to testtaking developmentally by making direct co mparisons across different age groups. Race/Ethnicity No differences in students self-report of total test anxiety were indicated across racial/ethnic groups. Results of a previous m eta -analyses indicated that African American and Hispanic students exhibit higher levels of test anxiety during elementary school but that these differences were no longer evident in highschool students (Hembree, 1988). Perhaps any existing differences in test anxiety associated with race have dissipated by the time students reach 7th, 8th, and 9th grade. Given that Hembrees research analyses were conducted over 20 years ago, it is possible that his findings are no longer applicable due to changing demographics, beliefs, and practices in the United States. Another consideration with regard to obtained si milar levels of test anxiety across race and ethnic groups is the processes of stereotype threat as they may serve to activate the experience of test anxiety at specific times. Ma rx and Stapel (2006) suggested th at stereotype threat increases anxiety specifically before taking a major test during the exact time the stereotype threat is
70 activated. After tests, students affected by stereo type threat reported highe r levels of frustration but not anxiety (Marx & Stapel, 2006). Therefore, st udents self-report of te st anxiety at a time when a specific stereotype threat is not activated would not necessa rily reflect the level of test anxiety experienced just prior to the test, for example. Correlates of Test Anxiety Academic Indicators Consisten t with previous research, test anxiety negativ ely correlated with various indicators of academic performance. Grade point average (GPA), Florida Comprehensive Achievement Test (FCAT) (Read ing Scaled Score), and studen t self-ratings of academic proficiencies (ACES Skills) negative ly correlated with total test anxiety. Student ratings about their academic enabling behaviors (ACES Enablers ) did not significantly relate to total test anxiety A large literature base supports the inverse relationship between academic performance and test anxiety (Hembree, 1988; Cassady & Johnson, 2002); however the strength of association with various types of academic indicators and test anxiety has not been examined within the same study. Performance on a high-stakes achievement test (FCAT Reading Scaled Score) appears to have a slightly stronger relationship with test anxiety than both GPA and selfratings of academic skills. One po ssible explanation is th at test anxiety is particularly salient during high-stakes academic achievement test s such as the FCAT, whereas GPA reflects performance over multiple occasions that may not be as anxiety-provoking as a major test that contributes to educational decisi ons. Another consideration is that performance on standardized, state-wide achievement tests could in some re present a more pure measure of academic skills in a high-stakes context while GPA may reflect differences in course or teacher demands, effort and engagement, and other intra-individual va riables (e.g., conscienti ousness, organizational skills, task completion) in addition to basic a cademic skills. Furthermore, several items on the
71 test anxiety inventory (TAI) refe r to important tests, and important exams, (Spielberger, 1980). The FCAT represents one such important test that may be pa rticularly salient to students in the state of Florida; therefore, it is possible that the wording of this measure led students to more automatically cognitively link this measure to the FCAT as they were completing the questionnaire. In general, items on the TAI do not assess academic anxiety in general, but specifically refer to test-taking situations. Relationship among Various Academic Indicators This study found that although student ratings of their own pro-academ ic behaviors (ACES Enablers) related to GPA, there was no apparent connection between self -reports of academic enablers and FCAT scores. Conceptually, acad emic enabling behaviors (ACES Enablers) such as study skills, engagement, motivation, and inte rpersonal skills logically could influence GPA but are not necessarily as pivotal in standardized achievemen t test scores. On the other hand, given the correlational nature of this study, the opposite may be true: students who tend to receive high grades in school could internalize academic feed back and develop beliefs about themselves as a good student. Finally, the validity of student ratings on the ACES could potentially be limited by social desirability bi ases or potentially rush ed responding on a long (68item) questionnaire. Gender differences were also observed with regard to the relati onship between various academic indicators. For females, GPA strongly re lated to self-ratings of academic skills (ACES Skills); however, no significant connection betw een grades in school and self-ratings of academic skills was obtained for male students. In this study, grades in school positively related to self-reports of academic enab ling behaviors (ACES Enablers) in both males and females. No known previous studies have examined this t ype of gender pattern in adolescent students; however, evidence from a sample of college studen ts has suggested that academic enablers may
72 relate more strongly to current GPA while academ ic skills may relate more to cumulative GPA (DiPerna, 2004). Perhaps adolescent males are more inclined to cogni tively separate their classroom performance from their actual abilitie s (e.g., Just because I d ont get good grades in English does not mean I cant read and write effectively,). To examine this possible explanation, partial correlations were calculated controlling for academic enabling behaviors (ACES Enablers). Although the strength of relationship between GPA and ACES Skills decreased slightly for both males and females when partialing out the vari ance associated with ACES Enablers, no changes in stat istical significance were obtained for either gender. The role of various academic indicatorsand the relation ship among themrepresents an important topic for educational research. Test Anxiety and Other Forms of Anxiety In addition to academ ic correlates, general anxi ety positively correlated with test anxiety. Although previous studies of test anxiety in youth have not used the Multidimensional Anxiety Scale for Children (MASC), other researchers report ed a direct relationship between test anxiety and broader forms of anxiety as measured on th e Revised Childrens Ma nifest Anxiety Scale (RCMAS; Reynolds & Richmond 1978) and the St ate-Trait Anxiety Inventory for Children (STAIC; Spielberger, 1973) (King et al., 1995). A consideration and potential limitation with regard to this association between general anxiety and test anxiety is that both of these measures relied exclusively on self-report. Therefore, part of the strong correlati on between test anxiety and general anxiety could partly reflect the shared variance associ ated with using similar selfreport rating scales with parallel Likert-type formats. For academic variables, external indicators (FCAT and GPA) were used in addition to self-report measures (ACES). Follow-up analyses explored the relationship among test anxiety and its components with subscales of the MASC. Total test anxiety re lated most to the scales measuring physical
73 symptoms, followed by social anxiety. This is consistent with Beidel and colleagues (1988) suggestion that test anxiety represen ts a subtype of social anxiety. Test Anxiety and Gender General anxiety and FC AT scores correlated with total test anxiety in both males and females. In female adolescents, total test anxi ety was not significantly related to GPA or selfratings of academic skills and enablers. In male s, however, higher levels of test anxiety were associated with lower GPA, lower FCAT, and lower self-ratings of academic skills and enablers. Although it is possible that test anxiety in males actually has a stronger relationship with academic indicators than in females, Pearson corre lations also reflect variation around the mean and measurement error in addition to valid re lationships between variables (Algina, 2005). Therefore, gender differences in observed correla tions should be interp reted with caution. The finding that total test anxiety negatively correlated with all additional academic indicators (ACES Skills, ACES Enablers, GPA) in male students but not in females may have interesting implications regarding gender differences in academic functioning and test anxiety in adolescents. Considering the perv asiveness of negative correlations of test anxiety with various academic indicators, one possible explanation is that deficient academic skills truly are more important in explaining test anxiety specifically in males, such as proposed within the traditional skill-deficit model of test anxiety (Culler & Hollahan, 1980) Given the correlational nature of the findings, these data may also lend support to th e idea that test anxiety could impair academic performance on tests and daily ev aluative tasks, as proposed by the interference model (Sarason, 1984). Furthermore, the information-processing or hybridized model also fits the findings for males, given the limits of correlational data (Naveh-Benjamin, 1991). However, none of these existing models linking test anxi ety to reduced academic performance consider the differences among daily academic performance (e.g., GPA), performance in a major test or exam, and
74 individual perceptions of academic skills. This w ill be discussed further in the next section of this text regarding predic tors of test anxiety. For females, the only academic indicator si gnificantly related to test anxiety was performance on a high-stakes statewide achievement test. Interpreted from the perspective of the skill-deficits model of test anxietywhich holds that test anxiety relates to deficient study and test-taking skillsthis could suggest that for gi rls, GPA does not represent a pure indication of underlying academic skills but in fact reflects several enabling behaviors such as motivation, study skills, and academic engagement. Interestin gly, females self-repor t of academic skills strongly related to GPA. This c ould suggest that girls internali ze GPA as a reflection of their academic skills. Perhaps test anxiety does not affect the daily academic engagement and performance reflected to GPA in female adolescents. An additional consideration is that perhaps female adolescents do exhibit more academic enabling behaviors regardless of te st anxiety, which could also account for the lack of significant correlation between GPA and test an xiety in adolescent girls. Ind eed, Elliot et al. (2004) found that teacher-ratings of academic enabling behavi ors were higher for female adolescents than male adolescents. Again, measurement error a nd slightly unequal variance observed between males and females particularly on the TAI may acc ount for this apparent gender variation in the relationship of test anxiety with academic variables. Predicting Test Anxiety No known studies have investig ated the predic tive utility of academic variables along with general anxiety and demographics as they relate to test a nxiety. Accounting for GPA, ACES Skills and Enablers, gender, and race, a major finding of this study was that two significant predictors of test anxiety we re general anxiety (MASC Total Scores) and performance on statewide achievement testing (FCAT Score). When entered jointly into the model (forced entry
75 procedures), none of the other variables made significant contributions. However, when variables were entered into the model using a hier archical procedure, ACES Skills emerged as an additional significant predictor. This finding supports the need for the pr oposed expansion of existing InformationProcessing Model linking test anxiet y to academic impairments in or der to jointly account for the influence of general anxiety and experiences of adolescent studentsnamely, those related to broader forms of anxiety. Furthermore, it supports that solely viewing test anxiety as a subtype of anxiety without consideration of academic sk ills represents too narrow of an understanding. In addition to lower academic skills that relate to test anxiety, self-reported general anxiety also plays an important role in the experience of test anxiety, even when controlling for gender and other academic indicators. This basic model offe rs a parsimonious and lo gical conceptualization of test anxiety: general anxi ety and performance on major test s represent key variables in understanding and predicting test anxiety. In ad dition, this model met statistical assumptions without concerns. Further hierarchical regression analyses revealed that beyond these two key predictors, the inclusion of students perceptions of academic skills (ACES Skills) significantly improves the initial model. Thus, student ratings of academic skills explain a portion of the variance in test anxiety beyond that accounted for by FCAT and MASC scores. In the initial forced-entry regression model, the unique variance associated with student perceptions of academic skills may have been undetected due to the reducti on in power associated with including more predictor variables. Theoretically, student perceptions of academ ic skills should account for some of the variance in test anxiety. For example, if a st udent possessed extremely deficient academic skills
76 and performed poorly on major tests but perceive d having typical or average skills, she or he may not experience as much test anxiety as a student with similar academic performance on major tests along with perceptions of weak academ ic skills. Although the test anxiety literature supports that student perceptions play an im portant role, it primarily considers students perceptions about the demand of academic ma terial (e.g., Hong, 1999) and peer comparisons rather than self-report of skills (Swanson & Howell, 1996). Predictors of Test Anxi ety in Male s and Females Given that FCAT score and MASC Total score we re identified as the primary predictors of test anxiety in the entire sample, these two vari ables were jointly entered into regression models conducted separately for males and females. Although standardized slopes (beta values) should be examined to compare relative influence amon g variables entered within a single model in order to account for different units of measurem ent, non-standardized slopes should be examined when comparing slopes of the same variables across populations (Algina, 2005). For both males and females, the slope for FCAT score was the same. However, MASC score appeared to be slightly more important for females than for males. Comparing relative predictive value within th e model for males, FCAT score appeared to be a slightly more important predictor than MASC score. Conversely, for females, general anxiety appeared to be slightly more important than FCAT score. A significant limitation of these gender-specific regression models is the omi ssion of other relevant variables. However, by splitting the sample between males and female s, the sample size for each group was majorly reduced and therefore fewer predictor variable s could be examined within each model to maintain adequate power.
77 Implications of Findings Theoretical Implications A central implication of this study is that bot h academ ic and emotiona l variables relate to test anxiety when controlling for demographics, grades in school, and self-ratings of academic skills and enablers. Likewise, when all these fact ors are considered together, the other variables do not account for a significant amount of the va riance in test anxiety. In this study, score on high-stakes achievement test represented the most important academic indicator in terms of its relationship with test anxiety. Beyond test performance and general anxiety, student beliefs and perceptions of academic skills also play a role in test anxiety. The role of individual perceptions of academic skills as a relevant predictor of test anxiety has important theoretical implications as well. Drawing from cognitive-behavioral conceptualizations of ot her, non-academic types of childhood anxiety (e.g., Kendall, 1993; Roblek & Piacentini, 2005), stud ents beliefs about their own skills and preparedness to engage in a given task (e.g., so cial interactions, academic evaluations) represent a crucial link between actual expe rienced difficulties with that ta sk and the experience of anxiety pertaining to the activity. This suggests that st udent perceptions of academic skills should be considered as an important variable to be incorporated into the existing information-processing model linking test anxiety to impaired academic performance. In addition to highlighting the joint importance of both academic and emotional factors in terms of predicting test anxiety, th e model offered in this study ma y help explain test anxiety in students without academic deficits. Indeed, Tobias (1985) noted concerns about explaining test anxiety in students with strong or adequate academic skills and study skills. This study may imply that with sufficient levels of general anxiety, a student may experience test anxiety regardless of academic skills. Likewise, even if a student is not generally anxious, he or she may
78 experience anxiety about test-taking if academic skills deficitsand the perception of suchare present. Implications for Research The finding that no t all indicators of acad emic performance are created equal has important implications for educational research. Pa rticularly as these variables relate to each other and test anxiety, GPA, FCAT Score, a nd ACES Skills and Enablers do not appear to measure the exact same construct. While past lite rature has at times focused solely on GPA, test scores, or student ratings of academic skills, futu re research should consider using more than one of these academic indicators or at least consider the limitations associated with solely relying on one academic measure. It would also be interest ing to examine these variables along with student performance on individually-administered standa rdized measures of achievement. Finally, the consideration of student percep tions of academic skills represen ts an important link between academic achievement and many relevant social -emotional experiences (e.g., motivation, selfefficacy). Implications for Practice Consistent with form er intervention studies (e.g., Beidel, Turner, & Taylor-Ferreira, 1999), these findings imply that intervention strategies exclusively focusing on emotional aspects of test anxiety (e.g., relaxation) may be less promisi ng approaches to intervention than a more comprehensive approach also including academic components. Including cognitive-behavioral strategies to cope with gene ral anxiety (e.g., elements of c ognitive restructuring, behavioral exposure) along with academic strategies (e.g., remediating skill deficits and teaching study skills and test-taking strategies) may be optimal ly beneficial for some students. This study implies that for some individuals, underlying symptoms of general anxiety and academic difficulties must be addressed in order to ove rcome the experience of test anxiety.
79 Study Limitations Several lim itations surround the design and natu re of this study. Regarding internal validity and potential for attribut ing causality, the relati onal nature of the data represents a key limitation. Without any experimental manipulation, causality cannot be as clearly determined. Although regression analyses were able to co ntrol for several academic indicators and demographic variables, several potentially-rel evant social-emotional variables were not controlled for in this particular study. Therefore, another limita tion of this study is the omission of potentially relevant predictor variables. For example, the literature supports that perfectionism (Stober & Joorman, 2001) trait anxiety (Beidel et al ., 1994), and related personality characteristics (Lufi et al., 2004) relate to test anxiety. Recently, Ciani and colleagues found that an individuals self-efficacy also may play a role in determining whether autonomic arousal (e.g., increased heart rate) is interpreted as a pleasant or unpleasant experience, particularly as it relates to test anxiety (Ciani, Easter, Summers, Posada, 2009). Although the ACES Skills measure may have tapped into some aspects of participants selfefficacy, this may be an important variable in terms of how the physiological aspects of test anxiety are actually interpreted by a student, which comprises a la rge portion of the test anxiety measure used in this study. Other possibly releva nt variables that were omitted from this study include information regarding stressful life ev ents and experiences, parental anxiety, and cognitive ability. Including these or similar variables may have fu rther clarified the relationship among test anxiety, academic performance, ge neral anxiety, and demographic variables. Including more of these variable s along with a much larger sample size would have improved the validity of overall conclusions drawn by this regression model without compromising power. Several possible limitations of this study stem from method biases. Several common method biases were introduced by the fact that the TAI, MASC, and ACES scales each rely on a
80 similar self-report rating scale format and were m easured at the same point in time, in the same setting (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003). Perhaps one of the biggest flaws in the overall design of this study relate s to common rater effects; that is, several of the predictor variables and the criterion vari able came from the same sour ce (Podsakoff et al., 2003). This introduces possible sources of systematic e rror including respondents tendency to try to maintain consistency in their responses (i.e., consistency motif,) which may be particularly problematic due to the construct similarities between test anxiet y and other types of anxieties (Podsakoff et al, 2003). In other words, regardle ss of whether a student experienced a specific subtype of anxiety, he or she may have responded in a way that is perceived as consistent with earlier answers. The effect of this bias woul d be to inflate the obtai ned correlation between general anxiety and test anxiety. Additional potentia l biases associated with rater effects include social desirability and potent ial implicit theories (Podsakoff et al., 2003). For example, if a student developed his or her own implicit theory that test anxiety is associated with lower academic skills, this may have influenced responses on rating scale measures to support this theory. By including GPA and FCAT scores as independently-obtained da ta in this study, the relationship of academic variables wi th test anxiety is not threaten ed by this particular type of bias. However, the relationship among emotionalbehavioral variables (e .g., test anxiety and other forms of anxiety) may be compromised by th is problem. Overall, the overreliance on selfreport measures represents a considerable limita tion to this study, potenti ally posing a serious problem with regard to conclusions surroundi ng the relationship among rating scale data. Although obtaining academic information about GPA and FCAT from a separate source reduced the threats associated with common method bias, this pa rticular methodological decision may have caused a tradeoff for another threat to the validity of the information obtained.
81 Specifically, this required that some identifying information be provided from participants in order to link their self-report with archival da ta, which possibly reduced their perceptions of anonymity. Perceiving limits to their anonymity ma y have introduced another realm of biases, particularly in adolescent stude nts providing self-report inform ation in a school setting. For example, social desirability biases may have been more salient due to the apparent limited of anonymity. Another limitation associated wi th the correlational nature of this study was the passage of time between taking a major test and completing self-report questionnaires. Collecting information about test anxiety im mediately after or even during a test-taking situation may have yielded more a more accurate depiction of stude nts symptoms during the test. The passage of time may have led students to have inconsistent awareness or memory of specific symptoms of test anxiety. For example, students may be mo re inclined to remember major physiological symptoms (e.g., feeling noticeably nauseous) over more subtle or comfortable sensations (e.g., increased heart rate). Also related to issues of timing, the self-report data in this study was collected towards the end of the academic year, a few weeks prior to final exams. It is unclear whether students would report simila r levels of test anxiety at diffe rent times in the school year and with varying temporal proximity to major evaluations (e.g., FCAT, final exams). Limitations also exist with rega rd to external validity. It is unknown whether these findings would fully generalize to students outside the relati vely narrow range of gr ade levels included in the sample. For example, test anxiety may operate slightly different in older adolescents or elementary-aged children. Academic skills or emotional characteristics may be more important depending upon age. Furthermore, it is unknown whether these findings would generalize to students in other cultures, school environments, or cohorts. The emphasis on high-stakes testing
82 that characterizes most U.S. school systems in recent decades may have shaped standardized-test scores (FCAT) as being particular ly relevant to test anxiety. However, in cultures, schools, or decades without this major emphasis on educational accountability and test-taking, performance on a major academic achievement test may not be nearly as salient in the experience of test anxiety. In some states, perf ormance on statewide tests has varying degrees of importance in terms of academic promotion and graduation. Finally, the academic climat e of the universityaffiliated developmental research school from which this sample was drawn may be more rigorous and achievement-oriented than other schools in the stat e of Florida. Given the strong emphasis on best practices in education and fr equency with which educational research is conducted at this school, students self-ratings of academic skills and enabling behaviors also may be slightly elevated than those obtained by other 7ththrough 9th-graders. In addition, unlike most public schools, this setting requires fa milies to initiate application for enrollment and agree to provide a childs transportation to a nd from school on a daily basis. Thus, the population of this school includes students w hose parents have self-selected their enrollment, a criteria that is significantly different from the enrollment process in most non-choice public schools. Despite its similar demographic composition, a school re quiring choice enrollment likely differs from typical public schools. In addition to differences associated with be ing a choice school, the research setting spans kindergarten through grades twelve which is considered a relatively uncommon feature in public school systems. Therefore, the majority of the participants in this st udy likely had attended the same school for several years. Changing schools introduces an additional potential source of anxiety. In typical ninth-grade students who ha d recently entered a new high school, anxiety may
83 be higher. Conversely, the affiliation with an ac hievement-oriented university potentially could introduce additional pressure for students to achieve academically. Another consideration with regard to genera lizability pertains to the use of the FCAT Reading score specifically. It is unknown whether high-stakes testing specific to other subject areas would have the same relationship with test anxiety as did r eading. Reading may be particularly important since important tests in ot her subject areas (e.g., history, science, and even math) tend to require reading skills. With regard to measurement, the reliance on self-report data for many important variables represents a significant limitation. In addition, students self-repor t about test anxiety may have drawn largely from each individuals knowledge of ones past test performance (e.g., FCAT) rather than reflecting a pure indication about their experience of anxiety. This represents a significant limitation in the self-report informati on obtained on the TAI. Ideally, an external, observable measure of test anxi ety would have also been used to avoid this overreliance on individuals self-report and provide information regarding the convergent validity about students self-report. The TAIalthough widely used in the literature does not appear to represent the best possible instrument with wh ich to measure test anxiety. For one, it only includes 20 items. Particularly given its importance as the key vari able of interest in this study, test anxiety should have been measured using a longer rating scale as well as observable, external information such as teacher or parent ratings and physiological data (e.g., heart rate, galvanic skin response) collected during test-taki ng situations. Summary & Future Directions Overall, bo th general forms of anxiety and academic indicators appear relevant to the experience of test anxiety in adolescent stud ents. High-stakes test performance and general
84 anxiety represent the strongest predictors when controlling fo r the effects of other academic indicators and demogra phic characteristics. Future studies should further examine the predic tors of test anxiety specific to certain subject areas (e.g., reading, math) in adolescent s. Examining the relationship among students self-report of skills within a specific academic dom ain with their anxiety specific to that subject would provide further support for the role of perc eptions and beliefs abou t ones skills as an additional predictor of test anxi ety. Although subject-specific subtypes of test anxiety have been studied in college students and graduate students, few studies have looked at this in younger students. Developmental changes in the expression of test anxiety should also be further explored. In younger students, would other characteristics more strongly predict test anxiety? It may be that early environmental and ear ly school experiences play a mo re important role in younger childrens test anxiety. For exam ple, if a student had not had strong early literacy experiences prior to beginning kindergarten, he or she may experience more test anxiety than same-aged peers. The construct of test anxiety may be different in younger children, too, given the changing expectations for assessment and eval uation. For example, in Florida and many other states, students do not take state-wide achievement tests until the third grade.
85 REFERENCES Ackerman, B.P., Izard, C.E., Kobak, R., Brown, E.D., & Smith C. (2007). Relation between reading problem s and internalizing behavior in school for preadol escent children from economically disadvantaged families. Child Development, 78 581-596. Altemus, M. (2006). Sex differen ces in depression and anxiety disorders: Potential biological determinants. Hormones and Behavior, 50 534-538. Bandalos, D.L., Yates, K., & Thorndike-Christ, T. (1995). Effects of math self-concept, perceived self-efficacy, and attributions for failure and success on test anxiety. Journal of Educational Psychology, 87 611-623. Beidel, D.C. (1991). Social phobia and ove ranxious disorder in school-age children. Journal of the American Academic of Child and Adolescent Psychiatry, 30 545-552. Beidel, D.C., & Turner, S.M. (1988). Comorbidity of test anxiety and other anxiety disorders in children. Journal of Abnormal Child Psychology, 16, 275-287. Biedel, D.C., Turner, S.M., & Taylor-Ferreira, J.C. (1999). Teaching study skills and test-taking strategies to elemen tary school students. Behavior Modification, 23 630-646. Beidel, D. C., Turner, S. M., & Trager, K. N. (1994). Test anxiety and childhood anxiety disorders in African American and White school children. Journal of Anxiety Disorders, 8, 169-179. Bernstein, G.A., Layne, A.E., Egan, E.A., & Tenni san, D.M. (2005). School-based interventions for anxious children. Journal of the American Academy of Child Psychiatry, 44 11181127. Benjamin, M., McKeachie, W.J., Lin, Y.G., & Holi nger, D.P. (1981). Test anxiety: Deficits in information processing. Journal of Educational Psychology 73, 816. Bernstein, G.A., Layne, A.E., Egan, E.A., & Tenni son, D.M. (2005). School-based interventions for anxious children. Journal of the American Acade my of Child and Adolescent Psychiatry, 44, 1118-1127. Birenbaum, M., & Nasser, F. (1994). On the relationship between test anxiety and test performance. Measurement and Evaluation in Counseling and Development, 27,293-302. Blatt, S.J. (1995). The destruc tiveness of perfectionism: Implications for the treatment of depression. American Psychologist, 50 1003-1020. Bodas, J., & Ollendick, T.H. (2005). Test anxiety: A cross-cu ltural perspective. Clinical Child and Family Psychology Review, 8, 65-88. Bryan, J.H., Sonnefeld, L.J., & Grabowski, B. ( 1983). The relationship between fear of failure and learning disabilities. Learning Disability Quarterly, 6, 217-222.
86 Cassady, J.C., & Johnson, R.E. (2002). Cognitive test anxiety and academic performance. Contemporary Educational Psychology, 27, 270-295. Cassady, J.C. (2004). The influence of cognitive te st anxiety across the le arningtesting cycle. Learning and Instruction, 14 569-592. Chapell, M.S., Blanding, Z.B., Silverstein, M. E., Takahashi, M., Newman, B., Gubi, A., & McCann, N. (2005). Test anxiety and academic performance in undergraduate and graduate students. Journal of Educational Psychology, 97 268-274. Corcoran, M.D., Macdougall, M.A., & Scarboro ugh, W.H. (1985). The interplay of worry and emotionality with anxiety and cognitive interf erence in predicting test performance. In H.M. van der Ploeg, R. Schwarzer,, & C.D. Spielberger (Eds.), Advances in test anxiety research (Vol 4, pp. 103-109). Lisse, Netherla nds: Swets & Zeitlinger, B.V. Culler, R.E., & Holahan, C.J. (1980). Test anxiety and academic performance: The effects of study-related behaviors. Journal of Educational Psychology, 72 16-20. Dierker, L.C., Albano, A.M., Clarke, G.N., et al (2001). Screening for a nxiety and depression in early adolescence. Journal of the American Acad emy of Child and Adolescent Psychiatry, 40 929-936 DiPerna, J.C., & Elliot, S.N. (1999). Developmen t and Validation of the Academic Competence Evaluation Scales. Journal of Psychoeducational Assessment, 17, 207-225. DiPerna, J.C., & Elliot, S.N. (2000). Academic competence evaluation scales. San Antonio, TX: The Psychological Corporation. DiPerna, J.C., Volpe, R.J., & Elliot, S.N. (2005). A model of academic enablers and mathematics achievement in the elementary grades. Journal of School Psychology, 43, 379-392. Di Maria, F., & Di Nuovo, S. (1990). Gender differences in social and test anxiety. Personality and Individual Differences, 11 525-530. Elliot, S.N., DiPerna, J.C., Mroch, A.A., & Lang, S.C. (2004). Prevalence and patterns of academic enabling behaviors: An analysis of teachers and students ratings for a national sample of students. School Psychology Review, 33, 302-309. Emery, J.R., & Krumboltz, R.D. (1967). Standa rd versus individualized hierarchies in desensitization to re duce test anxiety. Journal of Counseling Psychology, 14 204-209. Ergene, T. (2003). Effective in terventions on test anxiety re duction: A meta-analysis. School Psychology International, 24 313-328. Everson, H.T., Millsap, R.E., & Rodriguez, C.M. (1991). Isolating gender differences in test anxiety: A confirmatory factor analysis of the test anxiety inventory. Educational and Psychological Measurement, 51 243-251.
87 Eynsenck, M.W. (1985). Anxiety and cognitive-task performance. Personality and Individual Differences, 6, 579-586. Field, A. (2005). Discovering Statistics Using SPSS.(2nd ed.). London: Sage Fincham, F.D., Hokoda, A., & Sanders, R., Jr., ( 1989). Learned helplessness, test anxiety, and academic achievement: A longitudinal analysis. Child Development, 60 138-145. Firmin, M., Hwang, C., Copella, M., & Clark, S. (2004). Learned helplessness: The effect of failure on test-taking. Education, 124 688-693. Fisher, B.L., Allen, R., & Kose, G. (1996). Th e relationship between anxiety and problemsolving skills in children with and without learning disabilities. Journal of Learning Disabilities, 29, 439-446. Fletcher, C., Lovatt, C., Baldry, C. (1997). A stu dy of state, trait, and test anxiety, and their relationship to assessment center performance. Journal of Social Behavior and Personality, 12, 205-214. Florida Department of Edu cation (2008). Student Perfor mance Results: State Reading Demographic Report. Retrieved May 15, 2009, from https://app1.fldoe.org/FCATDemographics/S elections.aspx?level=State&subj=Reading Florida Departm ent of Edu cation (2008). No Child Left Behind (NCLB) School public accountability report. Retrieved May 29, 2009, from http://doewebprd.doe.state.fl.us/eds /nclbspar/index.cfm Frost, R.O., Marten, P.A ., Lahart, C., & Rosenblat e, R. (1990). The dimensions of perfectionism. Cognitive Therapy and Research, 14 449-468. Goetz, T., Frenzel, A. C., & Pekrun, R. (2006). The domain specificity of academic emotional experiences. The Journal of Experimental Education, 75 (1), 5-29. Graham, J. (1990). The MMPI-2: Assessing pers onality and psychopathology. New York: Oxford University Press. Guida, F.V., & Ludlow, L.H. (1989). A cr oss-cultural study of test anxiety. Journal of CrossCultural Psychology, 20 178-190. Haladyna, T.M., Downing, S.M., (2004). Construct irrelevant variance in high-stakes testing. Educational Measurement: Issues and Practice, 23 17-27. Hancock, D.R. (2006). Effects of test anxiety an d evaluative threat on students achievem ent and motivation. Journal of Educational Research, 94 284-290. Harrison, L.A., Stevens, C.M., Monty, A.N., & Coakley, C.A. (2006). The consequences of stereotype threat on academic performan ce of White and non-White lower income college students. Social Psychology of Education, 9, 341-357.
88 Hembree, R. (1988). Correlates, causes, effect s, and treatment of test anxiety. Review of Educational Research, 58 47-77. Hewitt, P.L., Caelian, C.F., Flett, G.L, Sherry, S.B., Collins, L., & Flynn, C.A. (2002).Perfectionism in children: associations with depres sion, anxiety, and anger. Personality and Indivi dual Differences, 32 1049-1061. Hewitt, P. L., & Flett, G. L. (1991b). Perfec tionism in the self and social contexts: conceptualization, assessment, a nd association with psychopathology. Journal of Personality and Social Psychology, 60, 456. 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. Hodapp, V. (1995). The TAI-G: A multidimensional a pproach to the assessment of test anxiety. In: Schwarzer, C. and Zeidner, M. (Eds.), Stress, anxiety, and Coping in Academic Settings, 95-130. Tu bingen, Germany: Francke. Hong, E. (1999). Test anxiety, perc eived test difficulty, and test performance: Temporal patterns their effects. Learning and Individual Differences, 11 431-447. Keogh, E., Bond, F.W ., French, C.C., Richards, A., & Davis, R.E. (2004). Test anxiety, susceptibility to distraction and examination performance. Anxiety, Stress, and Coping, 17, 241-252. King, N.I., Ollier, K., Iacuone, R., Schuster, S., Bays, K., Gullone, E., & Ollendick, T.H. (1989). Fears of children and adoles cents: A cross-sectional Austra lian study using the RevisedFear Survey Schedule for Children. Journal of Child Psychology and Psychiatry, 30, 775-784. King, N.I., Mietz, A., Tinney, L., & Ollendic k, T.H. (1995). Psychopathology and cognition in adolescents experiencing severe test anxiety. Journal of Clinical Child Psychology, 24, 49-54. Kondas, A. (1967). Reduction of examination anxi ety and stage-fright by group desensitization and relaxation. Behaviour Research and Therapy, 5, 275-281. Langley, A.K., Bergman, R.L., McCracken, J., Pi acentini, J.C. (2004). Impairment in childhood anxiety disorders: Preliminary examination of the Child Anxiety Impact ScaleParent Version. Journal of Child and Adolescent Psychopharmacology, 14 105-114. Lazarus, R.S., & Folkman, S. (1984). Stress, appraisal, and coping. New York: Springer. Leibert, R.M., & Morris, L.W. (1967) Cognitive a nd emotional components of test anxiety: A distinction and some initial data. Psychological Reports, 20 975-978.
89 Locker, J., & Cropley, M. (2004). Anxiety, depression,and self -esteem in secondary school children: An investigation into the impact of standard assessment te sts (SATs) and other important school examinations. School Psychology International, 25, 333. Lufi, D., Okasha, S., & Cohen, A. (2004). Test anxiety and its effect on the personality of students with learning disabilities. Learning Disability Quarterly, 27 176-184. March, J., Parker, J., Sullivan, K., Stallings, P., Conners, C.K. (1997). The Multidimensional Anxiety Scale for Children (MASC): Factor structure, reliabilty, and validity. Journal of the American Academy of Child and Adolescent Psychiatry, 36 554. March, J.S., Sullivan, K., & Parker, J. (1999). Test-retest reliability of the Multidimensional Anxiety Scale for Children. Journal of Anxiety Disorders, 13 349. Martindale, T., Pearson, C., Curda, L.K., & Pilcher, J. (2005). Effects of an online instructional application on reading and mathema tics standardized test scores. Journal of Research on Technology in Education, 37, 349-360. Marx, D.M, & Stapel, D.A. (2006). Its all in the timing: Measuring emotional reactions to stereotype threat before and after taking a test. European Journal of Social Psychology, 36, 687. Matters, G., & Burnett, P. C. (2003). Psychologica l predictors of the pr opensity to omit shortresponse items on a high-stakes achievement test. Educational and Psychological Measurement, 63, 239-256. Matthews, G., Hillyard, E.J., and Campbell, S.E. (1999). Megacognition and maladaptive coping as components of test anxiety Clinical Psychology and Psychotherapy, 6 ,11-125. McDonald, A.S. (2001). The prevalence and effect s of test anxiety in school children. Educational Psychology, 21 89-101. McLeod, B.D., Wood, J.J., & Weisz, J.R. (2007). Examining the association between parenting and childhood anxiety: A meta-analysis. Clinical Psychology Review, 27 155-172. Morris, L.W., Davis, M.A., & Hutchings, C.H. (1981). Cognitive and emotional components of anxiety: Literature review and a revised worry-emotionality scale. Journal of Educational Psychology, 73 541-555. Muris, P., Merckelbach, H., Ollendick, T., King, N., & Bogie, N. (2002). Three traditional and three new childhood anxiety questionnaires: th eir reliability and validity in a normal adolescent sample. Behaviour Research and Therapy, 40 753. Naveh-Benjamin, M. (1991). A comparison of trai ning programs for intended different types of test anxious students: Further support for inform ation processing model. Journal of Educational Psychology, 83, 134.
90 Nottelmann, E.D., & Hill, K.T. (1977). Test anxiety and off-task behavior in evaluative situations. Child Development, 48, 225-231. Ollendick, T.H., & Meyer, J. (1984). Sc hool phobia. In S.M. Turner (Ed.), Behavioral theories and treatment of anxiety (pp. 367-411). New York: Plenum. Ollendick, T.H., King, N.J., & Frary, R.B. (1989). Fears in children and adolescents:reliability and generalizability across ge nder, age, and nationality. Behaviour Research & Therapy, 27,19. Orbach, G., Lindsay, S., & Grey, S. (2007). A ra ndomised placebo-controlled trial of a self-help internet-based interven tion for test anxiety. Behaviour Research and Therapy, 45, 483496. Osborne, J.W. (2001). Testing stereotype threat: does anxiety explain race and sex differences in achievement? Contemporary Educational Psychology, 26 291. Paulman, R.G., & Kennelly, K.J. (1984). Test an xiety and ineffective te st taking: Different names, same construct. Journal of Educational Psychology, 76 279-288. Peleg-Popko, O. (2002). Childrens test an xiety and family interaction patterns. Anxiety,Stress, and Coping, 15, 45-59. Petrocelli, J.V. (2003). Hierar chical multiple regression in counseling research: Common problems and possible remedies. Measurement and Evaluation in Counseling and Development, 36, 9-22. Pintrich, P.R., & De Groot, E.V. (1990). Motiva tional and self-regulated learning components of classroom academic performance. Journal of Educational Psychology, 82, 33-40. Podsakoff, P.M., MacKenzie, S.B., Lee, J., & Podsakoff, N.P. (2003). Common method biases in behavioral research: A critical review of the literature and recommended remedies. Journal of Applied Psychology, 88, 879-903. Preckel, F., Zeidner, M., Goetz, T., & Schleyer E. (2008). Female big fish swimming against the tide:The big-fish-little-pond effect and gender-ratio in special gifted classes. Contemporary Educational Psychology, 33 78. Putwain, D.W. (2007). Test anxiety in UK sc hool children: Prevalence and demographic patterns. British Journal of E ducational Psychology, 77, 579-593. Journal ofPersonality and Social Psychology, 72 Reynolds, C. R., & Richmond, B. O. (1985). Revised Children s Manifest Anxiety Scale (RCMAS): Manual Los Angeles: Western Psychological Services.
91 Roblek, T.R., Piacentini, J. (2005). Cognitive-beha vioral therapy for childhood anxiety disorders. Child and adolescent psychiatric clinics of North America, 14, 863-876. Rocklin, T., & Ren-Min, Y. ( 1989). Development and adaptation of the Chinese Test Anxiety Inventory: A Research Note. In R. Schwarzer, H. M. van der Ploeg, & C. D. Spielberger (Eds.), Advances in test anxiety research (Vol. 6, pp. 245). Lisse, The Netherlands: Swets & Zeitlinger. Rosenthal, R.L. (1991). Meta-analytic procedures for social research (revised). Newbury Park, CA: Sage. Rynn, M.A., Barber, J.P., Kalid-Khan, S., Sique land, L., Dembiski, M., McCarthy, K.S.,Gallop, R. (2006). The psychometric properties of the MASC in a pediatric psychiatric sample. Journal of Anxiety Disorders, 20, 139-157. Sapp, M. (1999). Test anxiety: Applied research, a ssessment, and treatment interventions (2nd ed.) Lanham, MD: University Press of America. Sarason, I.G. (1984). Stress, anxiety, and c ognitive interference: Reactions to tests. Journal of Personality and Social Psychology, 46 929-938. Sarason, I. G., & Palola, E. G. (I960). The relationship of test and general anxiety, difficulty of task, and experimental inst ructions to performance. Journal of Experimental Psychology, 59, 185-191. Sarason, I.G., Sarason, B.R., & Pierce, G.R. (1990). Anxiety, cognitive interference, and performance. Journal of Social Behavior and Personality, 5 1-18. Sarason, I. G., & Stoops, R. (1978). Test anxiety and the passage of time. Journal of Consulting and Clinical Psychology, 46, 102-109. Scholze, S., & Sapp, M. (2006) Understanding te st anxiety and the multicultural learner. Multicultural Learning and Teaching, 1, 13-23. Schwarzer, R. (1984). Worry and emotionality as separate component s in test anxiety. International Review of Applied Psychology, 33, 205-220. Sharma, S., Sud, A., & Spielberger, C. D. ( 1983). Development of the Hindi form of Test Anxiety Inventory. In H. M. van der Ploeg, R. Schwarzer, & C. D. Spielberger (Eds.), Advances in test anxiety research (Vol. 2, pp. 183). Lisse, The Netherlands: Swets & Zeitlinger. Smyth, T.S. (2008). Who is No Child Left Behind leaving behind? Clearing House, 81, 133-137. Somers, J.M., Goldner, E.M., Waraich, P., & Hsu, L. (2006). Prevalence and incidence of anxiety disorders: A systema tic review of the literature. Canadian Journal of Psychiatry, 51 100-113.
92 Spielberger, C.D. (1972). Anxiety as an emotional state., in C.D. Spielberger (ed.) Anxiety: Current Trends in Theory and Research. New York: Academic Press. Spielberger, C.D. (1975). The measurement of state and trait anxiety: Conceptual and methodological issues. In L. Levi (Ed.), EmotionsTheir parameters and measurement (pp. 713-725). New York: Raven Press. Spielberger, C.D., Gonzalez, H.P., Taylor, C.J., Anton, E.D., Algaze, B., Ross, G.R., & Westberry, L.G. (1980). Test Anxiety Inventory Menlo Park, CA: Mindgarden. Spielberger, C.D. and Vagg, P.R. (1995). Treatm ent of test anxiety: Application to the transactional process model. In C.D. Spielberger & P.R. Vagg (eds.), Test Anxiety: Theory, Assessment, and Treatment, p. 197-212. Washington, D.C.: Taylor and Francis. Stober, J. (1998). The Frost Multidimensional Pe rfectionism Scale: More perfect with four (instead of six) dimensions. Personality and Individual Differences, 24, 481-491. Stober, J. (2004). Dimensions of te st anxiety: Relations to ways of coping with pre-exam anxiety and uncertainty. Anxiety, Stress, and Coping, 17 213-226. Stober, J. & Joorman, J. (2001). Worry, procra stination, and perfectio nism: Differentiating amount of worry, pathological wo rry, anxiety, and depression. Cognitive Therapy and Research, 25 49-60. Swanson, S., & Howell, C. (1996). Test anxiety in adolescents w ith learning disabilities and behavior disorders. Exceptional Children, 62 389-397. Tobias, S. (1985). Test anxiety: Interference, defective skills, and cognitive capacity. Educational Psychologist, 20, 135-142. Turner, B.G., Beidel, D.C., Hughes, S. & Tu rner, M.W. (1993). Test anxiety in African American school children. School Psychology Quarterly, 8, 140-152. Tryon, G. (1980). The measurement and treatment of test anxiety Review of Educational Research, 50 353. Van Boxtel, H. W., & Monks, F.J. (1992). General, social, and academic self-concepts of gifted adolescents Journal of Youth and Adolescence, 21, 169-186. Wachelka, D., & Katz, R.C. (1999). Reducing te st anxiety and improving academic self-esteem in high school and college student s with learning disabilities. Journal of Behavior Therapy and Experimental Psychiatry, 30 191-198. Warren, S.L., Ollendick, T.H., King, N.J. (1996). Test anxiety in girls and boys: A clinicaldevelopmental analysis. Behaviour Change, 13, 157-170. Williams, J.E. (1991). Modeling test anxiety, self concept and high school students academic achievement. Journal of Research and Development in Education, 25 51-57.
93 Whitaker Sena, J.D., Lowe, P.A., & Lee, S.W. (2007). Significant predictors of test anxiety among students with and without learning disabilities. Journal of Learning Disabilities, 40, 360. Wine, J.D. (1979). Test anxiety and evaluation threat: Children s behavior in the classroom. Journal of Abnormal Child Psychology, 7, 45-59. Wren, D. & Benson, J. (2004). Measuring Test Anxiety in Children: Scale Development and Initial Construct Validation. Anxiety, Stress, and Coping, 17 227-240. Zeidner, M. (1998). Test anxiety: The state of the art. New York: Plenum Press. Zeidner, M. & Safir, M. (1989). Sex, ethnic, and social differences in test anxiety among Israeli adolescents. Journal of Genetic Psychology, 150 175-185. Zeidner, M., & Schleyer, E. J. (1999). The big-fish-little-pond effect for academic self concept, test anxiety, and school gr ades in gifted children. Contemporary Educational Psychology, 24, 305. Zeidner, M., Nevo, B., & Lipschitz, H. (1988) The Hebrew version of the test anxiety inventory.Haifa: University of Haifa (Hebrew).
94 BIOGRAPHICAL SKETCH Anne Larm ore Bruehl was born in September 1981, in Richmond, Virginia. She grew up in Chesterfield County, Virginia and graduated fr om Monacan High School in 1999. Anne earned a bachelors degree in psychology from the College of William & Mary in Williamsburg, Virginia, graduating as a member of the alpha chapte r of the Phi Beta Kappa honor organization. Following the completion of her undergraduate de gree, Anne enrolled in the School Psychology doctoral program at the University of Florida. A nne completed her pre-doctoral internship with Prince William County Schools in Manassas, Virginia before graduating with her Ph.D. in 2009.