USF Libraries
USF Digital Collections

Multiple approaches to the validation of the scores from the study anxiety inventory

MISSING IMAGE

Material Information

Title:
Multiple approaches to the validation of the scores from the study anxiety inventory
Physical Description:
Book
Language:
English
Creator:
Lunsford, George Douglas
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Study skills and habits
Procrastination
Construct validity
Confirmatory factor analysis
Measurement invariance
Gender
Undergraduates
Dissertations, Academic -- Educational Measurement and Research -- Doctoral -- USF   ( lcsh )
Genre:
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: The Study Anxiety Inventory (SAI), consisting of the factors of worry and emotionality, was developed to measure college students' self-reported levels of anxiety while studying for an exam. Data from 2002 undergraduate students from four colleges (Arts and Sciences, Engineering, Business, and Education) at a southeastern state university were used to evaluate the validity of the scores from the 16-item Study Anxiety Inventory. Results of confirmatory factor analyses for the two factor model, conducted separately for each college, indicated marginally acceptable fit for the data (median fit measures across the four colleges: CFI =.915, SRMR=.049, RMSEA=.098), a pattern that was repeated for both males and females. Multigroup CFA was used to evaluate the factorial invariance of the SAI across gender within each college. Factor loadings (i.e., pattern coefficients) for the SAI items were not found to be significantly different between males and females (p > .05).Error variances for four items were found to be significantly different between males and females, indicating that there may be some difference in scale reliability by gender. Factor covariances were invariant for all four colleges (p > .05) and factor variances were invariant for all but the worry component for the College of Arts and Sciences where females had significantly greater variability on the worry factor. As was hypothesized, the SAI scores were positively correlated with scores on measures of test anxiety (median r=.74), trait anxiety (median r=.46), active procrastination (median r=.23), and passive procrastination (median r=.29), but negatively correlated with trait curiosity (median r=-.19). Contrary to what was hypothesized, no relationship was demonstrated between study anxiety and study skills and habits (median r=-.03).The nomological network was extended in this study by examining relationships between scores obtained from students on the SAI and measures of active and passive procrastination. This is the first study that systematically examines the factorial invariance of the SAI by gender, which is important because previous research using the SAI has shown men's scores to be consistently lower than women's scores. The results obtained in the current study provide support for gender invariance in a nonclinical population in the situation specific level of anxiety while studying.There is sufficient evidence of validity and reliability (median Cronbach alphas for males and females for the total score were .978 and .980, for worry were .968 and .973, and for emotionality were .947 and .951, respectively) that a researcher should feel confident that the SAI is a psychometrically sound research tool that holds up fairly well across a number of different types of students and that making mean comparisons on the SAI by gender is acceptable.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2009.
Bibliography:
Includes bibliographical references.
System Details:
Mode of access: World Wide Web.
System Details:
System requirements: World Wide Web browser and PDF reader.
Statement of Responsibility:
by George Douglas Lunsford.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 179 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002068291
oclc - 606600542
usfldc doi - E14-SFE0003125
usfldc handle - e14.3125
System ID:
SFS0027441:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam 2200421Ka 4500
controlfield tag 001 002068291
005 20100412141758.0
007 cr bnu|||uuuuu
008 100412s2009 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0003125
035
(OCoLC)606600542
040
FHM
c FHM
049
FHMM
090
LB3051 (Online)
1 100
Lunsford, George Douglas.
0 245
Multiple approaches to the validation of the scores from the study anxiety inventory
h [electronic resource] /
by George Douglas Lunsford.
260
[Tampa, Fla] :
b University of South Florida,
2009.
500
Title from PDF of title page.
Document formatted into pages; contains 179 pages.
Includes vita.
502
Dissertation (Ph.D.)--University of South Florida, 2009.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
520
ABSTRACT: The Study Anxiety Inventory (SAI), consisting of the factors of worry and emotionality, was developed to measure college students' self-reported levels of anxiety while studying for an exam. Data from 2002 undergraduate students from four colleges (Arts and Sciences, Engineering, Business, and Education) at a southeastern state university were used to evaluate the validity of the scores from the 16-item Study Anxiety Inventory. Results of confirmatory factor analyses for the two factor model, conducted separately for each college, indicated marginally acceptable fit for the data (median fit measures across the four colleges: CFI =.915, SRMR=.049, RMSEA=.098), a pattern that was repeated for both males and females. Multigroup CFA was used to evaluate the factorial invariance of the SAI across gender within each college. Factor loadings (i.e., pattern coefficients) for the SAI items were not found to be significantly different between males and females (p > .05).Error variances for four items were found to be significantly different between males and females, indicating that there may be some difference in scale reliability by gender. Factor covariances were invariant for all four colleges (p > .05) and factor variances were invariant for all but the worry component for the College of Arts and Sciences where females had significantly greater variability on the worry factor. As was hypothesized, the SAI scores were positively correlated with scores on measures of test anxiety (median r=.74), trait anxiety (median r=.46), active procrastination (median r=.23), and passive procrastination (median r=.29), but negatively correlated with trait curiosity (median r=-.19). Contrary to what was hypothesized, no relationship was demonstrated between study anxiety and study skills and habits (median r=-.03).The nomological network was extended in this study by examining relationships between scores obtained from students on the SAI and measures of active and passive procrastination. This is the first study that systematically examines the factorial invariance of the SAI by gender, which is important because previous research using the SAI has shown men's scores to be consistently lower than women's scores. The results obtained in the current study provide support for gender invariance in a nonclinical population in the situation specific level of anxiety while studying.There is sufficient evidence of validity and reliability (median Cronbach alphas for males and females for the total score were .978 and .980, for worry were .968 and .973, and for emotionality were .947 and .951, respectively) that a researcher should feel confident that the SAI is a psychometrically sound research tool that holds up fairly well across a number of different types of students and that making mean comparisons on the SAI by gender is acceptable.
538
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
590
Advisor: Robert Dedrick, Ph.D.
653
Study skills and habits
Procrastination
Construct validity
Confirmatory factor analysis
Measurement invariance
Gender
Undergraduates
690
Dissertations, Academic
z USF
x Educational Measurement and Research
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.3125



PAGE 1

Multiple Approaches to the Validation of th e Scores From the Study Anxiety Inventory by George Douglas Lunsford A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Educational Measurement and Research College of Education University of South Florida Major Professor: Robe rt F. Dedrick, Ph.D. Committee member: Bruce W. Hall, Ed.D. Committee member: Jeffrey D. Kromrey, Ph.D. Committee member: James A. Eison, Ph.D. Date of Approval July 14, 2009 Keywords: study skills and habits, procrastina tion, construct validity, confirmatory factor analysis, measurement invariance, gender, undergraduates Copyright 2009, George Douglas Lunsford All rights reserved.

PAGE 2

i Table of Contents Tables ........................................................................................................................ iiiFigures....................................................................................................................... ivAbstract ...................................................................................................................... .1Chapter 1 Introduction ................................................................................................3Purpose Statement ..................................................................................................18Significance............................................................................................................20Limitations .............................................................................................................21Chapter 2 ...................................................................................................................22Review of the Literature ...........................................................................................22Overview the Validation Process ...........................................................................22Evidence Based on Instrument Content ................................................................ 23Evidence based on Internal Stru cture of Item Responses ..................................... 24Evidence Based on Relations to Other Variables ................................................. 24Development and Initial Validation of the Study Anxiety Inventory .......................26Conceptualization of Study Anxiety ......................................................................26Types of Anxiety................................................................................................... 28Internal Structure: Assessment of State, Trait, and Situation Specific Anxiety ...................................................................................................................36State and Trait Anxiety ......................................................................................... 36Test Anxiety .......................................................................................................... 37Study Anxiety ....................................................................................................... 38External Evidence: Nomological Network ............................................................45Extending the Nomological Netw ork in the Present Study ...................................47Procrastination ...................................................................................................... 47Studying ................................................................................................................ 49Summary ................................................................................................................51Chapter 3 Method .....................................................................................................52Purpose ...................................................................................................................52Participants .............................................................................................................53Measures ................................................................................................................56Test Anxiety Inventory ......................................................................................... 56State-Trait Personality Inventory .......................................................................... 57Study Habits Evaluation and Instruction Kit ........................................................ 58The Passive (PPS) and The Active (APS) Procrastination Scale ......................... 58Procedure ...............................................................................................................59Data Analysis .........................................................................................................61Chapter 4 Results ......................................................................................................63

PAGE 3

ii Research Question 1: Confirmatory factor analysis and factorial invariance ........63College of Arts and Sciences ................................................................................ 64College of Engineering ......................................................................................... 72College of Business............................................................................................... 77College of Education............................................................................................. 82Research Question 2: Factor ial Invariance by Gender ..........................................88College of Arts and Sciences ................................................................................ 89College of Engineering ......................................................................................... 98College of Business............................................................................................. 107College of Education........................................................................................... 115Summary of Results for Research Question 2 .....................................................122Research Question 3: Relationships between SAI and Related Measures ..........126Reliability of Scores .............................................................................................127Correlational Results for SAI a nd Constructs of Interest ....................................128Anxiety Measures ............................................................................................... 128Curiosity and Study Skills and Habits ................................................................ 130Procrastination .................................................................................................... 131Summary of Results for Research Question 2 .....................................................133Chapter 5 Discussion ..............................................................................................134Background ..........................................................................................................134Research Question One: Evidence of Two-Factor Structure ...............................135Research Question Two: Eviden ce of Invariance by Gender ..............................137Research Question Three: Validity Evidence Based on Relations to Other Variables ..............................................................................................................140Significance of the Study .....................................................................................144Limitations ...........................................................................................................147Recommendations for Future Research ...............................................................149References ...............................................................................................................154Appendix A: Summary of findings from studies using the SAI .............................166Appendix B: Preamble to data collection ...............................................................168Explanation of the study and wh at the consent form says ...................................169Educational Debriefing ....................................................................................... 170Appendix C: Experimental Measures .....................................................................171Study Anxiety Inventory ......................................................................................172Test Anxiety Inventory ........................................................................................173Trait Anxiety and Trai t Curiosity scales ..............................................................174Study for Examinations scale ...............................................................................175Active and Passive Proc rastination scales ...........................................................177Scoring for the measures ......................................................................................178About the Author ....................................................................................................179

PAGE 4

iii Tables Table 1 Pearson Product Moment Correlati ons Between Anxiety Variables (n=536) ..................................................................................................................46Table 2 Gender and Ethnicity of 2,002 Participants Across Four Colleges ............56Table 3 Descriptive Statistics for the 16 Observed Variables Used in the Two-Factor Confirmatory Factor Anal yses of the Study Anxiety Inventory for the College of Arts and Sciences (n=662) ........................................................66Table 4 Modification Indices for Error Co variances for the Study Anxiety Inventory for Responses from All Arts and Science Students (n=625) ..................70Table 5 Descriptive Statistics for the 16 Observed Variables Used in the Two-Factor Confirmatory Factor Anal yses of the Study Anxiety Inventory for the College of Engineering (n=433) ................................................................73Table 6 Modification Indices for Error Co variances for the Study Anxiety Inventory for Responses from All Engineering Students .......................................76Table 7 Descriptive Statistics for the 16 Observed Variables Used in the Two-Factor Confirmatory Factor Anal yses of the Study Anxiety Inventory for the College of Business (n=399) ......................................................................78Table 8 Modification Indices for Error Co variances for the Study Anxiety Inventory for Responses from All Business Students (n=399) ...............................81Table 9 Descriptive Statistics for the 16 Observed Variables Used in the Two-Factor Confirmatory Factor Anal yses of the Study Anxiety Inventory for the College of Education (n=410) ....................................................................83Table 10 Modification Indices for Error Covariances for the Study Anxiety Inventory for Responses from All Education Students (n=410) ............................86Table 11 Fit Indices for the Confirmato ry Factor Analysis of the Hypothesized Two-Factor and One-Fact or Model for the Study Anxiety Inventory Across Four Colleges ............................................................................88Table 12 Descriptive Statistics for the 16 Observed Variables Used in the Two-Factor Confirmatory Factor Anal yses of the Study Anxiety Inventory for the College of Arts and Sciences by Gender (nM=215, nF=445) .....................90Table 13 Modification Indices for Error Covariances for the Study Anxiety Inventory for Responses from Male Arts and Science Students (n = 225) ............92Table 14 Modification Indices for Error Covariances for the Study Anxiety Inventory for Responses from Female Arts and Science Students (n = 410) ........93Table 15 Goodness-of-Fit Indices for Mode ls Tested for Invariance of Scores on the Study Anxiety Inventory by Gender (n = 660 A & S Students, nM=215, nF=445) ...................................................................................................95Table 16 Goodness-of-Fit Indices for Invari ance of Loadings on the Study Anxiety Inventory by Gender (n = 660 A & S Students, nM=215, nF=445) ...........95Table 17 Goodness-of-Fit Indices for Invari ance of Error Variances on the Study Anxiety Inventory by Ge nder (n = 660 A & S Students, nM=215, nF=445) ..................................................................................................................97

PAGE 5

iv Table 18 Descriptive Statistics for the 16 Observed Variables Used in the Two-Factor Confirmatory Factor Anal yses of the Study Anxiety Inventory for the College of Engineering by Gender (nM=243, nF=192) ............................100Table 19 Modification Indices for Error Covariances for the Study Anxiety Inventory for Responses from Male Engineering Students (n = 240) .................102Table 20 Modification Indices for Error Covariances for the Study Anxiety Inventory for Responses from Fema le Engineering Students (n = 188) ..............103Table 21 Goodness-of-Fit Indices for Mode ls Tested for Invariance of Scores on the Study Anxiety Invento ry by Gender (n = 428 Engineering Students, nM=240, nF=188) .................................................................................105Table 22 Goodness-of-Fit Indices for Invari ance of Error Variances on the Study Anxiety Inventory by Gende r (n = 428 Engineering Students, nM=240, nF=188) .................................................................................................106Table 23 Descriptive Statistics for the 16 Observed Variables Used in the Two-Factor Confirmatory Factor Anal yses of the Study Anxiety Inventory for the College of Business by Gender (nM=216, nF=193) ..................................108Table 25 Modification Indices for Error Covariances for the Study Anxiety Inventory for Responses from Fe male Business Students (n = 189) ...................111Table 26 Goodness-of-Fit Indices for Mode ls Tested for Invariance of Scores on the Study Anxiety Inventory by Gender (n = 399 Business Students, nM=210, nF=189) .................................................................................113Table 27 Goodness-of-Fit Indices for Invari ance of Error Variances on the Study Anxiety Inventory by Gende r (n = 399 Business Students, nM=210, nF=189) ................................................................................................................114Table 28 Descriptive Statistics for the 16 Observed Variables Used in the Two-Factor Confirmatory Factor Anal yses of the Study Anxiety Inventory for the College of Education by Gender (nM=210, nF=203) ...............................116Table 29 Modification Indices for Error Covariances for the Study Anxiety Inventory for Responses from Male Education Students (n = 208) .....................118Table 30 Modification Indices for Error Covariances for the Study Anxiety Inventory for Responses from Fema le Education Students (n = 202) .................119Table 31 Goodness-of-Fit Indices for Mode ls Tested for Invariance of Scores on the Study Anxiety Inventory by Gender (n = 410 Education Students, nM=208, nF=202) .................................................................................121Table 32 Goodness-of-Fit Indices for Invari ance of Error Variances on the Study Anxiety Inventory by Gende r (n = 410 Education Students, nM=208, nF=202) ................................................................................................................122Table 33 Fit Indices for the Confirmato ry Factor Analysis of the Hypothesized Two-Factor Model by Gender for the Study Anxiety Inventory Across Four Colleges ..........................................................................123Table 34 Item Pairs that Showed Signific ant Chi Squares for Modification Indices on the Study Anxiety Invento ry by Gender Across Four Colleges ..........125Table 35 Summary of Results for Models for the Study Anxiety Inventory Tested for Invariance by Gender by Colleges ......................................................126

PAGE 6

v Table 36 Cronbach’s Alpha for Internal Cons istency for Constructs of Interest for Four Colleges ....................................................................................128Table 37 Pearson Product Moment Correlations Between the SAI, SA/e and SA/w with Anxiety Constructs of Inte rest for Students from Four Colleges ........130Table 38 Pearson Product Moment Correlation for Other Constructs of Interest for Students from Four Colleges .............................................................131Table 39 Pearson Product Moment Correlations Between Each Study Anxiety Variable and Each Procrastina tion Scale for Arts and Sciences, Engineering, Business and Education Students ...................................................132

PAGE 7

iv Figures Figure 1. Published Articles Found on Psyc Info Concerning Test Anxiety. ...........30Figure 2. Information Processing Model. .................................................................30Figure 3. Lazarus’s Transactional Process Theory. ..................................................31Figure 4. A Suggested Expanded Model Showing Study and Test Worry and Emotionality. ...................................................................................................32Figure 5. Relationships of Items to Fact ors in the Two-Factor Model. ...................67Figure 6. Relationships of Items to Fact ors in a One-Factor Model. .......................72

PAGE 8

1 Multiple Approaches to the Validation of th e Scores From the Study Anxiety Inventory George Douglas Lunsford Abstract The Study Anxiety Inventory (SAI), cons isting of the factors of worry and emotionality, was developed to measure college students’ self-reported levels of anxiety while studying for an exam. Data from 2002 undergraduate students from four colleges (Arts and Sciences, Engineering, Business, and Education) at a southeastern state university were used to evaluate the va lidity of the scores from the 16-item Study Anxiety Inventory. Results of confirmatory factor analys es for the two factor model, conducted separately for each college, indicated margin ally acceptable fit for the data (median fit measures across the four colleges: CFI =.915, SRMR=.049, RMSEA=.098), a pattern that was repeated for both males and females. Multigroup CFA was used to evaluate the factorial invariance of the SAI across gender within each college. Factor loadings (i.e., pattern coefficients) for the SAI items were not found to be signi ficantly different between males and females ( p > .05). Error variances for f our items were found to be significantly different between males and fema les, indicating that there may be some difference in scale reliability by gender. Factor covariances were invariant for all four colleges ( p > .05) and factor variances were invari ant for all but the worry component for

PAGE 9

2 the College of Arts and Sciences where fema les had significantly greater variability on the worry factor. As was hypothesized, the SAI scores were positively correlated with scores on measures of test anxiety (median r =.74), trait anxiety (median r =.46), active procrastination (median r =.23), and passive procrastination (median r =.29), but negatively correlated with trait curiosity (median r =-.19). Contrary to what was hypothesized, no relationship was demonstrated between study anxiety and study skills and habits (median r =-.03). The nomological network wa s extended in this study by examining relationships between scores obtai ned from students on the SAI and measures of active and passive procrastination. This is the first study that systematically examines the factoria l invariance of the SAI by gender, which is important because pr evious research using the SAI has shown men’s scores to be consistently lower than women’s scores. The results obtained in the current study provide support for gender inva riance in a nonclinical population in the situation specific level of anxiety while study ing. There is sufficient evidence of validity and reliability (median Cronbach alphas for males and females for the total score were .978 and .980, for worry were .968 and .973, and for emotionality were .947 and .951, respectively) that a researcher should feel confident that the SAI is a psychometrically sound research tool that holds up fairly well across a number of different types of students and that making mean comparis ons on the SAI by gender is acceptable.

PAGE 10

3 Chapter 1 Introduction The construct of test anxiet y has been the subject of much research. Sarason and Mandler (1952) are generally credited with establishing test anxiety as an important psychological construct, which they defined as a “drive” with emotional arousal and worry cognitions evoked in examination s ituations that have a negative effect on performance (S. Sarason, Hill, & Zimbardo, 1964). Spielberger (1980) developed the Test Anxiety Inventory (TAI, 1980) to measure test anxiety as a situation-specific trait. This measure has become one of the most popular for research and has been used in thousands of studies that have examined the effects of anxiety duri ng testing. From this research, techniques have been developed th at are widely used in counseling centers across the country to help alleviate this anxiety (see Zeidner, 1998 for a review). It is interesting that, alth ough the anxiety felt during a te st has been researched in depth, there has been very lit tle published research concerning the anxiety that one experiences while studying for a test. Ge tting information into memory (encoding), retaining that information (storage) and getti ng that information back out (retrieval) may be influenced by anxiety at these different stages. Cognitive psychology suggests that if a student is unsuccessful in his/her attempt to encode the information due to some interference such as anxiety experienced during studying, then it follows that the retrieval performance of the student would reflect the lack of encoding. Studies have shown that when study skills have been used in conj unction with group counseling techniques to facilitate coping with anxi ety during studying, students’ grades improved (Gonzalez,

PAGE 11

4 1995). This may be due to improved skills in studying but may also be due to skill in learning to deal with anxiet y felt while studying, which may reduce interference with the encoding process. There have been a number of studies that show an interaction effect of study skills and test anxiety on test performan ce. When students were told that they were going to be evaluated, the hi gh test anxiety/poor study ha bits group performed more poorly than the high test anxiety/good study habits group and regardless of study habits, the low anxiety students perf ormed better than both of t hose two groups (I. Sarason & Smith, 1971). On the basis of further resear ch, Naveh-Benjamin (1991) concluded that test performance of test anxious students was influenced by both the interference of retrieval by worry and emotionality during te sts, and the organiza tion and encoding of material at the time of studying for a test. He suggested that the performance of students with high test anxiety and good study habits was reduced by the interference in the retrieval from memory during tests, wher eas the performance of high test anxious students with poor study habits was poorer beca use of both interference with retrieval and inability to organize and encode the material. The view that it would be beneficial to use both test and study anxiety relieving methods and teaching study skills is supporte d by an intervention study using behavioral modification and study counseling in which G onzales (1978) demonstrated that grade point average (GPA) improved for high test anxious students who had good study habits but did not improve for high test anxious students with poor study habits. Students who showed a substantial reduction in test anxi ety made the greatest improvement in GPA, indicating that reducti on of test anxiety in test anxi ous students with good study habits

PAGE 12

5 contributed to an improved GPA by eliminating the adverse interference effects of worry and emotionality while taking tests. Impr ovements were also found for students who showed a reduction in anxiety while studying. These results are consistent with Spielb erger’s reports, starting as early as 1966, that students complained that “anxiety re duced effectiveness in studying…” (p. 361), singling out study anxiety as an important explanatory variab le to understanding students’ performances on tests. The importan ce of study anxiety is suggested by the fact that Spielberger included items that dealt with anxiety felt prior to an exam (“ I worry a great deal before taking an important examination ”), even though the time reference for the items was inconsistent with the definition of test anxiety. In th at many studies have shown that anxiety during an exam can inte rfere with retrieval, it seems reasonable to suggest that those who worry before taki ng an important exam may have difficulty encoding information for later retrieval. This suggests that another situation-specific construct that may affect th e encoding of information prio r to taking an exam is study anxiety. A person suffering from study anxiet y would, while studying for exams, experience both worry and emotionality symptoms. The worry symptoms of study anxiety might include: thinking ab out grades or lack of prepar edness in a course so much that it interferes with learning; thought s freezing up; mind wandering; being easily distracted so that ot her thoughts interfere with learning; thinking of the consequences of failing that interferes with the learning pr ocedure or concentrati on; getting a mental block; increasing confusion as effort increases; having a sense of self defeat; an inability

PAGE 13

6 to retain what is studied for long or forget ting it quickly; thoughts of no longer being able to cope with school; wanting to drop cla sses; worrying about being disorganized physically and mentally; worryi ng about having to study longer than others to get the same results; and worry thoughts of doing poorly like “I’m not getting this” or “I can’t absorb the material properly.” The emotionality symptoms of study anxiety may include: getting tense, uneasy, upset feelings; feeli ng jittery; feeling nervous; feelings of panic; feeling stressed; increased speed or strength of heartbeat; shallow or difficu lt breathing; feeling hot, cold or breaking out in a sweat; showing signs of stress; having the st omach tighten; feeling physically ill (maybe nauseous); and feeling frustrated to the poi nt of distraction. The relevance of study anxiety as a fact or in test performance is supported by interviews with university students who were plagued with anxiet y in their pursuit of their degrees (Spielberger, 1966). The most in teresting conversations were held with those who explained that their anxiety did not hinder them during exams as they could reason that once they entered the exam room there was no more they could do to learn the material and hence they became calmer. This s uggests that the time factor for this anxiety separates the concept of test anxiety from study anxiety. It is also evident that this construct is different from test anxiety in th at the environment of the exam is set by the instructor of the course wh ile the study environment is established by the student. Comments from the students like “I often find that I think I must cook dinner before I can start studying” suggest that procrastination may be a symp tom of study anxiety. Finally,

PAGE 14

7 the construct of study anxiety differs from test anxiety in th at the interference experienced is for encoding instead of for retrieval. A search of the literature on study anxiety reveals that there have been several attempts to develop a measure of study anxi ety as a scale within inventories measuring various aspects of studying. For example, We lsh, Bachelor, and Wright (1990) developed the Study Anxiety Scale as part of the Study Attitudes and Methods Survey (SAMS). The problem with the Study Anxiety Sc ale is that is does not fo llow the theoretical guidelines of the construct of study anxiet y in a number of ways. Four of the 16 items reflect test anxiety, as the responses focus on feelings duri ng or just before st arting an exam. Four items reflect trait or social a nxiety as they indicate situations that are more general or social rather than referring specifically to the time of studying. Three items refer to “not understanding” but they do not identify a nxiety as the reason for this lack of understanding and this may simply be assessing a lack of prior preparation as the cause of confusion. One item is clearly a depression item rather than an anxiety item and another item raises two points that are not mutually exclusive “I become so anxious over small points I encounter in studying and reading that I miss the really important points and main trends.” It may be that individual s may become anxious over small points while studying but not while reading and they may mi ss the main trends but not the really important points. Based on the definition of the construct presented by Lunsford (2001), only two items in the SAMS subscale clearly reflect the construct of study anxiety. Another instrument that includes a m easure of anxiety concerning learning designed for college students is called The Learning and Study Skills Inventory (LASSI;

PAGE 15

8 Weinstein & Palmer, 1988). The authors subseq uently created a simplified version for high school students called the LASSI-HS (Wei nstein & Palmer, 2002). The LASSI is a 76-item cross-curricular self-report measure w ith an anxiety scale assessing the degree to which students worry about their performance An example of an item reflecting test anxiety is “While I am taking a test, worry a bout doing poorly gets in the way of keeping my mind on the test.” Both the LASSI and LASSI-HS are 10-scale inventories that are widely used although, in the test manuals the authors have described these as components of three basic factors of skill, will, and self-regulati on. In a confirmatory factor analysis (CFA) of this instrument however, three other factors were supported relating to effort-related activ ities, goal orientation, and cognitive activit ies (Prevatt, Petscher, Proctor, Hurst, & Adams, 2006). This instrument did not fulfill the requirements needed to measure anxiety while studying. Study anxiety was also investigated in a paper by Owens and Newbegin (1997) where SA was defined as state anxiety and wa s measured by asking students to indicate how they felt at a particular moment (here, while studying). The measure used in that study was Spielberger’s State-A nxiety Scale (which measures current feeling) from the State-Trait Anxiety Inventory with state anxiety items being prefaced with the words “When I am studying…” This definition doe s not follow the theories presented by Spielberger who clearly states that Test Anxi ety is a situation-speci fic trait anxiety and should be measured by asking how the student generally feels. These measures of study anxiety clearly do not address study anxiety as defined. The need for an instrument to measure study anxiety is particularly pressing given the

PAGE 16

9 increasing emphasis on the use of tests to make various accountability decisions at all levels of education and the use of test scor es to inform accountability decisions. Learning to deal with anxiety during st udying and testing is not just im portant in the high school or college setting. This problem may affect the prog ress of people in their jobs for the rest of their lives because tests do not finish wh en formal education ends. Employers are increasing their use of tests given before th ey take on new employees as they have found that pre-employment tests improve corpor ate productivity if given under the right conditions (Rudner, 1992). Based on observations and interviews s upporting the distinctiveness of study anxiety from test anxiety and the potentiall y important role study anxiety may play in students’ test performance, Lunsford ( 2001) developed a paper-and-pencil self-report measure of study anxiety for use as a research tool to examine study anxiety in college students. Because the conceptua lization of study anxiety was ve ry similar to test anxiety with the main difference being the time anxi ety is felt, study anxiety was posited to be a situation-specific anxiety with the same wo rry and emotionality components found in test anxiety. A pool of 40 items was created by Lunsford to assess study anxiety and its possible components. To make the reading leve l of the measure sufficiently low to cover a wide range of students incl uding those whose first langua ge was not English, wording on the survey was established at a sixth grad e level using the Flesch-Kincaid Grade Level test. Because the Study Anxiety Inventory items were, in part, modeled after the Test Anxiety Inventory, it was consid ered that those responding to both sets of items might miss the general instructions to think about their thoughts and fee ling at the time of

PAGE 17

10 studying or at the time of a test and hence beli eve that they were being asked to respond to the same item twice. To avoid this pr oblem, a phrase indicating the specific context was woven into each item. For example, the TAI item read, “While taking examinations I have an uneasy, upset feeling” while the corresponding SAI item read, “While studying for exams, I have an uneasy, upset feeling.” That 40-item pool was presented to 12 expe rts in the field of psychology and test development, and items were evaluated for c ontent validity. All item s were printed on a form using a 5-point scale with instructions as king for items to be marked that seemed to reflect the construct of study anxiety as operationally defined. Four of the items were eliminated as the experts pointed out that they reflected content different from anxiety (e.g., distractibility) leaving 36 items that received the highe st percentage agreement. Lunsford (2001) conducted a series of studies to evaluate the psychometric properties of these 36 items. Through analysis of the factor loadings (i.e., pattern coefficients) in combination with the conceptu al fit with the defin ition of study anxiety, eight items were selected as indicators of th e emotionality aspect and eight items for the worry aspect of study anxiety for the Study A nxiety Inventory (SAI). Each item on this measure enabled the respondent to indicate intensity on a scale from one to four (1=Not at all, 2=Sometimes, 3=Often, 4=Always or almost always). Analysis of the 16-item Study Anxiety Inventory included item analysis, exploratory f actor analysis (EFA), and test-retest reliability. Results of the exploratory factor anal ysis with 536 college students supported the two-factor (emo tionality and worry) structure underlying study anxiety and provided evidence of the internal consistency reliability ( =.96 for the overall index, and

PAGE 18

11 .94 and .94 for the worry and emotionality subs cales, respectively), and two-week testretest reliability ( r =.84 for the overall scale and .84 and .84 for worry and emotionality subscales, respectively) of th e scores from the SAI. Since the development of Lunsford ’s Study Anxiety Inventory, several researchers have used the instrument a nd provided additional ev idence of validity. Because it may be argued that study and te st anxiety are similar constructs, Kieffer, Reese and Cronin (2005) carried out a study in which they used both Spielberger’s Test Anxiety Inventory and Lunsford’s Study Anxi ety Inventory in one administration. Using exploratory factor anal ysis with a varimax-rotated solution, all test anxiety items emerged on a single factor with the 16 study anxiety/ worry and study anxiety/emotionality items emerging on the remaining tw o but with two of the worry items loading on the study anxiety/emotionality factor. Keiffer, Reese and Cronin (2004) also conducted confirmatory factor analysis of the 32 items For the confirmatory factor analysis, five competing, falsifiable models were developed fo r the 32 items: 1) a si ngle factor, 2) four 8-item factors (test anxiet y/worry, test anxiety/emoti onality, study anxiety/worry, and study anxiety/emotionality), 3) two 16-item worry and emoti onality factors, 4) two 16item test anxiety and study anxi ety factors, and 5) one 16-item test anxiety factor and two study anxiety factors (found in a pilot explor atory factor analysis). Only the second model evidenced an acceptable model-to-d ata fit as reflecte d in goodness-of-fit and adjusted goodness-of-fit (both above .83), comparative fit i ndex and Normed Fit Indexes (both above .90) and the Root Mean Squa re Error of Approximation (RMSEA=.066). The fit of the study anxiety items demonstrated a better fit than th e test anxiety items.

PAGE 19

12 The four subscales, Test Anxiety Worry (TA/W), Test Anxiety Emotionality (TA/E), Study Anxiety Worry (SA/W), and Study Anxiety Emoti onality (SA/E) had Cronbach alphas of .92, .93, .92 and .94 respectivel y with item to tota l correlations all greater than .68. Using 180 stude nts attending an e ffective study habits course, 10-week test-retest reliability coefficients were .73, .78, .67 and .81 for the same order of subscales mentioned above. Although these results provided initial vali dation of the scores from the SAI, validation is an ongoing pr ocess that is streng t hened through the collection of multiple sources of evidence. Because the CFA was carried out by Keiffer, Reese and Cronin (2004) on the 32 test and study anxiety items combined, their study does not provide an exact test of the measuremen t model underlying the SAI. Th erefore one of the purposes of the current study was to investigate further the late nt structure of the 16 SAI items using confirmatory factor analysis. In ps ychological assessment literature the most popular method for providing empirical support of construct validity is confirmatory factor analysis (AERA, APA, & NCME, 1999; Thompson & Daniel, 1996). When the measure of a construct has been developed us ing a theory, CFA is used to evaluate the latent structure behind the measure (Byrne, 1998; Hoyle & Panter, 1995). Stevens (1996) explains that exploratory fact or analysis is used to iden tify how many factors underlie a set of observed variables. It is considered to be a method of generating a theory rather than testing a theory-based instrument. While EFA was used in the development stages of the SAI, it was used primarily to enable th e elimination of item s that obtained poor loadings so that the remaining items woul d more clearly define the factors already

PAGE 20

13 established by theory. The next logical step in the construct validation process was to evaluate the two-factor structure (worry, emotionality) underlying the SAI using confirmatory factor analysis (CFA). Research has consistently found that self-re port scores of anxiety for females is higher than for males (Hewitt & Norton, 1993; Sp ielberger, 1975; Spielberger & Wasala, 1995) although there is little re search suggesting the reasons for this beyond a biological propensity towards anxiety. General anxiety is suffered by women about twice as much as men (Breslau, Schultz, & Peterson, 1995) wh ich is something that begins to show around puberty (Seeman, 1997) while prior to pub erty, males are more susceptible to anxiety. Reproductive hormones and cyclical hormonal patterns ar e therefore clearly important in the prevalence of anxiety as it relates to gender. Th ere is support of the evolutionary theory that pred icts that no differences would exist where the same adaptive problems have been faced but would exis t where problems have differed. It would therefore make sense that these differences would appear at the time of puberty if differences in anxiety have to do with sexua l selection (changes due to advantage in reproduction). The male pursues higher risk st rategies and, because he has a lower level of parental investment, therefor e develops a propensity for lo wer anxiety. This is not to say that social factors like sex roles, differe nces in economic power, perception of threat, or the impact of sexual selection should be ignored but these would be secondary factors to the biological ones. It has also been documented that there are sex differences in neurotransmitter and neuromodulatory systems th at are associated w ith anxiety (Carlsson & Carlsson, 1988; Wilson & Biscardi, 1994).

PAGE 21

14 Armstrong and Khawaja (2002) compared responses across gender and found that females considered the manifestations of th eir anxious worries to be more catastrophic and more dangerous. Females also reported mo re concern about emotional, physical and mental symptoms related to anxiety. Effect sizes, however, are typically moderate to low according to studies reported over the last decad e that have compared trait anxiety scores for males and females (Everson, Millsap, & Rodriguez, 1991; Foot & Koszycki, 2004; Marcus, 2001). Mean differences between male and female respondents, then, have been fairly well established, but the factor struct ure underlying the measures of anxiety have not been compared to determine whether ma les and females view the meaning of the items in the SAI in a similar manner. Ther efore, a second purpose of the study was to determine whether there was f actorial invariance of the SA I by gender. Invariance testing involves comparing the factor pattern coe fficients (loadings), uniquenesses (error variances), and factor varian ces and covariances across th e male and female groups. An equally important purpose for this st udy was to evaluate the validity of the SAI scores using the logic of the nomologi cal network proposed by Cronbach and Meehl (1955). Construct validation using this fram ework (AERA et al., 1999) involves carrying out tests of the relationship between study anxiet y and the related latent variables of test anxiety, trait anxiety, trait cu riosity, and procrastination. B ecause test anxiety has been researched extensively for a number of years, it has a fairly well established nomological network and since the SAI was developed using this construct as a model, it is to be expected that there would be a number of cons tructs that would also correlate with the

PAGE 22

15 SAI. To continue the ongoing validation pro cess, correlations were investigated using those existing parallels. The way that test anxiety is similar to st udy anxiety is that for both: (a) failure on a specific exam is a perceived threat, (b) the thre at is perceptual rather than actual, and (c) the nature of the response is worry and em otionality. The similarity of the antecedent suggests that study anxiety will correlate highl y with test anxiety. The similarity of perception of threat suggests that study anxiety will correlate highly with trait anxiety. The similarity of worry and emotionality responses suggests th at study anxiety will correlate positively with anxiety measures. The way that test anxiety is dissimilar to study anxiety is that: (a) the time of the perceived threat is while st udying instead of while taking the test, (b) the perceived control over what can be done about the unpl easant feelings is located within the individual rather than the test proctor, a nd (c) the difficulties faced by the individual with high study anxiety are in his/her ability to en code and retain information instead of in his/her ability to retrieve stored information. The dissim ilarity of the time of the perceived threat suggests that study anxiety sc ores will not correlate so highly with test anxiety that it should be considered the same construct. Individuals could have high study anxiety yet become calm as they walk into an exam realizing that there is nothing further they can do, or they may be unaware that the test will be as hard as it turns out to be, or that they are not as prepared as they should be so they could be calm while studying but feel high anxiety at th e time of the test.

PAGE 23

16 Curiosity is a motivational instinct desc ribed as the tendency to investigate a stimulus. According to the Optimal Stimul ation/Dual Process Theory presented by Spielberger and Starr (1994), the level of cu riosity a person has will change with the intensity of the stimulus. At low stimulus intensity, curiosity will be at a level that motivates exploration. At m oderate stimulus intensity, both curiosity and “mild-tomoderate anxiety” (p. 233) will be the motivating instincts. At high stimulus intensity, high levels of anxiety will cause avoidance behavior and as the intensity increases curiosity will decrease and anxiety will increase (Spielberger & Starr, 1994). Berlyn (1960) posits that the relationship between level of curiosity and level of anxiety is partly due to personality characteristics. Ther efore, different personalities will respond differently to different intensities of stimuli. Those with high trait anxiety will more quickly respond to a stimulus with state anxi ety and less curiosity while those with low trait anxiety will respond with less state anxiety and more curiosity. Each person will have a different optimal arousal level. Based on this theory, curiosity would be inhibited by anxiety. Thus, curiosity is predicted to corr elate negatively with anxiety; this result was found with trait curiosity and study a nxiety in a previous study (Lunsford, 2001). A number of researchers ha ve stated that people may avoid performing a task to avoid uncomfortable feelings of anxiety. At kinson (1974) suggests th at those who avoid failure tend to be more anxious about failing and will hence avoid tasks that will bring on that anxiety. Beswick, Rothblum and Mann (198 8) showed that as anxiety and low selfesteem increase, procrastination goes up, a nd grades go down. However, Ferrari, Johnson and McCown (1995) suggest that this relati onship is not so simple. Although Soloman

PAGE 24

17 and Rothblum (1984) defined pr ocrastination as needless dela y of tasks “to the point of experiencing subjective discomfort” (p. 503), th e measures of procras tination suggest that procrastination is a broader c onstruct than this. Chu and Choi (2005) suggested that there are two major types of proc rastination: passive and ac tive. The passive type is procrastination to avoid an unpleasant or anxiety-provoking task. The active type, is procrastination to increas e optimum performance by tim ing the event to cause appropriate pressure for purposeful use of time. They explain that an active procrastinator is more like a non-procrastinator than a passive procrastinat or. This would suggest that study anxiety would have a moderate positive correlation with passive procrastination while the relationship would be lower and positive with active procrastination. Chu and Choi (2005) suggest that activ e procrastinators ar e less like passive procrastinators than they are like non-procrastinators in th eir relationship with anxiety. This rethinking of procrastin ation as a two-factor constr uct with opposing relationships with situation-specific a nxiety points to a possible reason th at past studies that have used one factor measures of proc rastination have shown low or no relationships between the constructs of test anxiety and procras tination (Milgram & Tenne, 2000). Ackerman and Gross (2005) found that there wa s no relationship between pr ocrastination and fear or pressure to meet a deadline. Lee, Kelly and Edwards (2005) only found a moderate correlation in a study looking at the relationship between procra stination and neuroticism. Onwuegbuzie (2000), however, pointed out th at among the variables he studied, anxiety was a factor related to student s avoiding enrolling in statistic s classes as long as possible and tending to procrastinate on their assignments. It is necessary to look at these

PAGE 25

18 relationships as they pertai n to study anxiety and to whet her the relationships between these types of procrastinati on differ depending on the type of study anxiety, worry or emotionality. Purpose The purpose of this study was to extend research on the cons truct validity of responses from college students to the Study Anxiety Inventory (SAI). Several approaches were used. Because the SAI was de veloped using a theory that the construct consisted of two highly correlated factors, s upport for this two-factor model was needed in establishing factorial validit y. Confirmatory factor analysis was used to test the two factor model (worry and emotionality). As part of the CFA, the factorial invariance of the SAI for males and females also was examined. Additional evidence for construct validit y was collected using the nomological network framework. Based on the theoreti cal framework of study anxiety, it was predicted that there would be a positive re lationship between scor es on the Study Anxiety Inventory and scores on a measure of passive procrastination and active procrastination. Cronbach and Meehl (1955) argued that to suppo rt the validity of a construct, the test developer must show that the responses can be interpreted with specified hypothesized meaning; relationships between the constr uct and different or similar theoretical constructs or behaviors should be stated (nomological network). Cronbach and Meehl also explained that, although during the earl y stages of development the network will have few interrelations, more will be lear ned about a construct by “elaborating the nomological network” (p. 290). Construct validity is supported as the nomological

PAGE 26

19 network is enriched with observable behavior s like responses to related and unrelated measures that appropriately correlat e with the constr uct in question. In the previous study using these meas ures (Lunsford, 2001), correlations were computed between study anxiet y, test anxiety, trait anxiety, trait depression, trait anger, trait curiosity, and st udy skills and habits. Given the rela tionships that were reported between these constructs in earlier studies and using the Optimal Stimulation/Dual Process Theory discussed earlier, it was hypothesized that study anxiety and the two components of study anxiety, worry and emoti onality, would have a positive correlation with test anxiety (overall, worry and em otionality) and trait anxiety. A negative relationship was predicted be tween study anxiety and trait curiosity. A weak positive relationship between study anxiety and study skills a nd habits was hypothesized. Based on the findings of Choi and Chu (2005) it was expected that the SAI total, worry, and emotionality scores would correlate positively with the passive procrastination scores and with the active procrastination scores. In summary, the purpose of this study was to evaluate the construct validity of the scores from the SAI with evidence obtained by: 1. Evaluating the two-factor measurement model underlying the Study Anxiety Inventory in a sample of college students; 2. Evaluating the factorial invariance of the two-factor measurement model underlying the Study Anxiety I nventory across male and female college students;

PAGE 27

20 3. Examining the relationship between the scores on the Study Anxiety Inventory and scores on two measur es of procrastination (active procrastination and passive procras tination), a measure of study skills and habits, and two of the four tra it personality measures of the StateTrait Personality Inventory, the Tr ait-Anxiety Scale (T-Anx) and the Trait-Curiosity Scale (T-CY). Significance Additional evidence supporting the valid ation of the responses to the SAI provides the users with greater confidence in employing the SAI for research. Support for the factorial validity of the Study Anxiety Inventory gives confidence to researchers that this measure can be used to continue i nvestigating the constr uct of study anxiety. Research can then be carried out to determ ine the effect of st udy anxiety on encoding information for college students. To estab lish measurement invariance by gender for the responses to the SAI would give researcher s confidence in comparing differences in means between males and females. Findings of relationships of study anxiety with the two types of procrastination will exte nd the understanding of the construct. Although at this stage the SAI is intended as a research t ool, it is conceivable that treatments might also be developed for study an xiety in the same way that they have been developed for test anxiety, and counselors may start using the measure for assessments that could guide treatment or for making decisi ons regarding the type of help that might be given to a student.

PAGE 28

21 Limitations This study was carried out at only one st ate university in Florida and the sample was not randomly selected but was a samp le of convenience made up of students who elected to attend social sciences statistics classes or certain education, business, and engineering classes. The implications of study anxiety may reach outside the population of university students, so us ing only university students is a limitation of this study which may be dealt with in future studies. A second limitation of this study was that it measured at one point of time students being mostly ar ound the age of 21. Another limitation of this study is the use of paper-and-pencil self-repo rt methods which tend to raise concerns about the validity of any causal conclusions th at may be made from their use because of social desirability, response-set bias, or measurement error (Graziano & Raulin, 2007; Razavi, 2001). Although every effort was made to simplify the language of each item, because of the nature of the measure, there is a potential that a participant might not understand the wording of an item and w ould respond with guesses. Finally, the participant may either be unaware of thei r anxious responses or deny they had them.

PAGE 29

22 Chapter 2 Review of the Literature The purpose of this study was to collect va rious types of eviden ce to evaluate the validity of the inferences derived from th e Study Anxiety Inventory (SAI). Given this focus, the present review begins by overview ing the validation proce ss and the traditional types of evidence that are collected as part of this process. Th ese types of evidence involve an analysis of item content (content validity), intern al structure of the responses to the items (exploratory and confirmatory fact or analysis), and rela tionships between the construct and other variable s (concurrent, predictive, and construct validity). Following this overview of the validati on process, the theoretical work and research studies focusing on the construct of st udy anxiety are examined as they relate to the development and validation of the scor es from the SAI. Articles that focus on university students, the intended audience of th e SAI, are included; articles that focus on other populations are not include d unless the information is relevant to the construct validation process. Overview of the Validation Process The Standards for Educational and Psychological Testing (American Educational Research Association, American Psychol ogical Association, & National Council on Measurement and Education, 1985) emphasizes th at validity is “the most fundamental consideration in developing and evaluating te sts” (p. 9). According to Messick (1989), validity is the degree to which evidence co mbined with rationales based in theory “support the adequacy and appropriateness” (p. 13) of inferences made from the scores

PAGE 30

23 obtained from a test. Various types of evidence may be used to support the validity of test scores. Evidence based on instrument content. After a construct has been conceptualized, a way must be established to measure it. For a construct like study anxiety, a common measurement approach is to present written st atements (items) that are indicators of the construct for the respondent to rate him or he rself. Items are designe d to elicit responses that are theoretically aligned with the conceptualization of the construct. These items and instructions for completing them should th en undergo content analysis, which is an examination by experts of both item developm ent and the construct being operationally defined to determine if the sample of items re presents the construct of interest (Cronbach, 1949). Completion of the content validation proc ess clears the way fo r further evaluation of the measure. The next stage is for the instrument to be completed by a number of participants and a statistical analysis of the responses car ried out to determine that the item scores have a reasonable level of reliability. M easures of internal consistency, such as Cronbach’s alpha, are frequently used to assess reliability. Reliability may also be viewed in terms of the consistency of scores over time This type of consistency or stability may be assessed using te st-retest reliability. Once it has been established that the res ponses to the items on the measure relate to one another sufficiently and that a reasonable degree of stab ility exists in the responses from one time to another then it is important to establish a level of construct validity. Although construct validity is a unitary concept, it is mo re difficult to a ssess the validity

PAGE 31

24 of inferences made from the scores of a m easure than to assess its reliability because validity requires that a rational argument be given as to how a construct should be measured and then empirical evidence must be gathered to support that argument. Special types of evidence may be collected including evidence based on the internal structure of item responses and evidence based on relations to other variables. Evidence based on the internal structure of item responses Two types of factor analysis are used commonly to evaluate the internal structure of item responses, exploratory (EFA) and confirmatory factor anal ysis (CFA). Factor analysis identifies the way related items cluster and enables evaluati on of the dimensionality underlying a set of scores. Because the SAI was conceived as an in strument with two factors rather than one, the CFA should support the multidimensionality of the responses. The exploratory and confirmatory factor analyses that have been used with the SAI are presented later in this chapter as a rationale for including an examin ation of factorial invariance between males and females. Evidence based on relations to other variables. In a landmark paper by Cronbach and Meehl (1955), construct valid ity was given more clarificat ion and the concept of the nomological net was introduced as a framework for providing evidence of the validity of psychological constructs. Construc t validity of the scores is more than one coefficient. It is an ongoing process involving examination of the relationships that should theoretically exist between a scale score and other variab les. For the present study, it was hypothesized that the SAI scores should be positively corr elated with scores on measures of test and trait anxiety.

PAGE 32

25 Campbell and Fiske (1959), motivated by the fact that, in many areas, there is not a “Gold Standard” criterion with which to compare new measures, introduced the multitrait-multimethod approach which evalua tes construct validity of a measure by investigating relationships f ound through correlations between two or more traits, each assessed by two or more methods. They introduced convergent validity as high correlations between measures of the same construct assessed by different methods and divergent or discriminant valid ity as low correlations between constructs that should not relate to one another. Even though there have been some objections to this approach, the approach is widely accepted as adding to th e empirical evidence of the construct validity of the scores of a measure. Examination of the scores from the SA I with other theoretically meaningful constructs and observable attributes (e.g., gender) provides deeper insight into the construct validity of this inst rument. The following sections d eal with the development of the SAI and go through each step in the valid ation process as it relates to the SAI.

PAGE 33

26 Development and Initial Validati on of the Study Anxiety Inventory Because the first step in the development of an instrument is detailed analysis of the construct being measured, the following sections present the theoretical background of anxiety and study anxiety. This background provides the framework for the creation of the items used to measure the construct of study anxiety and the appr oaches used in the construct validation process. It is important to understand the nature of anxiety in or der to be able to measure anxiety in a specific and meaningful way. Becau se the number of articles on this topic is large, this chapter will briefly touch upon the highlights of the gene ral topic of anxiety and situation specific anxiety, but will include appr opriate articles on the constructs of trait anxiety, anxiety at the time of studying or testing, study skills, and procrastination. Conceptualization of Study Anxiety Anxiety as a typical respons e to fear is a central pr oblem in our society. Rollo May, in his book, The Meaning of Anxiety, described the significant impact of anxiety in the arts, in the social sciences, and in so ciety (May, 1950). Many exam ples of this impact can be found in popular literature and newspa per articles and incl ude everything from fears of sexual predators (e .g., Hong, 2007) to concern about the amount of coffee being drunk by young people (e.g., Fiely, 2007). There are also many articles about causes of anxiety or how it may be overcome (e .g., Roysdon, 2006). The most common mental health disorders are the anxiety disorders according to Mental Help Net for Anxiety Disorders, accounting for close to half a billion dollars in healthcare costs each year (Anxiety Disorder, 2001).

PAGE 34

27 Darwin (1872) recognized the importan ce of fear and considered that it had evolved as an adaptive response in both anim als and humans with the purpose of arousing the motivation to cope with some danger. He reported those signs of fear (e.g., racing heart, perspiring, etc.) that were fairly easy to observe, whereas othe r researchers in later years reported less obvious chemical changes (Pitts, 1971). Darwin also suggested that fear is an adaptive signal of danger so the organism may escape or fight the feared stressor. It is interesting that Darwin ma de the observation that these responses might lead to disaster if too little or too much fear is elicited such that the individual’s behavior might attack foolishly or be overcome by exces sive fear responses. Creatures with these overor under-reactions would, according to th e theory of evolution, be less likely to survive and be less likely to continue c ontributing to the gene pool. Those with appropriate reactions woul d survive to reproduce. Freud (1936) was more interested in anxi ety as experienced feelings – the state characterized by unpleasant feelin gs of apprehension. Freud’s vi ew was that the presence of these feelings served as a warning that action was needed to avoid or eliminate a stressor. He considered that, as well as feelings of a pprehension, the experience of anxiety includes tension and thoughts of worry. He also pointed out symptoms of anxiety to be increased heartbeat, increased breat hing rate, shaking and possibly nausea or dizziness (Freud, 1936). His work agreed with Da rwin’s view on anxiety as a response to the presence of real danger but diverged when he introduced the idea of an xiety being a response to danger that was ne ither present nor imminent.

PAGE 35

28 According to Walter Cannon (1929), duri ng an emergency reaction, blood is redistributed to the body areas that will be active so that the en ergy supplies will reach the critical muscles and organs while the en ergy used for digestion can be sacrificed. Thus "fight-or-flight response" is an adaptiv e response occurring when energy is needed for that purpose. The responses that make up this reaction were considered to be mediated by part of the autonomic nervous sy stem called the sympathetic nervous system (Bernstein, Roy, Srull, & Wickens, 1988). Pavlov also investigated fear and anxiety but studies could only be done on animals and ha d to be done without a reliable measure of anxiety (Kalechstein, Hocevar, Zimmer, & Kl echstein, 1989). Othe r researchers had differing theories. Heinrich Neumann (1814 188 4) spoke of unsatisfied drives, as the cause of "anxiety" and Karl Ideler (1795 1860) suggested unfulfilled sexual longings as being important in the cause of nervous disorders (Stone, 1996). Anxiety is complicated because in differe nt contexts it means different things. Many think of it as a mood state, having to do with emotions or physical symptoms, while others discuss its cognitive aspects. The following information is included to clarify the current views on the meaning of a nxiety and to examine cr itically the state of the field. Types of anxiety. The Diagnostic and Statis tical Manual of Mental Disorders 4th Edition (DSM-IV), a publication primarily deali ng with disorders, presents anxiety as being of differing extreme types: Panic Di sorder, Agoraphobia, Specific Phobias, Social Phobias, Post-Traumatic Stress Disorder Obsessive-Compulsive Disorder, and Generalized Anxiety Disorder. Test anxiety is mentioned und er the category of Social

PAGE 36

29 Phobia (300.23) as a symptom or associated de scriptive feature, and is cited as a reason that sufferers of social phobia may perf orm poorly in school (American Psychiatric Association, 1994). It is one of the types of anxiety that is more generally accepted to be part of every person rather than a malady th at should be considered severe enough to warrant a classification. The term State Anxiety (S-Anx) is most ofte n used to describe an existing state or feeling of fear of impending danger while th e term Trait Anxiety (T -Anx) refers to the overall tendency towards such feelings that remain stab le over time and situations (Spielberger, 1975). These two constructs ma y vary in intensity and often influence individuals differently in their reactions to stress. Those low in T-Anx will experience SAnx less often than those high in T-Anx. Test anxiety falls under the umbrella of trait anxiety but is referred to as a situation specific trait anxiety as it occurs at a specific time and has to do with a situation in which the person experiencing it must vi ew the test as a fo rm of evaluation and therefore a threat to some so cial standing. There would, for example, be no perception of threat by college students if given a test of first grade mathematics because they know that they would not fail to obtain a high eval uation, while threat w ould be perceived if given a test on college mathematics as they may fail. Situation specific traits are most commonly measured using questionnaires, which have provided evidence in over 2400 studies conducted on test anxi ety since 1966 at a rate of 200400 every five years for 40 years (see Figure 1).

PAGE 37

30 Figure 1. Published Articles Found on Ps ycInfo Concerning Test Anxiety. Measures of test anxiety are theory-based with the information processing model (Figure 2) being the most influential in gui ding the development of the most popular of these instruments. According to this model, when a student is cued with a question in a test situation, he/she perceive s it and makes an appraisal of its threat to his/her position and goes through information processing a nd retrieval to answ er the question. Figure 2. Information Processing Model. Test Ques ti o n Perception Appraisal & Rea pp raisal Information Processing and Retrieval Answer Tas k -Relevan t

PAGE 38

31 When a person appraises a question as a th reat, anxiety in the form of worry and emotionality interferes with information processing and retrieval (see Figure 3). Figure 3. Lazarus’s Transactional Process Theory. The information processing model in Figure 2 and Lazarus’s Transactional Process Theory assume encoding but Zeidne r (1998) suggests that those suffering from high test anxiousness may have more difficulty encoding information than those low in this trait. This implies that the model need s to be increased in its scope to include encoding of information. If the information to be learned has been presented in an acceptable fashion, it may be anxiety while atte mpting to learn the information that stops a student from encoding the information in th e first place, therefore giving rise to a metacognitive awareness that the material ha s not been committed to long term memory. It would seem evident that the techniques that should be taught to improve learning would be those that would help someone su ffering from anxiety felt when attempting to learn – study anxiety (SA). Fi gure 4 represents the expande d model that includes study anxiety. Test Ques ti o n Perception Appraisal & Rea pp raisal Information Processing and Retrieval Answer Tas k -Relevan t Test Emotionalit y Test Worry Distraction/ Task Irrelevant Behavior Study Habits Deficit & Testtaking Skills Deficit

PAGE 39

32 Figure 4. A Suggested Expanded Model Showing St udy and Test Worry and Emotionality. This model distinguishes between test anxi ety -that begins when the student is given a question in a test situ ation -and study anxi ety -the anxiety fe lt during the time that the student is studying for an exam. Th ere may be those who feel anxiety in both situations but some who don’t feel anxiety un til they enter the exam room. Still there may be others who feel anxiety up to the moment they walk into the room but calm down at that point believing there is nothing more they can do. The model suggests that study skills, worry, and emotionality determine th e level of encoding into long term memory (LTM) while a student is studying for a test and that test-taki ng skills, worry, and emotionality determine the level of interference with retrieval during a test. The earliest mention of the effects of anxiety during studying was by Spielberger (1966) who discussed research initiated in 1955 on students who complained that their anxiety increased around exam time. Anxi ety concerning performance was either the salient symptom or an important background f actor. These students i ndicated that their ability to absorb information was being affected by the anxiet y they felt while studying. Question Perception Appraisal &Rea pp raisal Information Processing & Retrieval Answer TaskRelevan t Distraction/ Task Irrelevant Behavio r Information Perception Appraisal & Reappraisal Information Processing and Encoding to Long Term Memory Test Emotionalit y Test Worry Study Skills Deficit Test-taking Skills Deficit Distraction/ Task Irrelevant Behavio r Study Emotionalit y Study Worry

PAGE 40

33 When Owens and Newbegin (1997) attempted to examine this concept using Spielberger’s State Anxiety Scale from the St ate Trait Anxiety Invent ory they did not use an approach that was in line with the situa tion-specific basis presented by Spielberger in the development of the Test Anxiety Inventory. They asked participants to complete the State Anxiety Scale while studying when the construct concerns general traits while studying for an exam. Not surprisingly, the re lationships found using this approach were in line with those obtained us ing the State Anxiety Scale without the extra instructions. State Anxiety Scale scores we re not significantly correlated with grades of high school students aged 12 to 16 years. Other cognitive psychologists have examined the mechanism through which anxiety exerts influence on mental functi ons including attention, memory, levels of processing and retrieval. Tobi as (1985) reproduced a resear ch model of the effects of anxiety on learning from instruc tion, originally presented in Anxiety, Learning and Instruction, as early as 1977. He suggested that when cognitive resources are taken by anxiety, the resources to st udy would be lacking. The con cept that anxiety hinders encoding was also introduced by Eysenck ( 1991) who indicated that a considerable amount of evidence shows that anxiety level an d the functioning of the attentional system are related and that the effect s of anxiety are an increased susceptibility to distraction (Eysenck, 1979; Wachtel, 1967). Eysenck, MacLeod and Mathews (1987) showed that anxious individuals are more di stracted by threatening distract ers, which in this context could refer to consequences of failure to learn. Williams, Watts, MacLeod and Mathews

PAGE 41

34 (1988) showed that anxiety affects the pa ssive, automatic aspect s of encoding, thus affecting pre-attention and attentional processes more frequently than memory. Various authors have noted that the a nxiety may occur as many as four days before an exam (Bolger, 1990; Lay, Edwa rds, Parker, & Endler, 1989). Covington and Omelich (1985) suggested that task-irrelev ant worry about ability interferes with effective information processing. Their resear ch also suggests that for people who are perfectionistic, anxiety discourages deep-l evel processing duri ng original learning. Eysenck, MacLeod and Mathews (1987) showed that threat (i.e., appraisal of negative consequences) causes more distraction for high anxious than for low anxious individuals. Williams, Watts, MacLeod and Mathews (1988) showed that, rather than directly affecting memory encoding, anxiety affects the attention and pre-atte ntive processes that are automatic. One important study on trait an xious students suggested that hypervigilant students responded more to stimuli they percei ved as threatening and focused on any task irrelevant stimuli presented (Eysenck & By rne, 1992). Zeidner (1998) suggested that denial, wishful thinking, and avoidance may disrupt studying. Various authors have noted that the anxiety sometimes occurs days be fore an exam (Bolger, 1990; Lay, Edwards, Parker, & Endler, 1989). Naveh-Benjamin (1991) suggested that ther e are different types of test anxiety sufferers with some having poor study hab its and others having good study habits. Naveh-Benjamin also posited that there are some individuals who would benefit most by interventions that help them encode and organi ze as they study. It has even been shown in studies on rats being trained to run a maze th at stress produced by exposure to a cat for 30

PAGE 42

35 minutes each day results in an absence of de ndritic spine density that indicates encoding of information into long term memory (LTM). Further results indicate d that stress before the training started blocked information from getting into LTM, and stress before the retrieval test blocked access to stored memory (Diamond, Park, Heman, & Rose, 1999). Since anxiety is the result of appraisal of threat, one might ask what causes that appraisal of threat. One could answer this question from the Cognitive, Rational-Emotive therapy angle by saying that the individual’s appraisal is flawed, causing worry. One could ignore the reason for the appraisal a nd approach a solution from the Systematic Desensitization, Relaxation, or Biofeedback trai ning angle, which attempts to deal with the emotionality of the individual. Bette r results may come from combining these approaches but, although studies using them have demonstrated a decrease of anxiety during tests, no consistent improvements in performance as measured by GPA have resulted. The general conclusion is that pe rformance on tests is not improved by merely decreasing test anxiety (Vagg & Papsdorf, 1995) A possible reason for this may lie in the metacognition of the individual that he/she ha s not learned the material. Is this because he/she did not try hard enough? The evidence su ggests that there is not a uniform answer to this question. There are, of course, those whose anxiety prior to the exam causes them to accept failure and therefore do not study and as the time of the exam gets closer, the anxiety increases (Covington & Omelich, 1985). There are also those who procrastinate excessively, delaying studying due, in part to this anxiety (Kalechstein, Hocevar, Zimmer, & Kalechstein, 1989). Anxiety level is not a reflection of intelligence, as anxiety does not discriminate between th e more or less intelligent.

PAGE 43

36 Internal Structure: Assessment of State, Trait, and Situation Specific Anxiety Because autonomic nervous system respons es are difficult to control voluntarily, and therefore would not be influenced by faki ng, defensiveness, and social desirability, the initial way to measure anxiety was with physiological measures. This approach gave way to self-report which is now the more commonly used approach because researchers using respiration, heart rate, galvanic skin response, blood pressure, pulse pressure, and oral and skin temperatures found that resu lts using these physiological measures were disappointing (Hopkins & Chambers, 1966; Levi tt, 1967). They concl uded that: (a) these physiological measures were unrelated and did not provide a basis for identifying specific anxiety, (b) each person responded differently, and (c) the measures did not relate to test scores obtained under different treatments. Self-report measures, how ever, did correlate moderately with performance and were ab le to tap components of anxiety that physiological measures did not assess such as worry or perception of severity of anxiety. Additionally, self-re port inventories have been f ound to be acceptably reliable while physiological symptoms have been found to be present when a person does not feel anxiety and not to be present when a pe rson does feel anxiety (Spielberger, 1975). State and Trait Anxiety. The Taylor Manifest Anxiet y Scale (MAS), a measure based on the idea that the level of anxiety is an indicator of emotionality and motivation or drive, was the first objective measure of anxiety to be published (Taylor, 1953). Spielberger, using the Liebert and Morris ( 1967) concept that there are two components of anxiety, worry and emotionali ty, and realizing that there was also a need to measure anxiety states as well as gene ral tendencies, developed the State-Trait Anxiety Inventory

PAGE 44

37 (STAI; Spielberger, Gorsuch, & Lushene, 1970) In research over the past 50 years the STAI has become one of the most wide ly used measures for assessing anxiety. Spielberger (1966) introduced the constructs of state and trait anxiety and created the STAI with two self-report 20-item scales inte nded to provide brief but reliable measures of a person’s current and gene ral level of anxiety (Spielbe rger, Gorsuch, & Lushene, 1970). The best 10 items from each of the scales from this measure have been included in the anxiety scales of Spielber ger’s State-Trait Personality I nventory (STPI; Spielberger et al., 1979). As was the design of these measur es, the state anxiety (S-Anx) scale item responses reflect the feelings of the particip ant at the time the measure is administered while the trait anxiety (T-Anx) scale items and test-retest reli ability show that responses are stable over time and in di fferent administration situations. Test anxiety. The Test Anxiety Inventory (TA I) (Spielberger, 1980) is the most popular measure of the construct of test anxiet y and has been used in thousands of studies published in scholarly articles. In 1990, Ware, Galassi and Dew used the responses from a sample of 752 college students in a confirmatory factor analysis to investigate the factor structure of the TAI. They compared a 2-fact or oblique model with a 2-factor orthogonal model, and both a null and single-factor model. The oblique solution gave the best fit, giving support to the theory that the construc t contains two correlate d factors (worry and emotionality), although the question of the neces sity for more than 16 items was raised. Based on these findings and analys is of items by Spielberger, the author of the measure, four items were removed, leaving the 16 best items. The most current version of the TAI uses these 16 items (see the TS AI measure in Appendix B).

PAGE 45

38 Everson, Millsap and Rodriguez (1991) conducted a study using 501 undergraduates to investigate th e factor structure and factor invariance across gender of the TAI. Although females generally reported a higher level of test anxiety, factor invariance across gender was supported which su ggests that although th e meanings of the items are similar for males and females, the le vel of test anxiety is higher for females. Study anxiety. The Study Anxiety Inventory (S AI; Lunsford, 2001) was developed as a research tool to examin e the construct of study anxiety. The SAI was posited to have two scales reflecting worry and emotionality. Study anxiety is defined as a situationspecific personality trait of a nxiety felt while a person is study ing for an exam. A sufferer would experience both worry and emotionality while studying for exams. Worry cognitions while studying for an exam would in clude: not being able to organize material mentally, getting a mental block to absorbing material, worrying to the point of engaging in distracting behaviors, wo rrying about being capable of learning material, and being unable to keep focused on the subject. Emoti onality while studying for an exam would include feeling uneasy, panicky, upset, jittery, or nervous. Theoretica lly this construct and its components should correlate highly with test anxiety, and le ss highly with other measures of personality such as anger a nd curiosity. The information processing model also suggests that study anxiety should ha ve significant relationships with academic achievement. Prior to 2001, the only measure purporti ng to measure the construct of study anxiety was a scale in the Study Attitudes and Methods Survey (SAMS) developed by Welsh, Bachelor, and Wright (1990). This scal e was limited in that only two items in the

PAGE 46

39 SAMS subscale clearly reflect the construc t of study anxiety. Given this limitation, Lunsford developed the Study A nxiety Inventory. Since the Te st Anxiety Inventory has a great deal of support for the validity of the scores and has been factor analyzed (Spielberger, 1980) with both e xploratory and confirmatory factor analysis, it seemed prudent to start the developm ent of the SAI by considering the items used in the Test Anxiety Inventory (e.g., “During tests I feel very tense”, “D uring examinations I get so nervous that I forget facts I really know”). There are 20 item s on this measure, of which 16 are associated with the subscales of worry and emotionality. Each item included words that approximated “While taking a test.” In developing the initial pool of items for the Study Anxiety Inventory, an effort was made to include an approximately equal number of worry and emotionality items. As both of th ese factors had been shown to be present in the Test Anxiety Inventory, it was assumed that the same factors would be established in the Study Anxiety Inventory. Th e items were selected by adapting items from the Test Anxiety Inventory to create new items that were approximately equal in meaning except they specifically targeted th e time period of studying rather than the time period during test-taking, and they also sp ecified that the studying was for an upcoming test (e.g., the words "taking a test" being repl aced with "studying for a te st”). The College Adjustment Scales (Anton, 1991) were also found to have a number of the items that suggested difficulties in studying except they did not specify that the studying should be for an exam. These items from the College Adjustment Scales were adapted for use in the SAI by adding that component. This increased the number of items in the SAI to 30. Finally, discussions were held with a person who su ffers from the symptoms of anxiety while

PAGE 47

40 studying and an additional 10 items that dea lt with specific symptoms like difficulty breathing were developed resul ting in a total of 40 items. Each item on the SAI was worded such th at it contained the time element “while studying for an exam” along with a cognitive or emotional symptom. Then, in each item, either the word “feel” was used or strongl y implied to tap into the emotionality component or the word “think” was used or strongly implied to tap into the worry component. Using the Flesch-Kincaid Grade Le vel test, the reading level of the measure was determined to be at a sixth grade level. In preparation for the items to be rated by a team of experts, all items were printed on a form with a 5-point scale (1=unsuitable to 5=suitable). Instructions asked for items to be marked as suitable that seemed to re flect the construct of study anxiety as defined. A clinical psychology professo r, 15 clinical psychology graduate students, and Dr. Spielberger, the author of the Test Anxi ety Inventory, completed the form and made comments that suggested that four of the ite ms indicated content different from anxiety (e.g., distractibility) leaving 36 items that were viewed as suitable by the majority of the reviewers. Once this pool of items had been evalua ted and found to be acceptable, the test form was created for completion by particip ants. A 4-point res ponse scale indicating frequency of experience was used. This was the same response scale used on the Test Anxiety Inventory. The response for each item assesses severity using a 4-point response with 1 = “Almost Never”, 2 = “Sometimes,” 3 = “Often,” and 4 = “Almost Always.” The instructions were worded similarly to the instructions on Spielberger’s Test Anxiety

PAGE 48

41 Inventory (see Appendix B) To evaluate the psychometric properties of the 36-item SAI, 55 undergraduate students attending a large stat e university were offered the opportunity to take part in a psychometric study in re turn for extra credit points toward their psychology classes. Eleven part icipants were lost to attr ition by the posttest. The age range was from 18 to 48 in both pretest and postt est. In the pretest, there were 46 (85%) females and 8 (14%) males. One participant di d not disclose his/he r gender. The ethnic composition of the sample was 22 (46%) Ca ucasian, 13 (27%) African Americans, 8 (17%) Hispanics, and 1 (2%) other. Seven par ticipants chose not to disclose ethnicity. The inventory was administered and afterwards collected for scoring. Two days later, the same procedure was followed with the only ch ange being the location of the classroom. The results of a test-retest reliability an alysis and an alpha reliability analysis indicated that, from a possibl e range of scores of 36 to 144, the responses ranged from 38 to 127 on test administration one and 39 to 114 on test administration two. The mean for administration one was 68.72 ( SD =22.02), with a median score of 70. The mean for the second administration was 64.97 ( SD =19.9) with a median score of 62. Scores were positively skewed (0.72) with the 25th percentile of the first administration at 48, the 50th percentile at 70 and the 75th percentile score at 81. Any score over 81 fell in the top 25% of these data. Analysis of data collected on this measure showed an alpha coefficient of .97 for the first administration on the overall scale (.95 for the Worry and .92 for the Emotionality subscales), and .94 for the pos ttest on the overall scal e (.96 for the Worry and .92 for the Emotionality subscales). The te st-retest reliability coefficient showing

PAGE 49

42 stability of the overall scale scores on the SAI over time was .79, with the two-day testretest reliability of the worry and emotiona lity scales equal to .82 and .71, respectively. Item analyses indicated good item to total co rrelations so no items were deleted as all items positively influen ced scale reliability. Because the items were constructed to represent the factors of worry and emotionality, principal axis exploratory factor analysis with promax rotation was used to evaluate the internal struct ure of the SAI (Lunsford, 2001). Evidence of three factors of worry, emotionality and physical responses appeared. Most of the nine items with dominant loadings on factor three referred to physical symptoms (e.g., sweating, upset stomach, heart beating fast, difficulty breathing, etc.), but because the third factor was not part of the theory underlying the developmen t of the SAI, these items were dropped from further analysis. In selecting the items with the best potential for measuring emotionality, the 14 items with consistently high loadings after rotation on factor one for the combined sample, and for both sexes, were retained for further study. Two items were dropped because the loadings for these items were inc onsistent for males and females. In selecting the best worry items, the 10 items with dom inant salient loadings on factor two after rotation for the combined sample, and for both sexes, were retained for further study. The item with the smallest loading on factor one for the combined sample and with inconsistent loadings on the two factor s for males and females was dropped. Responses to the 24 retained items were fu rther evaluated in a principal axis twofactor analyses with promax rotation, and in separate analyses w ith promax rotation for males and females. The 10 items with the highe st consistent loadings for both males and

PAGE 50

43 females were selected from the pool of emoti onality items and all of the 10 worry items were retained for further study. The three item s on factor one with the smallest factor loadings (less than .60 for the combined sample and for both males and females after promax rotation) were dropped from further analysis. Two items had dual loadings for males. One of these items was retained becau se loadings were larg er for the principal factor before rotation, and for the combined sample, and for males and females after rotation. All but one of the items in factor tw o had dominant salient loadings consistently across males and females. A two-factor principal axis factor analysis was perf ormed on these 20 items. A reexamination of the items showed that one of the items did not refer to the time of studying but asked about worry cognitions after the study period. One of the items designed to measure worry had high loadings on the emoti onality scale and one of the items designed to measure emotionality had high loadings on the worry scale. One of the worry items seemed also to be asking about self esteem. This process allowed the number of items to be narrowed down to eight worry items and eight emotionality items for a total of 16 items in the inventory (Lunsford, 2001). This revised 16-item version of the SAI was used in a multi-site study by Keiffer, Reese and Cronin (2004) consisted of 165 underg raduates. Results of 10-week test-retest reliability supported the stability of the scores Test-retest reliability was .88 for the total score with .67 and .81 for the Study Wo rry and Study Emotionality subscales, respectively, indicating a satis factory level of stability ove r 10 weeks. Cronbach’s alpha

PAGE 51

44 for the overall scale was high at .96 with each subscale at .94 (Keiffer, Reese, & Cronin, 2004).

PAGE 52

45 External Evidence: Nomological Network Cronbach and Meehl (1955) developed the idea of the nomological network as a framework for evaluating cons truct validity. They argued that this network would represent a theoretical framework of the c onstruct being measured, a framework of how to measure it, and the relationships between constructs embedded in the framework. The principles that guide establis hing construct validity are to make clear what the construct is so that relationships of the construct to other constructs ca n be established. By increasing the number of variables that relate to the construct of in terest, the nomological network increases thus providing additional in sight into whether the measures used to represent the construct are operating as theorized. Using the logic of the nomological ne twork, Lunsford (2001) evaluated the relationships of the Study Anxiety Inventor y with the Test Anxi ety Inventory, trait anxiety and trait curios ity scales from the State Trait Pe rsonality Inventory, and the selfesteem and academic problems scales from th e College Adjustment Scale. Data were collected from 536 students. Since the study anxiety scales (worry and emotionality) were developed using the items from the TAI, and the basis of the construct is anxiety, it was predicted that the scores from the SA I would be positively correlated with these other measures. Results from this study suppor ted these predictions wi th the correlations between scores from the SAI and scores from measures of these constructs being between .39 and .79 (see Table 1).

PAGE 53

46 Table 1 Pearson Product Moment Correlations Between Anxiety Variables (n=536) SAI SA/E SA/WTAI TAE TAW TANXTCY SH SA/E .95 SA/W .95 .80 TAI .79 .78 .73 TAE .75 .74 .68 .96 TAW .77 .74 .72 .94 .84 TANX .45 .43 .42 .44 .43 .39 TCY -.25 -.20 -.29 -.17 -.14 -.16 -.50 SH .46 .40 .49 .38 .36 .35 .75 -.49 AP .56 .45 .61 .45 .39 .50 .34 -.25 .47 Note : SAI = Study Anxiety Inventory TAI = Test Anxiety Inventory SA/E = SA Emotionality TA/E = TA Emotionality subscale SA/W = SA Worry TA/W = TA Worry subscale TANX = Trait Anxiety TCY = Trait Curiosity SH = Self for Examinations AP = Academic Problems N = 536 all correlations were significant at <.0001 The validity of the scores from the SA I as a situation-specific construct was supported by these high correlations and by the high correlations ( r =.50 to .63) with the academic problems scale (Lunsford, 2001). Give n these high correlations, it is important to differentiate between the c onstruct of study anxi ety and test anxiety lest the reader conclude they are measures of the same constr uct. Conceptually ther e is a clear difference between test anxiety and study anxiety in the situation in which the anxiety is experienced. Test anxiety is experienced duri ng a test and the stress involved is imposed on the student by the instructor, the nature of the test, and the testing environment. Study anxiety on the other hand is experienced prior to the exam and the stress involved is selfimposed; studying is self-arrange d; and the environment is self-selected. Another major conceptual difference is that test anxiety is defined as interfering with retrieval of information during a test while study anxiet y interferes with the process of encoding information.

PAGE 54

47 Using the same sample of 536 college students, Lunsford (2001) also found relationships between other rela ted constructs. The high correl ation with the trait anxiety scores of the State-Trai t Personality Inventory ( r =.43 to .46) gives support to the concurrent validity of the scores as a meas ure of anxiety. Discriminant validity, the confirmation that this instrume nt is not measuring other co nstructs, is supported by lower correlations with trait depression measures ( r =.32 to .35), decreasing still further with trait anger ( r =.25 to .28), and trait curiosity ( r =-.23 to -.30). The correlation between study anxiety and study habits was not signifi cantly different from zero; however there was a moderate and negative correlation between study anxiety and testwiseness. Extending the Nomological Netw ork in the Present Study In the following section, a theoretical ar gument is presented linking study anxiety with procrastination. This section also pres ents an argument for why study anxiety may be unrelated to study skills and habits. Taken to gether, the pattern of relationships that is described represents an extension of the nom ological network that is used to evaluate further the construct validity of the SAI. Procrastination. The definition of procrastinatio n is the tendency to put off starting or finishing tasks (Lay, 1986) or th e avoidance of unpleasan t situations to the point of feeling discomfort (Soloman & Rothblum, 1984). Extension of the nomological network that shows the relationship of study a nxiety to other traits like procrastination requires an examination of the theory behind the construct of study a nxiety. Because it is a common belief that people avoid what they perceive to be unpleasant, the information processing theory would support that there woul d be a relationship between scores on the

PAGE 55

48 SAI and scores on measures of procrastination. Figure 4 illustrates how both worry and emotionality lead to task irrelevant behavior which may partly take the form of procrastination. The relations hip between the need to a void failure and anxiety has already been shown to be positive (Atkinson, 1974), as has the relationship between trait anxiety scores and a measure of procra stination (Beswick, Rothblum, & Mann, 1988). This relationship between a nxiety and procrastination is not yet fully and clearly established, however, and ma y be more complicated (Ferrari, Johnson, & McCown, 1995). Although some believe that people delay tasks to avoid experiencing discomfort (Soloman & Rothblum, 1984), Chu and Choi (2005) suggest that there are two major types of procrastination: passive and active. The traditional view of procrastination involving avoidance of discomfort is how these researchers define passive procrastination. They suggest though that an other reason for postponi ng certain activities is to increase motivation and enhance pe rformance achieved when a challenge is presented. Scores from a measure of passive procrastination would th eoretically correlate positively with scores from the SAI while scores from a measure of active procrastination would also correlate positively but not as high ly with scores from the SAI. Chu and Choi (2005) found that active and passive procrastin ators are not much alike but that active procrastinators are more like non-procrastina tors in their relationship with anxiety. In previous studies, the relationship betw een anxiety and procrastination has been weak (Ackerman 2005; Milgram & Tenne, 2000). Procrastination and neuroticism returned only a moderate correlation in a study to establish a relationship using academically-undecided college students (Lee & Edwards, 2004). It may be that, like

PAGE 56

49 study and test anxiety, procrastin ation can be factored into two more situa tion-specific constructs. In a specific situation concer ning statistics anxiety, Onwuegbuzie (2000) established that anxiety was a factor related to students’ procrastinating as long as possible to enroll into class and procrastin ating on assignments. In the present study, additional evidence of construct validity of the scores from the SAI was collected by examining the relationship between scores fr om the SAI and measures of passive and active procrastination devel oped by Chu and Choi (2005). Studying. Research has shown that in college most classes use exams to evaluate the progress of students. These exams are so mewhat high-stakes in that there are a number of classes that act as prerequisites for students to be able to get into upper level courses and may, in certain circumstances, be key in judgments as to whether a student gets admitted into a specific program of study (e.g., nursing programs usually require good grades in anatomy). Because of the importance of preparation for exams, researchers have looked at th e area of studying and have de veloped measures of study techniques and habits using items that addr ess exam study habits su ch as, “I read my notes over several times” and “I do less than one hour’s study for an exam” (Brown & Holtzman, 1984). The best predic tor of grades according to some researchers is study skills and habits (Gadzella, Goldston & Zimmerman, 1976; Pace, 1990; Walters & Sherk, 1990). Regular and serious study has been shown to have a positive relationship with academic performance (Fontana, 1986; Howard, 1993; Rau & Durand, 2000; Silverman & Riordan, 1974; Trapey & Harris, 1979).

PAGE 57

50 Anxiety can be due to real or perceived threat. For example, if a person knows that he or she has not studied sufficiently, it would be expected that he or she would fear the outcome of an impending exam. Also, it would seem logical that if a person consistently gets poor results because of low academic ability that he or she would develop a dread of coming exams. This has not been supported by research; however research has shown that anxiety affects the highly intelligent and skillful person as much as and sometimes more than those who logi cally should feel the anxiety. It has been shown that perfectionists are often filled w ith anxiety, which interferes with encoding more than for non-perfectionists (Covington & Omelich, 1985). This theory that anxiety affects those regardless of their intelligence and skill sets would suggest that the relationship between the scores on the SAI and scores on a m easure of study habits would be small or not significantly different from zer o. The fact that scores from the SAI did not correlate with study habits in the 2001 study by Lunsford supports this theory that study anxiety is not affected by knowledge of how to study or w ith regularity in the use of techniques to study effectively, and that it is a construct i ndependent of study skills and habits. The relationship between study skills an d habits and study anxiety is examined in the current study. Research into coping skills suggests that those who have the ability to respond to test questions correctly, and feel confident in their answers even when they have not studied, will be less anxious during a test. This is supported by the findings of a moderate negative relationship of the SAI scores with test-taking skill scores ( r = -.28 to -.30). This suggests that when a student has confidence in his or her knowledge, there is more of an

PAGE 58

51 ability to cope with worry a nd emotionality even when studyi ng is ineffective, and this perception will lead to decr eased anxiety while studying. Summary The purpose of this study is to evaluate th e validity of the inferences that can be made using scores from the Study Anxiety Inventory. The overview of the validation process has established a basis for collection of further data to continue the validation of the scores from this measure. Analysis of item content (content validity) has been described and some initial findings concerning the internal structure of the responses to the items (exploratory and confirmatory factor analysis) have been described; factorial invariance has been introduced with a focu s on gender differences. Relationships between study anxiety and other variable s have been discussed and directions for establishing further relationships based on theory have b een indicated. A summary of the results of four previous studies that support the reli ability and validity of the SAI scores is presented in Appendix A.

PAGE 59

52 Chapter 3 Method In accordance with the objective to obtain ev idence to support the validity of the inferences from the Study Anxiety Inventory (SA I), this chapter desc ribes the procedures used to collect data using the SAI and other theoretically relevant variables that were used in the validation process: test anxiety, trait anxiety, trait curiosity, study skills and habits, and active and passive procrastination. The chapter begins w ith a review of the purposes of the study, followed by a descrip tion of the characteristics of those participating in the validation process, data collection procedures, and statistical analysis of the data. Purpose The purpose of this study was to extend research on the cons truct validity of responses from college students to the Study Anxiety Inventory (SAI). Several approaches were used. Because the SAI was de veloped using a theory that the construct consisted of two highly correlated factors, s upport for this two-factor model was needed in establishing factorial validit y. Confirmatory factor analysis was used to test the two factor model (worry and emotionality). As part of the CFA, the factorial invariance of the SAI for males and females also was examined. Additional evidence for construct validity was collected by examining the relation between the SAI and other th eoretically relevant variables that were part of the nomological network framework. Specific purposes were to:

PAGE 60

53 1. Evaluate the two-factor measurem ent model underlying the Study Anxiety Inventory in a sample of college st udents from various disciplines; 2. Evaluate the factorial invariance of the two-factor measurement model underlying the Study Anxiety Invent ory across male and female college students; 3. Examine the relationship between the scores on the Study Anxiety Inventory and scores on two measur es of procrastination (active procrastination and passive procras tination), a measure of study skills and habits, a measure of test a nxiety, and two of the four trait personality measures of the Stat e-Trait Personality Inventory, the Trait-Anxiety Scale (T-Anx) and th e Trait Curiosity Scale (T-CY). Participants The participants for this first study we re 2,002 undergraduate students (939 males and 1,054 females, 9 did not indicate gender). Stud ents were recruited from each of four colleges at a large state university in the sout heast: College of Arts and Sciences, College of Business, College of Education, and Colle ge of Engineering. These students were recruited by asking for volunteers from classes of professors who agreed to allow data collection in their class. The researcher approached 12 professors teaching in the summer semester and 16 professors teaching in the fall semester, all of whom agreed. It was made optional as to whether the professor offered an extra credit point to those who would be prepared to complete the measures and one psychology professor offered that point. For that professor a sign-up sheet was provided at the table where the completed measures

PAGE 61

54 were returned and this list of names was late r given to that professor. Those who allowed data to be collected but did not offer ex tra credit merely asked if students would volunteer for the study and extra credit was not mentioned. The Arts and Sciences (A & S) students were recruited from social scien ce statistics classes as this class attracts enrollment from most of the different discip lines offered in the College of Arts and Sciences; the social sciences set statistics as part of the graduati on requirements. Students from the College of Arts and Sciences we re recruited from classes taught by the researcher in the summer and fall semester s during the first week of each class to minimize the influence of social desirability on responses resulting from participants knowing the instructor. The students from the College of Business were obtained by handing out the questionnaire package to two large classes (Ethics and the Law, and Economics), both of which were required cour se for majors in business. The students from the College of Engineering were obt ained by attending and handing out the questionnaire in a number of smaller (11-30 students) classes and also by handing out the questionnaire package in a central meeting ar ea to those who were waiting for classes to start. The students from the College of Education were obtained by handing out questionnaire packages in ESOL (English for Speakers of Other Language) classes, which is a required class for those wishing to obtain teaching certificates in Florida schools and, as such, would have students from many different majors of the college. Recruitment provided sufficient responses to analyze the 16 SAI items for both males and females based on a recommendation by DiStefano and Hess (2005) that there be no fewer than five responders for each item. Of the 2,002, 1,964 fully completed at

PAGE 62

55 least two measures, indicated gender, and were a part of one of the four colleges under investigation. There were 218 male and 446 fema le College of Arts and Science students, 261 male and 195 female College of Engin eering students, 237 male and 194 female College of Business students, and 210 ma le and 203 female College of Education students. Obtaining more than 80 of each gender from each college provided more than the expected ratio of between 5 and 17 responders for each item. The ethnicity of each college was similar in most respects except the College of Engineering which had fewer than half of the African Americans found in the other colleges, and the College of Education wh ich had almost no representation of Asian Americans (2%). The ethnicity percentages for the combined sample was 11% African American, 8% Asian, 12% Hispanic, 54% Ca ucasian, 1% American Indian/Alaskan Native, 4% more than one race, and 10% not recorded. The design of the study was that students over the age of 64 and under the age of 18 would be excluded to expedite review by the IRB; however no students fell in either category. The students were aged from 18 to 64 years with a mean of 22.16 although the median age of 21 may more accurately reflect the ages of those in this sample.

PAGE 63

56 Table 2 Gender and Ethnicity of 2,002 Partic ipants Across Four Colleges Variable Total Arts & Sciences Business Education Engineering X2 Gender 2002 678439422463 Male 939 225237216261 Female 1063 453202206202 79.84Race/ethnicity none 153 25445925 Afr. Am./Black 265 96567439 Asian Am/ 130 37411042 Hispanic/Latino 237 106403556 White 1050 336228227259 American Indian 42 20697 Other 66 3212319 2 or more 59 2612516 122.12 Measures Along with the Study Anxiety Inventory (SAI), the following questionnaires were administered to the participan ts during the class period: the Test Anxiety Inventory (TAI), the trait anxiety scale (T-Anx) and trait curiosity scal e (T-CY) from the State-Trait Personality Inventory (STPI), study for exam inations (SH) scale from the Study Habits Evaluation and Instru ction Kit (SHEIK), The Passive Procrastination Scale (PPS), and The Active Procrastination Scale (APS). The Test Anxiety Inventory (TAI), wh ich was developed by Spielberger (1980), is a 16-item self-report measure that was designed to assess i ndividual differences in test anxiety as a situation-specific personality trait. Besides the total scale score, two scales of test anxiety, Test Anxiety Emotionality (TA/E) and Test Anxiety Worry (TA/W), measure the two major components of test a nxiety, emotionality and worry, as identified by Liebert and Morris (1967). Th e TA/E scale was developed from the prototypic item “While taking examinations I have an uneas y, upset feeling” and the TA/W scale was

PAGE 64

57 developed from the prototypical statement “Dur ing examinations I get so nervous that I forget facts I really know” (see Appendix B). Relative to Huck and Jacko’s (1974) form and format concerns, all 20 items of the original TAI were administered to the participants using the original multiple-choice format. Frequencies of symptoms are reported on a 4-point scale: 1 = “Almost Never,” 2 = “Sometimes,” 3 = “Often” and 4 = “Almost Always.” Spielberger (1980) reports that this scale exhi bits good test-retest reliability ranging from .62 ( over 6 months) to .80 (over 1 month). The Cronbach alphas for the TAI total scale are uniformly high at .92 or higher for the tota l scale score and .94 for each of the two subscales. Exploratory a nd confirmatory factor analysis has been carried out on this measure using the re sponses from 752 and 1537 university students, the results from which support the two-factor st ructure of this measur e (Kieffer, Reese, & Cronin, 2004; Ware, Galassi, & Dew, 1990). The trait anxiety and trait curiosity sc ales from the State-Trait Personality Inventory (STPI Form Y) (Spielberger, 1995) are from an 80-item measure of four state (S-) and four trait (T-) constructs: curios ity (S-Cy and T-Cy), anxiety (S-Anx and TAnx), anger (S-Ang and T-Ang), and depres sion (S-Dep and T-Dep). Responses from only 20 trait (10 T-Anx and 10 T-CY) items were included in the battery of tests and used in the analysis of the data (see Appendix B) These scales use a 4-point response scale indicating frequency of experience (1 = "Almos t Never", 2 = "Sometimes", 3 = "Often", 4 = "Almost Always"). Forward and reverse scor ings are used in bot h scales. Spielberger (1980) reports that these scales exhibit good reliability with al pha coefficients of between .80 and .96 for the entire sample and for male s and females. Correlations of this scale

PAGE 65

58 with the State Trait Anxiety Inventory, the current gold standard measure of anxiety, were high for both males and females (.95 for both) (Spielberger, 1980). Descriptive statistics, scale intercorrelations and item-re mainder correlations are provided in the manual for male and female college students ( n = 280) and navy recruits ( n = 270). The study for examinations (SH) scale of Study Habits Evaluation and Instruction Kit (SHEIK) from the New Zealand Council for Educational Research (1979) served as a measure of the degree of knowledge and appl ication concerning a student’s study habits (SH) (see Appendix B). The SH consists of 25 self-report items using a 5-point degree of response: 1 = "Never or Almost Never", 2 = "About of the Time", 3 = "About of the Time", 4 = "About of the Time", and 5 = "Almost Always". Both forward and reverse scoring is used in the SHEIK. The studying for examinations scale (SH) items address exam study habits (e.g., “I read my note s over several times”). The SHEIK manual reports that reliability of the SH is good w ith KR20 value of .86, and split half value at .86 (New Zealand Council for Educational Rese arch, 1979). Analysis from a sample of 536 university students indicated that the rela tionship between this measure and scores from the SAI was not significantly differe nt from zero (Lunsford, 2001). This measure has been used in studies by the New Zeala nd Council for Educational Research and is being examined for improvement (W. R. Brown, personal comm unication, May 2, 2000). The Passive Procrastination Scale (PPS ) and The Active Procrastination Scale (APS) measure two types of procrastination (Chu & C hoi, 2005). The first (PPS) measures the procrastination construct in th e traditional sense in that those with high scores are those who are paralyzed by their indecision and often fail to complete tasks on

PAGE 66

59 time. The second (APS), however, measures a mo re productive kind of procrastination in that those with high scores ar e those who make a conscious d ecision to delay in order to increase pressure as they find that in th ese circumstances they complete work and perform better. Concerning academic performance, those high in APS are more like nonprocrastinators than they are li ke passive procrastinators. Th e APS is a 6-item scale with acceptable internal consistency ( =.82) while the coefficient for the 12-item PPS is less acceptable ( =.67). The response format for both of these measures is a 7-point scale ranging from 1 = "Not at all True" to 7 = "Very True" (See Appendix B). Procedure The battery of tests printed in light grey was administered to groups of undergraduate students in soci al science statistics classes in the College of Arts and Sciences, and in various cla sses in the College of Busine ss, College of Education, and College of Engineering at a state university in the sout heast. Undergraduate course instructors were contacted in advance in or der to obtain permissi on to administer all related materials to their stude nts, which took less than 20 mi nutes of class time in total. To maintain anonymity of th e participants, no identifying information was requested. Participants were tested within their regular classrooms. During the last 30 minutes of class time, the researcher was intr oduced and participants were informed that the goals of the study were to learn about the feelings and a ttitudes of stud ents through the use of several questionnair es. The researcher then read the instructions aloud and briefly explained the format of the questionnai res after which the participants filled them out as instructed. Finally, part icipants were provided with an informational debriefing (in

PAGE 67

60 Appendix A) and were informed that a fee dback session would be scheduled at a later time during which any questions could be addressed.

PAGE 68

61 Data Analysis Analysis of the data using the two sta tistical computer programs, Statistical Package for the Social Science (SPSS) 16.0 and Mplus 3.0 (Muthn & Muthn, 1998), included calculating the internal consistenc y coefficient for all measures (Cronbach alpha). This was followed by a confirmatory fact or analysis (CFA) to evaluate the fit of the two-factor model (worry and emotiona lity). For the CFA, Hu and Bentler (1999) recommend using the standardized root m ean square residual (SRMR) in conjunction with another fit index like the comparative fit (CFI) or root mean square error of approximation (RMSEA). It is desirable to have the standardized difference between a covariance and a predicted covariance as clos e to zero as possible (z ero indicates perfect fit), but it is typical for the SRMR to range from .05 to .10, although Hu and Bentler (1999) suggest that a cutoff of .08 or below shoul d be used to indicate good model fit. It has been argued that cut offs for CFI shoul d be at least .90, which would indicate that 90% of the covariati on in the data set can be repr oduced by the model, and, although Bollen (1989) suggests that these cut-offs are arbitrary, Hu and Bentler (1999) suggest that minimum type I and type II errors w ill occur with a CFI of .95. Good model fit is reflected by an RMSEA of .06 or less (Hu & Bentler, 1999). To evaluate the equivalence of the tw o-factor measurement model underlying the SAI, multigroup CFA was used. The data were divided by gender, and variancecovariance matrices were calculated for each group (males, females). The fit of the twofactor model was evaluated separately for males and females followed by the evaluation of the invariance of the model for males a nd females. Invariance tests were conducted

PAGE 69

62 using a multiple group analysis to test the equa lity of the factor loadings (i.e., factor patterns), residual variances, factor variances, and the c ovariance between factors by gender. Equality restrictions were impos ed across males and females for tests of invariance. A Chi-square-difference test for re lative fit for a nested sequence of models was used for this test. Analyses were conduc ted separately for each college and and the significance level was set at .01 except in situ ations where a more stringent significance level was used to take into account multiple statistical tests. In order to extend the nomological network, Pearson’s product moment correlations were calculated be tween the SAI and the relevant theoretical variables for each of the four colleges.

PAGE 70

63 Chapter 4 Results In keeping with the purpose of the study to provide evidence to evaluate the validity of the inferences from the Study Anxi ety Inventory, data were collected from undergraduate students from four colleges at a southeastern state university: Arts and Sciences, Engineering, Business, and Education. In this chapter, resu lts of the analyses are presented by research question in three sections. Section one contains the results of the conf irmatory factor analyses that were used to evaluate the factorial validity of the SAI scores. These results include fit indices from the CFA two-factor model (worry and emo tionality) for each college. In the second section, results from the tests of factorial i nvariance of the SAI for males and females are reported. Finally, additional evidence for cons truct validity using relationships between the SAI and other measures (i.e., nomologi cal network framework) is presented. Research Question 1: Confirmatory Factor Analysis Confirmatory factor analyses (CFA) we re performed using Mplus, version 3.0 (Muthn & Muthn, 1986) to evaluate the two-factor model underlying the Study Anxiety Inventory (see Figure 5). Analyses were based on the variance-covariance matrix of the 16 observed va riables and maximum likelihood estimation was used to estimate the model parameters. Fit indices that were used to evaluate the model included the chi-square test ( 2) test, an indicator of the fit of the responses to the model; the standardized root mean square residual (SRMR), an indicator of the mean of the differences between the predicted variances and covariances and th eir observed values;

PAGE 71

64 comparative fit index (CFI), an indicator of th e percentage of covari ation in the data set that can be reproduced in the model; and th e root mean square error of approximation (RMSEA), an indicator of the discrepancy per degree of freedom. Listwise deletion was employed in the calculations for the confir matory factor analysis which had a small influence on the sample sizes reported. Sample size dropped from 2002 to 1867 when using listwise deletion. Tests of statistical significance were conducted at the .01 level given the large sample size. Analyses were c onducted separately for each college and are presented in the next section. College of Arts and Sciences. Table 3 provides descrip tive statistics for the 16 observed variables used in the confirmatory factor analysis for responses from students from the College of Arts and Sciences. The responses for each item ranged from one to four and the means for the 16 items ranged from 1.67 to 2.22 (median = 1.93) with standard deviations ranging fr om 0.78 to 1.08 (median = 0.91). The normality of the distributions for the 16 items was evaluated using univariate skewness and kurtosis measures for each of these variables. All skewness values were close to zero with the smallest skewness va lue of 0.37 and the largest skewness value of 1.17 (median = 0.68) which reflects approximate symmetry in each of the items. All kurtosis values were less than 1.0 for each of the items with values ranging from -0.84 to 0.62 (median = -0.31), which suggest that the peak and tails of the distribution were similar to the normal curve. Multivariate normality evaluate using Mardia’s test of multinormality was not demonstrated in the College of Arts and Sciences for the total

PAGE 72

65 group or by gender ( p <.001 in every case) so the a ssumption of multivariate normality was not tenable.

PAGE 73

66 Table 3 Descriptive Statistics for the 16 Obser ved Variables Used in the Two-Factor Confirmatory Factor Analyses of the Study An xiety Inventory for the College of Arts and Sciences (n=662) Item # Mean SD Skewness Kurtosis 1e 1.92 0.88 0.78 0.01 2w 2.02 0.78 0.52 0.09 3w 2.21 0.87 0.38 -0.43 4e 1.94 0.93 0.64 -0.52 5e 1.76 0.87 0.96 0.17 6w 2.22 0.89 0.38 -0.44 7w 2.00 0.91 0.60 -0.35 8e 2.02 1.08 0.66 -0.84 9e 1.88 0.96 0.80 -0.37 10w 2.00 0.93 0.58 -0.47 11w 1.99 0.98 0.59 -0.64 12e 1.87 0.91 0.75 -0.25 13e 1.67 0.87 1.17 -0.62 14w 1.92 0.91 0.70 -0.27 15w 1.83 0.84 0.73 -0.09 16e 1.85 0.91 0.78 -0.21 Note : Emotionality (e) items are 1, 4, 5, 8, 9, 12, 13, 16; Worry (w) items are 2, 3, 6, 7, 10, 11, 14, 15 The 2 fit statistics for the CFA of the two-factor (worry, emotionality) model for the students from the College of Arts and Scie nces indicated that the fit of the model was not acceptable, 2 (103, N = 660) = 772.52, p < .001. Alternative measures of fit were included because one of the limitations of the 2 is that it is sensitive to sample size. The SRMR of .047 indicated acceptable fit whil e the CFI of .918 and the RMSEA of .099 indicated less than acceptable fit.

PAGE 74

67 E1 E4 E5 E8 E9 E12 E13 E16 E2 E3 E6 E7 E10 E11 E14 E15 Figure 5. Relationships of Items to Fact ors in the Two-Factor Model. Emotionality I1 I4 I5 I8 I9 I12 Worry I3 I6 I7 I10 I11 I15 I13 I16 I2 I14

PAGE 75

68 The unstandardized factor lo adings, excluding the one fi xed to 1.0 to identify the model, ranged from 1.18 to 1.39 (median = 1.22 ) for the factor of emotionality and from 1.01 to 1.45 (median = 1.34) for the factor of worry. All loadings were significantly different from zero ( p < .01). An examination of the sta ndardized factor loadings showed that the loadings ranged from .69 to .86 for th e emotionality items and from .63 to .85 for the worry items. The correlation be tween emotionality and worry was .87. Sources of misfit of the model were e xplored by examining the modification indices that indicated the expected decrease in model fit chi square ( 2) that would result when a specific parameter, constraine d initially to zero, s ubsequently was freely estimated. The largest modification indices we re for covariances between errors. Because similarities in the item content for pairs of items may result in large error covariances, examination of the modification indices a nd wording of the pairs with chi-square differences larger than the critic al chi-square statistic of 6.64 ( p = .01) was carried out. Table 4 lists all pairs of items with changes in chi square that were statistically significant and shows details of the six pairs of items w ith the highest chi-square differences. This value indicates the improvement in model fit if a covariance was posited between the errors of the items. The largest modification index for the m odel was for the covariance between the errors for items 6 and 3 ( 2 = 189.75). These items were ve ry similar and shared the word “interfere”. Although the change in chi-square was substantially lower for the remaining five pairs of error covariances ( 2 ranged from 29.56 to 52.49), examination

PAGE 76

69 of those items revealed similar wording in th e pairs of items. Items 4 and 5 were similar because the word “nervous” could be view ed as synonymous with “uneasy” and/or “upset.” Items 9 and 12 used the words “p anicky” and “very tense”, which could be viewed as representing the same feeling. Exam ination of the wording in items 13, 14, and 15 (all worry items) revealed that “freezing up,” “mental block,” and “can’t get my brain to organize” may be sufficiently simila r to suggest redundancies in the items. Examination of items 10 and 7 reveal that “can ’t absorb the material” and “not being able to learn the material” are sufficiently simila r that they may have been viewed by the respondents as nearly identical items. The m odification indices for the pairs of items are shown in Table 4.

PAGE 77

70 Table 4 Modification Indices for Error Covarian ces for the Study Anxiety Inventory for Responses from All Arts and Science Students (n=625) Items with errors covarying Chi-Square Difference Pair 1: 3. I can’t keep my mind on the subj ect when studying for an exam because other thoughts interfere (Worry) 6. While studying for tests, other thoughts interfere with my learning (Worry) 189.75 Pair 2: 4. Even when I have plenty of time, I feel nervous when I tr y to study for an exam (Emotionality) 5. While studying for exams, I have an uneasy, upset feeling (Emotionality) 48.89 Pair 3: 9. I feel panicky when studying for an important exam (Emotionality) 12. While studying for exams, I feel very tense (Emotionality) 38.65 Pair 4: 13. I freeze up while studying for an important test (Emotionality) 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 49.85 Pair 5: 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 15. When I am studying for a test, I can’t get my brain to organize the information (Worry) 52.49 Pair 6: 10. While studying for exams, I am stressed with thoughts like “I can’t absorb the material properly” (Worry) 7. While studying for a test, I worry about not being able to learn the material (Worry) 29.56 2 with 1 12 with 4 15 with 1 4 with 2 14 with 7 9 with 4 10 with 3 16 with 3 13 with 7 27.05 27.04 22.58 22.02 18.58 16.51 15.72 15.15 14.33 9 with 3 4 with 3 15 with 10 16 with 11 15 with 13 6 with 4 10 with 9 13 with 12 13 with 8 13.44 13.20 12.69 12.36 12.59 12.17 11.52 11.26 11.16 5 with 1 7 with 2 9 with 8 12 with 6 9 with 6 16 with 6 12 with 2 5 with 2 14 with 9 10.78 10.75 9.01 8.67 8.55 8.43 7.34 6.99 6.83 The correlation of .85 between the emotionali ty and worry factors was statistically significantly different from 0 ( p < .001). The strength of the correlation between the two factors led the researcher to entertain the possibility of a one-factor model underlying the Study Anxiety Inventory (see Figure 6). Based upon the 2, the fit of the one-factor model was also not acceptable with 2 (104, N = 660) = 1187.42, p < .001, which is

PAGE 78

71 larger than the two-factor model. The SRMR for the one-factor model was .06, which indicated poorer fit compared with the two-factor model (SRMR = .047 for the two-factor model). The CFI of .868 for the one-factor mode l indicated that the fit was not as good as the two-factor model (CFI = .918). The RM SEA of .126 for the one-factor model was larger than the RMSEA for the two-factor model (.099), again suggesting that the onefactor solution provided an ev en less acceptable fit compared with the two-factor model. The standardized factor loadings for the one-factor model ranged from .56 to .83 and were all significantly different from zero ( p < .01).

PAGE 79

72 E1 E4 E5 E8 E9 E12 E13 E16 E2 E3 E6 E7 E10 E11 E14 E15 Figure 6. Relationships of Items to F actors in a One-Factor Model. College of Engineering. Table 5 provides descriptive statistics for the 16 observed variables used in the confirmatory factor analysis for responses from the students in the College of Engineering. The means for th e 16 items ranged from 1.59 to 2.08 (median = 1.83) with standard deviations rangi ng from 0.74 to 1.02 (median = 0.84). I4 I5 I8 I9 I12 Study Anxiety I3 I6 I7 I10 I11 I15 I13 I16 I2 I14 I1

PAGE 80

73 The normality of the distributions for the 16 items was evaluated using univariate skewness and kurtosis measures for each of these variables. All skewness values were close to zero with the smallest skewness va lue of 0.44 and the largest skewness value of 1.30 (median = 0.81). All kurtosis values were near zero for each of the items with values ranging from -0.30 to 1.21 (median = 0.03), which suggests that the peak and tails of the distribution are similar to the normal curv e. Multivariate normality evaluate using Mardia’s test of multinormality was not demonstrated in the College of Engineering for the total group or by gender ( p <.001 in every case) so the assumption of multivariate normality was not tenable. Table 5 Descriptive Statistics for the 16 Obser ved Variables Used in the Two-Factor Confirmatory Factor Analyses of the St udy Anxiety Inventory for the College of Engineering (n=433) Item # Mean SD Skewness Kurtosis 1e 1.88 0.86 0.67 -0.12 2w 1.95 0.74 0.46 0.29 3w 2.06 0.84 0.44 -0.18 4e 1.82 0.83 0.77 0.17 5e 1.62 0.77 1.03 0.65 6w 2.08 0.89 0.50 -0.20 7w 1.84 0.81 0.64 -0.04 8e 1.85 1.02 0.91 -0.30 9e 1.77 0.92 0.92 -0.05 10w 1.86 0.85 0.69 -0.11 11w 1.86 0.92 0.77 -0.22 12e 1.78 0.86 0.86 0.10 13e 1.59 0.81 1.30 1.21 14w 1.73 0.83 0.90 0.31 15w 1.70 0.82 0.97 0.45 16e 1.73 0.84 0.88 0.10 Note : Emotionality (e) items are 1, 4, 5, 8, 9, 12, 13, 16; Worry (w) items are 2, 3, 6, 7, 10, 11, 14, 15

PAGE 81

74 The 2 fit statistics for the CFA of the two-factor (worry, emotionality) model for the students from the College of Engineering indicated that the fit of the model was not acceptable, 2 (103, N = 428) = 521.82, p < .001. Alternative meas ures of fit were included because one of the limitations of the 2 is that it is sensitive to sample size. The SRMR of .050 indicated acceptable fit, wh ile the CFI of .912 and the RMSEA of .097 indicated less than acceptable fit. The unstandardized factor lo adings, excluding the one fi xed to 1.0 to identify the model, ranged from 1.20 to 1.51 (median = 1.30 ) for the factor of emotionality and from 1.08 to 1.38 (median = 1.25) for the factor of worry. All loadings were statistically significantly different from zero ( p < .01). An examination of the standardized factor loadings showed that the loadings ranged fr om .61 to .85 for the emotionality items and from .66 to .80 for the worry items. The co rrelation between emotionality and worry was .89. Sources of misfit of the model were e xplored by examining the modification indices that indicated the expected decrease in model fit chi square ( 2) that would result when a specific parameter, constraine d initially to zero, subsequently was freely estimated. The largest modification indices we re for covariances between errors. Because similarities in the item content for pairs of items may result in large error covariances, examination of the modification indices and word ing of the pairs of items with chi square differences larger than the critical chi square of 6.64 ( p = .01) was carried out. Table 6 lists all pairs of items with changes in chi square that were statistically significant and shows the six pairs of items with th e highest chi-square differences ( 2 > 15). This value

PAGE 82

75 indicates the improvement in model fit if a covariance was posited between the errors of the items. The largest modification index for the m odel was for the covariance between the errors for items 6 and 3 ( 2 = 123.22). These items were ve ry similar and shared the word “interfere”. Although the change in chi-square was substantially lower for the remaining five pairs of error covariances ( 2 ranged from 15.99 to 28.24), examination of those items revealed similar wording in th e pairs of items. Items 9 and 10 used the words “panicky” and “stressed”, which could be viewed as representing the same feeling. Examination of the wording in items 13, 14 and 15 (all worry items) revealed that “freezing up,” “mental block,” and “can’t ge t my brain to organize” may be sufficiently similar to suggest redundancies in the items The modification indices for the pairs of items are shown in Table 6.

PAGE 83

76 Table 6 Modification Indices for Error Covarian ces for the Study Anxiety Inventory for Responses from All Engineering Students Items with errors covarying Chi-Square Difference Pair 1: 3. I can’t keep my mind on the subj ect when studying for an exam because other thoughts interfere (Worry) 6. While studying for tests, other thoughts interfere with my learning (Worry) 123.22 Pair 5: 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 15. When I am studying for a test, I can’t get my brain to organize the information (Worry) 28.24 Pair 2: 2. While I am studying for an exam I often think “I’m not getting this” (Emotionality) 1. I feel very uneasy just before starting to study for an exam (Emotionality) 27.06 Pair 4: 13. I freeze up while studying for an important test (Emotionality) 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 19.14 Pair 3: 9. I feel panicky when studying for an important exam (Emotionality) 10. While studying for exams, I am stressed with thoughts like “I can’t absorb the material properly” (Worry) 19.01 Pair 6: 4. Even when I have plenty of time, I feel nervous when I try to study for an exam (Emotionality) 1. I feel very uneasy just before starting to study for an exam (Emotionality) 15.99 12 with 11 3 with 2 5 with 4 4 with 2 15 with 1 12 with 9 13 with 1 9 with 3 14.83 14.72 14.46 12.61 12.57 12.18 11.78 11.70 14 with 1 4 with 3 15 with 12 12 with 2 12 with 5 13 with 12 9 with 6 12 with 4 11.01 10.55 10.26 9.88 9.42 8.90 8.70 7.84 7 with 5 14 with 2 13 with 3 13 with 6 12 with 8 14 with 12 16 with 12 5 with 1 7.66 7.53 7.39 7.22 6.96 6.82 6.79 6.67 The correlation of .89 between the emotionali ty and worry factors was statistically significantly different from 0 ( p < .001). The strength of the correlation between the two factors led the researcher to entertain the possibility of a one-factor model underlying the Study Anxiety Inventory. Based upon the 2, the fit of the one-fact or model also was not acceptable with 2 (104, N = 410) = 613.45, p < .001, which is larger than the two-factor model. The SRMR for the one-factor m odel was .0501, which indicated poorer fit

PAGE 84

77 compared to the two-factor model (SRMR = .050 for the two-factor model). The CFI of .902 for the one-factor model indicated that the fit was not as good as the two-factor model (CFI = .912). The RMSEA of .109 for th e one-factor model was larger than the RMSEA for the two-factor model (.097), again suggesting that the on e-factor solution provided an even less acceptable fit. The standardized factor loadings for the one-factor model ranged from .59 to .82 and were all statis tically significantly different from zero ( p < .01). College of Business. Table 7 provides descriptive statistics for the 16 observed variables used in the confirmatory factor analysis for responses from the students from the College of Business. The means for the 16 items ranged from 1.70 to 2.21 (median = 1.93) with standard deviations rangi ng from 0.74 to 1.04 (median = 0.87). The normality of the distributions for the 16 items was evaluated using univariate skewness and kurtosis measures for each of these variables. All skewness values were less than 1.0 with the smalle st skewness value of 0.25 and th e largest skewness value of .97 (median = 0.68), which shows that the symmetry of the it ems was acceptable. All kurtosis values were less than 1.0 for each of the items with values ranging from -0.62 to 0.22 (median = -0.22), which suggests that the peak and tails of th e distribution were similar to the normal curve. Multivariate normality evaluate using Mardia’s test of multinormality was not demonstrated in the College of Business for the total group or by gender ( p <.001 in every case) so the assump tion of multivariate normality was not tenable.

PAGE 85

78 Table 7 Descriptive Statistics for the 16 Obser ved Variables Used in the Two-Factor Confirmatory Factor Analyses of the Study Anxiety Inventory for the College of Business (n=399) Item # Mean SD Skewness Kurtosis 1e 1.94 0.84 0.63 -0.09 2w 1.98 0.74 0.52 0.22 3w 2.21 0.85 0.25 -0.58 4e 1.90 0.88 0.71 -0.23 5e 1.72 0.82 0.94 0.21 6w 2.18 0.88 0.39 -0.52 7w 1.98 0.87 0.59 -0.24 8e 1.98 1.04 0.75 -0.62 9e 1.91 0.93 0.70 -0.42 10w 1.97 0.90 0.61 -0.38 11w 1.95 0.95 0.66 -0.50 12e 1.83 0.87 0.86 0.12 13e 1.70 0.86 0.97 0.10 14w 1.87 0.85 0.71 -0.06 15w 1.82 0.80 0.65 -0.11 16e 1.85 0.89 0.76 -0.21 Note : Emotionality (e) items are 1, 4, 5, 8, 9, 12, 13, 16; Worry (w) items are 2, 3, 6, 7, 10, 11, 14, 15 The 2 fit statistics for the CFA of the two-factor (worry, emotionality) model for the students from the College of Business i ndicated that the fit of the model was not acceptable, 2 (103, N = 399) = 527.43, p < .001. Alternative meas ures of fit were included because one of the limitations of the 2 is that it is sensitive to sample size. The SRMR of .05 indicated acceptable fit, while the CFI of .908 and the RMSEA of .102 indicated less than acceptable fit. The unstandardized factor lo adings, excluding the one fi xed to 1.0 to identify the model, ranged from 1.15 to 1.41 (median = 1.30 ) for the factor of emotionality and from 1.07 to 1.49 (median = 1.27) for the factor of worry. All loadings were statistically

PAGE 86

79 significantly different from zero ( p < .01). An examination of the standardized factor loadings showed that the loadings ranged fr om .67 to .84 for the emotionality items and from .65 to .82 for the worry items. The co rrelation between emotionality and worry was .89 Sources of misfit of the model were e xplored by examining the modification indices that indicated the expected decrease in model fit chi square ( 2) that would result when a specific parameter, constraine d initially to zero, s ubsequently was freely estimated. The largest modification indices we re for covariances between errors. Because similarities in the item content for pairs of items may result in large error covariances, examination of the modification indices a nd wording of the pairs with chi square differences larger than the critical chi square of 6.64 ( p = .01) was carried out. Table 8 lists all pairs of items with changes in chi square that were statistically significant and shows the six pairs of items with th e highest chi-square differences ( 2 > 17). This value indicates the improvement in model fit if a covariance was posited between the errors of the items. The largest modification index for the m odel was for the covariance between the errors for items 6 and 3 ( 2 = 77.00). These items were very similar and shared the word “interfere”. Although the change in chi-square was substantially lower for the remaining five pairs of error covariances ( 2 ranged from 16.33 to 32.00), examination of those items revealed similar wording in th e pairs of items. Item 4 would be responded to in a similar way to items 6 and 3 because the participant may feel that interference is causing their nervousness. Items 9 and 12 used the words “panicky” and “very tense”,

PAGE 87

80 which could be viewed as representing a si milar feeling. Examination of the wording in items 14 and 15 (all worry items) revealed that “mental block” and “can’t get my brain to organize” may be sufficiently similar to s uggest redundancies in the items. Examination of items 5 and 7 reveal that “uneasy” and “worry about not bei ng able to learn the material” may be identified as having suffici ently similar meaning that they may have been viewed by the respondents as very sim ilar items. Examination of the way item 4 might be related to both items 6 and 3 di d not reveal any obvi ous relationship. The modification indices for the pairs of items are shown in Table 8.

PAGE 88

81 Table 8 Modification Indices for Error Covarian ces for the Study Anxiety Inventory for Responses from All Business Students (n=399) Items with errors covarying Chi-Square difference Pair 1: 3. I can’t keep my mind on the subj ect when studying for an exam because other thoughts interfere (Worry) 6. While studying for tests, other thoughts interfere with my learning (Worry) 77.00 Pair 2: 4. Even when I have plenty of time, I feel nervous when I tr y to study for an exam (Emotionality) 6. While studying for tests, other though ts interfere with my learning (Worry) 32.00 Pair 3: 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 15. When I am studying for a test, I can’t get my brain to organize the information (Worry) 27.90 Pair 4: 5. While studying for exams, I have an uneasy, upset feeling (Emotionality) 7. While studying for a test, I worry about not being able to learn the material (Worry) 22.73 Pair 5: 2. While I am studying for an exam I often think “I’m not getting this”(Worry) 12. While studying for exams, I feel very tense (Emotionality) 21.85 Pair 6: 4. Even when I have plenty of time, I feel nervous when I tr y to study for an exam (Emotionality) 5. While studying for exams, I have an uneasy, upset feeling (Emotionality) 19.84 4 with 3 13 with 6 16 with 12 4 with 1 14 with 13 10 with 9 2 with 1 14 with 7 12 with 11 10 with 5 18.21 17.16 16.49 16.33 15.55 15.47 15.37 15.37 14.87 14.67 3 with 2 9 with 3 9 with 6 10 with 3 9 with 5 6 with 1 16 with 2 15 with 3 14 with 8 13 with 1 14.50 13.34 12.52 11.52 10.49 9.52 9.01 8.95 8.81 8.65 5 with 1 12 with6 12 with 3 15 with 1 13 with 4 3 with 1 10 with 6 11with 4 5 with 3 4 with 2 8.22 8.20 7.80 7.70 7.25 7.10 7.07 6.89 6.67 6.65 The correlation of .89 between the emotionali ty and worry factors was statistically significantly different from 0 ( p < .001). The strength of the correlation between the two factors led the researcher to entertain the possibility of a one-factor model underlying the Study Anxiety Inventory. Based upon the 2, the fit of the one-fact or model was also not acceptable with 2 (104, N = 399) = 694.53, p < .001, which is larger than the two-factor

PAGE 89

82 model. The SRMR for the one-factor model wa s .06, which indicated poorer fit compared with the two-factor model (SRMR = .05 for the two-factor model). The CFI of .873 for the one-factor model indicated that the fit wa s not as good as the two-factor model (CFI = .908). The RMSEA of .119 for the one-factor m odel was larger than the RMSEA for the two-factor model (.107), agai n suggesting that the one-facto r solution provided an even less acceptable fit. The standard ized factor loadings for the one-factor model ranged from .60 to .81 and were all statistically significantly different from zero ( p < .01). College of Education. Table 9 provides de scriptive statistics for the 16 observed variables used in the confirmatory factor analysis for the responses from students in the College of Education. The means for the 16 items ranged from 1.72 to 2.26 (median = 1.98) with standard deviations rangi ng from 0.79 to 1.13 (median = 0.89). The normality of the distributions for the 16 items was evaluated using univariate skewness and kurtosis measures for each of these variables. All skewness values were less than 1.0 with the smalle st skewness value of 0.21 and th e largest skewness value of 0.90 (median = 0.57), indicating that the it ems reflected acceptable symmetry. All kurtosis values were near zero for each of the items with values ranging from -1.05 to 0.24 (median = -0.40), which suggests that the p eak and tails of the distributions were similar to the normal curve. Multivariate normality evaluate using Mardia’s test of multinormality was not demonstrated in the Co llege of Education for the total group or by gender ( p <.001 in every case) so the assumption of multivariate normality was not tenable.

PAGE 90

83 Table 9 Descriptive Statistics for the 16 Obser ved Variables Used in the Two-Factor Confirmatory Factor Analyses of the St udy Anxiety Inventory for the College of Education (n=410) Item # Mean SD Skewness Kurtosis 1e 1.93 0.90 0.70 -0.19 2w 2.05 0.79 0.56 0.24 3w 2.26 0.87 0.21 -0.45 4e 1.98 0.94 0.61 -0.55 5e 1.81 0.89 0.90 0.07 6w 2.24 0.90 0.31 -0.55 7w 2.10 0.89 0.44 -0.45 8e 2.08 1.13 0.58 -1.05 9e 1.97 0.97 0.65 -0.57 10w 2.03 0.89 0.54 -0.36 11w 2.01 0.94 0.57 -0.56 12e 1.87 0.90 0.77 -0.20 13e 1.72 0.87 0.90 -0.07 14w 1.98 0.89 0.54 -0.47 15w 1.90 0.81 0.49 -0.23 16e 1.89 0.90 0.74 -0.22 Note : Emotionality (e) items are 1, 4, 5, 8, 9, 12, 13, 16; Worry (w) items are 2, 3, 6, 7, 10, 11, 14, 15 The 2 fit statistics for the CFA of the two-factor (worry, emotionality) model for the students from the College of Education indicated that the fit of the model was not acceptable, 2 (103, N = 410) = 483.58, p < .001. Alternative meas ures of fit were included because one of the limitations of the 2 is that it is sensitive to sample size. The SRMR of .044 indicated acceptable fit, wh ile the CFI of .927 and the RMSEA of .095 indicated less than acceptable fit. The unstandardized factor lo adings, excluding the one fi xed to 1.0 to identify the model, ranged from 1.08 to 1.29 (median = 1.11 ) for the factor of emotionality and from 1.03 to 1.27 (median = 1.22) for the factor of worry. All loadings were statistically

PAGE 91

84 significantly different from zero ( p < .01). An examination of the standardized factor loadings showed that the loadings ranged fr om .75 to .85 for the emotionality items and from .69 to .83 for the worry items. The co rrelation between emotionality and worry was .92. Sources of misfit of the model were e xplored by examining the modification indices that indicated the expected decrease in model fit chi square ( 2) that would result when a specific parameter, constraine d initially to zero, s ubsequently was freely estimated. The largest modification indices we re for covariances between errors. Because similarities in the item content for pairs of items may result in large error covariances, examination of the modification indices a nd wording of the pairs with chi square differences larger than the critical chi square of 6.64 ( p = .01) was carried out. Table 10 lists all pairs of items with changes in chi square that were statistically significant and shows the six pairs of items with th e highest chi-square differences ( 2 > 17). This value indicates the improvement in model fit if a covariance was posited between the errors of the items. The largest modification index for the m odel was for the covariance between the errors for items 6 and 3 ( 2 = 93.34). These items were very similar and shared the word “interfere”. Although the change in chi-square was substantially lower for the remaining five pairs of error covariances ( 2 ranged from 17.9 to 31.39), examination of those items revealed similar wording in the pairs of items. Items 3 and 6 were similar because the word “interfere” is the basis of the item and could therefore be responded to in the same way. Examination of the word ing in items 13, 14 and 15 (all worry items)

PAGE 92

85 revealed that “freezing up,” “mental block,” and “can’t get my brain to organize” may be sufficiently similar to suggest redundancies in the items. Examination of items 10 and 7 reveal that “can’t absorb the material” and “not being able to retain the material” are sufficiently similar that they may have been viewed by the respondents as nearly identical items. Item 10 may be related to item 3 becau se both address the thoughts that interfere with learning. The modification indices for the pairs of items are shown in Table 10.

PAGE 93

86 Table 10 Modification Indices for Error Covarian ces for the Study Anxiety Inventory for Responses from All Education Students (n=410) Items with errors covarying Chi-Square difference Pair 1: 3. I can’t keep my mind on the subj ect when studying for an exam because other thoughts interfere (Worry) 6. While studying for tests, other thoughts interfere with my learning (Worry) 93.34 Pair 5: 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 15. When I am studying for a test, I can’t get my brain to organize the information (Worry) 31.39 Pair 4: 13. I freeze up while studying for an important test (Emotionality) 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 30.35 Pair 6: 10. While studying for exams, I am stressed with thoughts like “I can’t absorb the material properly” (Worry) 7. While studying for a test, I worry about not being able to learn the material (Worry) 24.62 Pair 3: 10. While studying for exams, I am stressed with thoughts like “I can’t absorb the material properly” (Worry) 3. I can’t keep my mind on the subject wh en studying for an exam because other thoughts interfere (Worry) 23.32 Pair 2: 7. While studying for a test, I worry about not being able to learn the material (Worry) 5. While studying for exams, I have an uneasy, upset feeling (Emotionality) 19.64 14 with 7 15 with 13 14 with 5 10 with 6 9 with 6 9 with 8 11 with 2 18.73 18.32 17.99 17.08 16.09 14.44 13.24 15 with 7 10 with 9 7 with 3 9 with 3 2 with 1 4 with 3 9 with7 12.83 11.20 10.89 10.12 9.72 9.49 9.34 4 with 2 14 with 3 6 with 4 11 with 3 9 with 4 13 with 7 9 with 2 7 with 2 9.11 8.60 8.19 8.09 7.62 7.56 7.18 6.89 The correlation of .92 between the emotionali ty and worry factors was statistically significantly different from 0 ( p < .001). The strength of the correlation between the two factors led the researcher to entertain the possibility of a one-factor model underlying the Study Anxiety Inventory (see Figure 6). Based upon the 2, the fit of the one-factor model was also not acceptable with 2 (104, N = 428 = 692.48, p < .001, which is larger

PAGE 94

87 than the 2 for the two-factor model. The SR MR for the one-factor model was .058, which indicated a poorer fit compared with the two-factor model (SRMR = .044 for the two-factor model). The CFI of .877 for the one -factor model indicated that the fit was not as good as the two-factor model (CFI = .927). The RMSEA of .115 for the one-factor model was larger than the RMSEA for the tw o-factor model (.095), again suggesting that the one-factor solution provided an even less acceptable fit. The standardized factor loadings for the one-factor model ranged fr om .64 to .83 and were all statistically significantly different from zero ( p < .01). Table 11 presents an overview of the f it indices for the oneand two-factor confirmatory factor analysis of the St udy Anxiety Inventory across four colleges and shows that chi-square values for both mode ls indicated less than acceptable fit for each college. The results, however, were consisten tly poorer for the one-f actor model than for the two-factor model. Although the SRMRs fo r both the hypothesized two-factor and one-factor model for each college were less th an .08, indicating acceptabl e fit, the indices indicated poorer fit for th e one-factor model (SRMRs ranged from .050 to .060) for each college than for the two-factor model (SRM Rs ranged from .047 to .050). The CFIs for the two-factor model for each of the college s indicated acceptable fit ranging from .908 to .927 while the one-factor model indicated le ss than acceptable fit for each of the colleges except the College of Engineer ing with CFIs ranging from .868 to .902, and even that CFI was lower for the one-factor model than the two-f actor model. Although the RMSEA values for both the two-factor (.095 to .092) and one-f actor model (.109 to .126) indicated less than acceptable fit for each college, the results were consistently

PAGE 95

88 poorer for the one-factor model than for th e two-factor model. So, although for each college the correlation between the two factors of emotionality and worry ranged from .87 to .92 (.87 to .91 for males, and .85 to .92 for females), it still seems more acceptable to consider that the two-factor model is a be tter fit than the one-f actor model. Although multivariate normality was not found to be te nable, rerunning the data using maximum likelihood estimation with standard errors and a mean-adjusted chi-square test statistic (MLM in Mplus), which is robust to this vi olation, produced results that pointed to the same conclusions. Table 11 Fit Indices for the Confirmatory Factor A nalysis of the Hypothesized Two-Factor and One-Factor Model for the Study Anxi ety Inventory Across Four Colleges Arts and Sciences Business Education Engineering 2-factor 1-factor 2-factor 1-factor 2-factor 1-factor 2-factor 1-factor 2 772.52 1187.42 527.43694.53483.58692.48 521.82 613.45 SRMR .047 .060 .050.060.044.058 .050 .050 CFI .918 .868 .908.873.927.877 .912 .902RMSEA .099 .126 .102.119.095.115 .097 .109 Research Question 2: Factorial Invariance by Gender To evaluate the equivalence of the tw o-factor measurement model underlying the SAI, multigroup CFA was used. The data were divided by gender, and variancecovariance matrices were calculated for each group (males, females). The fit of the twofactor model was evaluated separately for males and females followed by the evaluation of the invariance of the model for males a nd females. Invariance tests were conducted using a multiple group analysis to test the equality of the factor loadings, residual variances, factor variances, and the covariance between factors by gender. Equality

PAGE 96

89 restrictions were imposed across males and fe males for tests of invariance. A Chi-squaredifference test for relative fit for a nested sequence of models was used for this test. Analyses were conducted sepa rately for each college and are presented in the next section. College of Arts and Sciences. Table 12 provides descrip tive statistics by gender for the 16 observed variables used in the conf irmatory factor analysis. The responses for each item ranged from one to four and the means for the 16 items for males ranged from 1.42 to 2.11 (median = 1.72) with standard deviations ranging from 0.73 to 1.93 (median = 0.84), and for females ranged from 1.79 to 2.28 (median = 2.04) with standard deviations ranging from 0.78 to 1.12 (median = 0.92). Examination of the kurtosis and skewness values showed that for each of the items both kurtosis and skewness values were close to zero for both males and female which suggests that th e peak and tails of the distribution were similar to the norm al curve and reflected acceptable symmetry. The mean of every item score for the ma les was significantly lower than the mean item score for the females. Effect sizes were calculated using: Effect sizes for the indiv idual items are disp layed in Table 12 and ranged from fairly low at -0.13 to moderate at -0.36 with a median effect size of -0.24. Th e effect size of the overall scale was moderate at 0.32 and the effect sizes of the subscales were moderate at 0.30 for emotionality and 0.29 for worry.

PAGE 97

90 Table 12 Descriptive Statistics for the 16 Obser ved Variables Used in the Two-Factor Confirmatory Factor Analyses of the Study An xiety Inventory for the College of Arts and Sciences by Gender (nM=215, nF=445) Males Females Males Females Males Females Mean SD Mean SD t es Skewness Kurtosis 1e 1.76 0.79 2.00 0.92 -3.46** -0.20 0.86 0.71 0.30 -0.17 2w 1.84 0.76 2.11 0.78 -4.24** -0.25 0.73 0.44 0.41 0.09 3w 2.09 0.84 2.27 0.88 -2.54* -0.15 0.49 0.32 -0.26 -0.47 4e 1.76 0.87 2.03 0.95 -3.62** -0.21 0.91 0.52 0.01 -0.66 5e 1.65 0.83 1.81 0.89 -2.27* -0.13 1.18 0.87 0.67 0.01 6w 2.11 0.89 2.28 0.89 -2.30* -0.14 0.52 0.31 -0.38 -0.42 7w 1.71 0.81 2.15 0.92 -6.25** -0.36 0.96 0.45 -0.31 -0.47 8e 1.73 0.93 2.16 1.12 -5.20** -0.30 1.02 0.48 -0.05 -1.07 9e 1.61 0.86 2.01 0.99 -5.33** -0.31 1.25 0.62 0.61 -0.59 10w 1.72 0.86 2.14 0.93 -5.72** -0.33 1.03 0.41 0.28 -0.54 11w 1.75 0.91 2.10 0.99 -4.50** -0.26 0.96 0.44 -0.10 -0.73 12e 1.69 0.78 1.96 0.95 -3.87** -0.22 0.96 0.63 0.36 -0.50 13e 1.42 0.73 1.79 0.91 -5.62** -0.32 1.76 0.96 2.51 -0.18 14w 1.69 0.84 2.04 0.92 -4.86** -0.28 1.11 0.54 0.57 -0.41 15w 1.67 0.83 1.91 0.84 -3.47** -0.27 1.13 0.57 0.62 -0.20 16e 1.66 0.84 1.95 0.92 -4.03** -0.30 1.09 0.65 0.37 -0.35 SAI 1.74 0.65 2.05 0.71 -5.57** -0.32 1.03 0.53 0.55 -0.11 SAe 1.66 0.69 1.97 0.79 -5.15** -0.30 1.11 0.70 0.55 -0.10 SAw 1.83 0.68 2.12 0.71 -5.06** -0.29 .912 0.47 0.50 -0.05 Note : es = effect size, p <.05, ** p < .01. Emotionality (e) items are 1, 4, 5, 8, 9, 12, 13, 16; Worry (w) items are 2, 3, 6, 7, 10, 11, 14, 15; SAI = Study Anxiety Inventory Index Score, SAe=Study Anxiety Emotionality subscale score, SAw=Study Anxiety Inventory Worry subscale score. The 2 fit statistics for the CFA of the two-factor (worry, emotionality) model for male and female students from the College of Arts and Sciences indicated that the fit of the model was not acceptable, 2 (103, NM=215) = 337.64 and 2 (103, NF=445) = 617.76, respectively. The same alternative measur es of fit were included because the 2 is sensitive to sample size. The SRMRs for the male and female groups of .050 and .053 respectively indicated sim ilar and acceptable fit, while the CFIs of .915 and .904 respectively, and the RMSEAs of .103 and .106 both indicated less than acceptable fit.

PAGE 98

91 The unstandardized factor loadings for th e responses, excluding the one fixed to 1.0, ranged from 1.07 to 1.40 with a mean of 1.30 ( SD = 0.12) for the males and from 1.17 to 1.38 with a mean of 1.27 ( SD = 0.09) for the females for the factor of emotionality. Loadings ranged from 0.82 to 1.33 with a mean of 1.16 ( SD = 0.17) for the males and from 1.14 to 1.50 with a mean of 1.37 ( SD = 0.13) for the females for the factor of worry. An examination of the standardized factor loadings showed that the loadings of all items on the emotionality and worry scale we re statistically signi ficant as hypothesized (>.68 for males and >.54 for females). The correlation between emo tionality and worry was .90 for males and .85 for females. Sources of misfit of the models for males and females were further explored by comparing the modification indices that indicat ed the expected decrease in model fit chi square ( 2) that would result when a specific para meter, constrained initially to zero, subsequently was freely estimated. As with th e combined sample of Arts and Sciences students, the largest expected chi square ch ange was between covari ances of the errors for items 3 and 6 ( 2=70.96 for males and 116.67 for females). Of the eight pairs of items with the highest chi-square change for the females, three of the error covariances were also found in the list of the highest eight for males (6 and 3, 14 and 13, and 15 and 14).

PAGE 99

92 Table 13 Modification Indices for Error Covarian ces for the Study Anxiety Inventory for Responses from Male Arts and Science Students (n = 225) Items with errors covarying Chi-Square Difference Pair 1: 3. I can’t keep my mind on the subject when studying for an exam because other thoughts interfere (Worry) 6. While studying for tests, other thoughts interfere with my learning (Worry) 116.67 Pair 2: 13. I freeze up while studying for an important test (Emotionality) 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 38.42 Pair 3: 9. I feel panicky when studying for an important exam (Emotionality) 12. While studying for exams, I feel very tense (Emotionality) 38.23 Pair 4: 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 15. When I am studying for a test, I can’t get my brain to organize the information (Worry) 33.21 Pair 5: 1. I feel very uneasy just before starting to study for an exam (Emotionality) 2. While I am studying for an exam I often think “I’m not getting this” (Worry) 26.77 Pair 6: 4. Even when I have plenty of time I feel nervous when I try to study for an exam (Emotionality) 2. While I am studying for an exam I often think “I’m not getting this” (Worry) 23.25 Pair 7: 4. Even when I have plenty of time I feel nervous when I try to study for an exam (Emotionality) 2. While studying for exams, I feel very tense (Emotionality) 23.37 Pair 8: 10. While studying for exams, I am stressed with thoughts like “I can’t absorb the material properly” (Worry) 7. While studying for a test, I worry about not being able to learn the material (Worry) 22.72 15 with 13 10 with 9 15 with 1 4 with 3 14 with 7 6 with 4 9 with 4 21.31 19.66 18.51 16.20 16.14 14.43 14.25 10 with 6 13 with 7 10 with 3 9 with 3 12 with 6 7 with 2 11 with 9 12.96 12.85 11.95 11.69 11.29 10.30 9.32 12 with 3 5 with 1 11 with 5 5 with 2 15 with 10 13 with 12 15 with 9 13 with 6 8.90 8.80 8.64 8.41 7.49 7.42 7.18 6.70

PAGE 100

93 Table 14 Modification Indices for Error Covarian ces for the Study Anxiety Inventory for Responses from Female Arts and Science Students (n = 410) Items with errors covarying Chi-Square Difference Pair 1: 3. I can’t keep my mind on the subj ect when studying for an exam because other thoughts interfere (Worry) 6. While studying for tests, other thoughts interfere with my learning (Worry) 70.96 Pair 2: 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 15. When I am studying for a test, I can’t get my brain to organize the information (Worry) 18.69 Pair 3: 9. I feel panicky when studying for an important exam (Emotionality) 8. I wish studying for tests did not upset me so much (Emotionality) 17.21 Pair 4: 12. While studying for exams, I feel very tense (Emotionality) 5. While studying for exams, I have an uneasy, upset feeling (Emotionality) 16.90 Pair 5: 3. I can’t keep my mind on the subj ect when studying for an exam because other thoughts interfere (Worry) 6. I feel jittery while studying for important exams (Emotionality) 14.69 Pair 6: 7. While studying for a test, I worry about not being able to learn the material (Worry) 8. I wish studying for tests did not upset me so much (Emotionality) 12.17 Pair 7: 13. I freeze up while studying for an important test (Emotionality) 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 12.00 Pair 8: 13. I freeze up while studying for an important test (Emotionality) 8. I wish studying for tests did not upset me so much (Emotionality) 11.91 15 with 9 5 with 4 9.92 9.27 11 with 1 16 with 4 8.78 8.45 16 with 11 16 with 7 7.08 6.70 After looking at the CFA model separa tely for males and females, multigroup CFA was conducted to compare the paramete r estimates (factor loadings, residual variances, covariance between factors, and factor variances) for males and females. Models were tested sequentially beginning with the least restrictive model and continuing with the addition of specific constraints. Table 15 contains the fit indice s corresponding to each of the models that were tested in th e confirmatory factor analysis to evaluate

PAGE 101

94 factorial invariance of the scores on the Study Anxiety Inventory by gender. Model 1 was the baseline model in which there were no equality constraints across the male and female groups. For this model, factor load ings, residual variances (i.e., uniquenesses), and factor variances and covariance were freely estimated in each group (males and females). Model 2 is a more restrictive mode l that imposes equality constraints on the loadings by gender. Because two factors were hypothesized, one loading from each factor was fixed to 1.0 to identify the model. This le ft 14 pairs of loadings free to vary, seven from each factor. This results in an increase in the degrees of freedom for Model 2 of 14 and a critical value of chi square of 29.121 ( p = .01). Model 3 adds additional restrictions by imposing equality constraints on the item residual variances for males and females. Model 4 adds an additional equality constraint restricting the factor covariance to be equal across gender and Model 5 imposes e quality constraints on the factor variances across males and females. Table 15 shows the different models tested to determine invariance. To test the hypothesis of equal loadings across gender, th e more restrictive Model 2 is compared with Model 1 (loadings freely estimated in each group). The change in the chi square value of 30.74 relative to the cha nge in degrees of freedom ( df = 14) suggested that invariance of the factor loadings may be unt enable (critical value of chi square for 14 degrees of freedom at p = .01 is 29.12).

PAGE 102

95 Table 15 Goodness-of-Fit Indices for Models Tested for Invariance of Scores on the Study Anxiety Inventory by Gender (n = 660 A & S Students, nM=215, nF=445) Model Model # 2 df 2 df p 1. Baseline 1 955.39206 2. Equal Loadings 2 986.13220 30.74 14<.01 3. Equal Residual Variances 3 1082.78236 96.65 16<.01 3a. Equal Residual Variance for all but 4 items 3a 1031.19232 45.06 12<.01 4. Equal Factor Covariances 4 1032.35233 1.16 1>.01 5. Equal Factor Variances 5 1044.90235 12.55 2<.01 Table 16 Goodness-of-Fit Indices for Inva riance of Loadings on the St udy Anxiety Inventory by Gender (n = 660 A & S Students, nM=215, nF=445) Item # 22p 3w 960.47 5.08 .0224 4e 958.26 2.87 .0902 5e 957.32 1.93 .1648 6w 957.24 1.85 .1738 7w 958.40 3.01 .0828 8e 955.42 0.03 .8625 9e 955.40 0.01 .9203 10w 956.81 1.42 .2334 11w 956.71 1.32 .2506 12e 956.14 0.75 .3865 13e 956.74 1.35 .2453 14w 957.94 2.55 .1103 15w 956.22 0.83 .3623 16e 956.72 1.33 .2488 Note : degrees of freedom (df) = 207; change in df = 1; p -value shows 4 decimal places to compare with .05/14=.0035; 2 for the baseline model was 955.39, df =206.

PAGE 103

96 Because the overall hypothesis of equal loadings was rejected ( p < .01), follow-up testing of each item loading was done to id entify the source of the difference. A .0035 (05/14 = .0035) level of statisti cal significance was used to c ontrol the type I error rate. Results are displayed in Table 16. These re sults indicated that no item loading was significantly different across gender. The p -value closest to being statistically significant was .0224 and was for item 3 (“I can’t keep my mind on the subject when studying for an exam because other thoughts interfere”). The next model that was tested (Model 3) imposed equality constraints on the error variances. The 2 was 96.65 relative to a change in degrees of freedom of 16 indicating that inva riance of errors was not tenable.

PAGE 104

97 Table 17 Goodness-of-Fit Indices for In variance of Error Variances on the Study Anxiety Inventory by Gender (n = 660 A & S Students, nM=215, nF=445) Item # 22p 1e 993.99 7.86 .0051 2w 992.25 6.12 .0134 3w 987.07 0.94 .3323 4e 1000.85 14.72 .0001 5e 990.51 4.38 .0364 6w 987.58 1.45 .2285 7w 995.24 9.11 .0025 8e 1000.80 14.67 .0001 9e 989.86 3.73 .0534 10w 991.32 5.19 .0227 11w 990.17 4.04 .0444 12e 987.52 1.39 .2384 13e 990.12 3.99 .0458 14w 989.90 3.77 .0522 15w 986.46 0.33 .5657 16e 997.50 11.37 .0007 Note : degrees of freedom (df) = 221; the change in df = 1; p -value shows 4 decimal places to compare with .05/16 = .0031; 2 for Model 2 was 986.13, df =220. Because the overall hypothesis of eq ual error variances was rejected ( p < .01), follow-up testing of each item error variance was done to identify the source of the difference. A .0031 (.05/16 = .0031) level of stat istical significance was used to control the type I error rate. Results are displayed in Table 17. These results indicated that item error variance was statistically significantly different across gender for items 7, a worry item, and items 4, 8, and 16, three emotionality items. Model 3a removes the restrictions on the four items that demonstrated an in equality of residual variance. Model 4 and Model 5 were then run disallowing the rest rictions on the four residual variances. When the covariance between the worry a nd emotionality factors was set equal, the resulting chi square was 1032.35, which repr esented a change of 1.16 from Model 3a.

PAGE 105

98 The change in chi square was not statistically significant at the .01 level indicating that invariance of the covariance between the tw o factors was tenable. When the factor variances of emotionality and worry were se t equal across the male and female groups, the resulting chi square was 1044.90 repr esenting a change of 12.55, which was statistically significant at the .01 level indicating that invariance of this parameter was not tenable. Because the overall hypothesis of equa l factor variances was rejected ( p < .01), follow-up testing of each variance was done to identify the source of the difference. A .025 (05/2 = .025) level of statistic al significance was used to c ontrol the type I error rate. When the factor variance of emotionality wa s allowed to vary across gender, the resulting chi square was 1032.37, a chi square change of 0.05, which was not statistically significant at the .01 level indi cating that invariance of emot ionality was tenable. When the factor variance of worry was allowed to vary across, the resulting chi square was 1039.02, a chi square change of 6.67, which was statistically significant at the .01 level indicating that invari ance of worry was not tenable. College of Engineering. Table 18 provides descriptive statistics by gender for the 16 observed variables used in the confirmatory factor analysis. The responses for each item ranged from one to four and the means for the 16 items for males ranged from 1.50 to 2.05 (median = 1.72) with standard devi ations ranging from 0.72 to 0.94 (median = 0.83), and for females the means ranged from 1.70 to 2.15 (median = 1.98) with standard deviations ranging from 0.73 to 1.09 (median = 0.84). For the emotionality scale, the

PAGE 106

99 means for every item were significantly lower for males vs. females. For the worry scale, the means for five of the eight items were significantly lower for males vs. females. Effect sizes for the individual items ar e displayed in Table 18 and ranged from fairly low at -0.04 to moderate at -0.39 with low median effect size of -0.18. The effect size of the overall scale was low at -0.11 a nd the effect sizes of the subscales were moderate at -0.12 for emotionality and low at -0.07 for worry.

PAGE 107

100 Table 18 Descriptive Statistics for the 16 Obser ved Variables Used in the Two-Factor Confirmatory Factor Analyses of the St udy Anxiety Inventory for the College of Engineering by Gender (nM=243, nF=192) Males Females Males Females Males Females Mean SD Mea n SD t es Skewness Kurtosis 1e 1.79 0.84 1.99 0.88 -2.40** -0.20 0.85 0.48 0.47 -0.60 2w 1.89 0.75 2.02 0.73 -1.82* -0.13 0.48 0.46 0.42 0.20 3w 2.00 0.88 2.15 0.78 -1.88* -0.15 0.49 0.46 -0.27 0.02 4e 1.71 0.80 1.96 0.84 -3.15** -0.25 0.97 0.57 0.81 -0.29 5e 1.57 0.72 1.70 0.83 -1.72* -0.13 0.94 1.06 0.52 0.52 6w 2.05 0.93 2.11 0.84 -0.71 -0.06 0.50 0.54 -0.25 -0.14 7w 1.75 0.80 1.96 0.81 -2.70** -0.21 0.68 0.61 -0.01 -0.01 8e 1.73 0.94 2.01 1.09 -2.82** -0.28 1.07 0.71 0.32 -0.84 9e 1.60 0.83 1.99 0.98 -4.41** -0.39 1.19 0.62 0.87 -0.69 10w 1.73 0.85 2.02 0.83 -3.58** -0.29 0.88 0.53 0.20 -0.21 11w 1.74 0.89 2.02 0.95 -3.14** -0.28 0.96 0.57 0.33 -0.64 12e 1.66 0.82 1.93 0.89 -3.25** -0.27 1.02 0.68 0.68 -0.35 13e 1.50 0.76 1.70 0.86 -2.53** -0.20 1.37 1.21 1.49 0.86 14w 1.71 0.85 1.76 0.79 -0.63 -0.05 0.96 0.84 0.38 0.23 15w 1.68 0.85 1.72 0.77 -0.51 -0.04 1.08 0.80 0.69 0.03 16e 1.62 0.79 1.86 0.88 -2.95** -0.24 0.99 0.75 0.48 -0.25 SAI 1.74 0.63 1.94 0.64 -3.26** -0.20 0.82 0.71 1.04 0.22 SAe 1.65 0.66 1.89 0.72 -3.58** -0.24 0.95 0.74 1.05 -0.05 SAw 1.83 0.66 1.98 0.64 -2.39** -0.15 0.71 0.69 0.75 0.22 Note : p <.05, ** p < .01. es = effect size. Emotionality (e) items are 1, 4, 5, 8, 9, 12, 13, 16; Worry (w) items are 2, 3, 6, 7, 10, 11, 14, 15 Kurtosis and skewness values for each of the items were close to zero for both males and females, which suggests that the peak and tails of the dist ribution were similar to the normal curve and reflected acceptable symmetry. The 2 fit statistics for the CFA of the two-factor (worry, emotionality) model for male and female students from the College of Engineer ing indicated that the fit of the model was not acceptable, 2 (103, NM=240) = 359.11 and 2 (103, NF=188) = 345.52. respectively. The same alternative measur es of fit were included because the 2 is sensitive to sample size. The SRMRs for the male and female groups of .053 and .060,

PAGE 108

101 respectively, indicated similar and acceptabl e fit, while the CFI of .907 for males and .882 for females, and the RMSEAs of .102 for males and .112 for females indicated less than acceptable fit. The unstandardized factor loadings for the items, excluding the one fixed to 1.0, ranged from 1.12 to 1.36 with a mean of 1.26 ( SD = 0.10) for the males and from 1.28 to 1.65 with a mean of 1.44 ( SD = 0.15) for the females for the factor of emotionality. These loadings ranged from 1.14 to 1.39 with a mean of 1.26 ( SD = 0.09) for the males and from 0.99 to 1.38 with a mean of 1.18 ( SD = 0.13) for the females for the factor of worry. Standardized factor loadings of all items on the emotiona lity and worry scale were statistically significant as hypothesized (>.62 for males and >.58 for females). The correlation between emotionality and worry was .90 for males and .88 for females. Sources of misfit of the models for males and females were further explored by comparing the modification indices that indicat ed the expected decrease in model fit chi square ( 2) that would result when a specific para meter, constrained initially to zero, subsequently was freely estimated. As with th e combined sample of engineering students, the largest expected chi square change was be tween covariances of the errors for items 3 and 6 ( 2=59.99 for males and 67.08 for females). Of the eight pairs of items with the highest chi-square change for the females, two of the correlated errors were also found in the list of the highest eight for males (6 and 3, and 15 and 14).

PAGE 109

102 Table 19 Modification Indices for Error Covarian ces for the Study Anxiety Inventory for Responses from Male Engineering Students (n = 240) Items with errors covarying Chi-Square Difference Pair 1: 3. I can’t keep my mind on the subj ect when studying for an exam because other thoughts interfere (Worry) 6. While studying for tests, other thoughts interfere with my learning (Worry) 59.99 Pair 2: 1. I feel very uneasy just before starting to study for an exam (Emotionality) 2. While I am studying for an exam I often think “I’m not getting this” (Worry) 25.67 Pair 3: 4. Even when I have plenty of time, I feel nervous when I tr y to study for an exam (Emotionality) 1. I feel very uneasy just before starting to study for an exam (Emotionality) 16.44 Pair 4: 12. While studying for exams, I feel very tense (Emotionality) 15. When I am studying for a test, I can’t get my brain to organize the information (Worry) 14.86 Pair 5: 11. I worry so much when I study for a test that I do things that distract me (Worry) 2. While studying for exams, I feel very tense (Emotionality) 12.10 Pair 6: 3. I can’t keep my mind on the subj ect when studying for an exam because other thoughts interfere (Worry) 2. While I am studying for an exam I often think “I’m not getting this” (Worry) 11.41 Pair 7: 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 1. I feel very uneasy just before starting to study for an exam (Emotionality) 10.88 Pair 8: 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 15. When I am studying for a test, I can’t get my brain to organize the information (Worry) 9.69 14 with 10 14 with 13 4 with 3 8.45 8.27 8.11 11 with 10 9 with 6 16 with 15 8.10 7.69 7.40 13 with 10 14 with 7 9 with 7 6.99 6.87 6.69

PAGE 110

103 Table 20 Modification Indices for Error Covarian ces for the Study Anxiety Inventory for Responses from Female Engi neering Students (n = 188) Items with errors covarying Chi-Square Diff Pair 1: 3. I can’t keep my mind on the subj ect when studying for an exam because other thoughts interfere (Worry) 6. While studying for tests, other thoughts interfere with my learning (Worry) 67.08 Pair 2: 5. While studying for exams, I have an uneasy, upset feeling (Emotionality) 4. Even when I have plenty of time, I feel nervous when I try to study for an exam (Emotionality) 23.42 Pair 3: 15. When I am studying for a test, I can’t get my brain to organize the information (Worry) 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 15.63 Pair 4: 12. While studying for exams, I feel very tense (Emotionality) 9. I feel panicky when studying for an important exam (Emotionality) 13.56 Pair 5: 12. While studying for exams, I feel very tense (Emotionality) 5. While studying for exams, I have an uneasy, upset feeling (Emotionality) 13.32 Pair 6: 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 12. While studying for exams, I feel very tense (Emotionality) 12.72 Pair 7: 13. I freeze up while studying for an important test (Emotionality) 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 10.34 Pair 8: 9. I feel panicky when studying for an important exam (Emotionality) 10. While studying for exams, I am stressed with thoughts like “I can’t absorb the material properly” (Worry) 9.52 15 with 2 12 with 7 9 with 3 8.76 8.33 8.13 9 with 5 11 with 6 13 with 1 7.74 7.68 7.49 4 with 2 5 with 2 16 with 8 7.10 6.76 6.71 After looking at the CFA model separa tely for males and females, multigroup CFA was conducted to compare the paramete r estimates (factor loadings, residual variances, covariance between factors, and factor variances) for males and females. Models were tested sequentially beginning with the least restrictive model and continuing with the addition of sp ecific constraints.

PAGE 111

104 Table 21 shows the different models tested to evaluate invariance. The constraint of equal loadings by gender is the more rest rictive Model 2. The change in the chi square value of 6.38 relative to the cha nge in degrees of freedom ( df = 14) suggested that invariance of the factor loadings is tenable.

PAGE 112

105 Table 21 Goodness-of-Fit Indices for Models Tested for Invariance of Scores on the Study Anxiety Inventory by Gender (n = 428 Engineering Students, nM=240, nF=188) Model Model # 2 df 2 df p 1. Baseline 1 704.62206 2. Equal Loadings 2 710.93220 6.38 14>.01 3. Equal Residual Variances 3 759.34236 48.41 16<.01 3a. Equal Residual Variance for all but 2 items 3a 736.87234 25.94 14>.01 4. Equal Factor Covariances 4 736.88235 0.01 1>.01 5. Equal Factor Variances 5 744.61237 7.73 2>.01 The next model that was tested (Model 3) imposed equality constraints on the error variances. The 2 was 48.41 relative to a change in degrees of freedom of 16 indicating that inva riance of errors was not tenable.

PAGE 113

106 Table 22 Goodness-of-Fit Indices for In variance of Error Variances on the Study Anxiety Inventory by Gender (n = 428 Engineering Students, nM=240, nF=188) Item # 22 p 1e 712.10 1.17 .2794 2w 711.37 0.44 .5071 3w 714.26 3.33 .0680 4e 710.97 0.04 .8415 5e 713.79 2.86 .0908 6w 711.38 0.45 .5023 7w 711.37 0.44 .5071 8e 716.45 5.52 .0188 9e 719.76 8.83 .0030 10w 711.16 0.23 .6315 11w 713.53 2.60 .1069 12e 711.56 0.63 .4274 13e 722.83 11.9 .0006 14w 710.95 0.02 .8875 15w 716.30 5.37 .0205 16e 713.41 2.48 .1153 Note : degrees of freedom (df) = 221; change in df = 1; p -value shows 4 decimal places to compare with .05/14=.0035; 2 for model 2 was 710.93, df =220; Emotionality (e) items are 1, 4, 5, 8, 9, 12, 13, 16; Worry (w) items are 2, 3, 6, 7, 10, 11, 14, 15 Because the overall hypothesis of eq ual error variances was rejected ( p < .01), follow-up testing of each item error variance was done to identify the source of the difference. A .0031 (.05/16 = .0031) level of stat istical significance was used to control the type I error rate. Results are displayed in Table 22. These results indicated that item error variance was statistically significantly different across gender for items 9 and 13, two emotionality items. Mode l 3a removes the restricti ons on the two items that demonstrated an inequality of residual variance. Models 4 and 5 were then run and compared with Model 3a which constrained the factor loadings and 14 out of the 16 residual variances to be equal across gender.

PAGE 114

107 When the covariance between the two fact ors was set equal, the resulting chi square was 736.88, which represented a change of 0.01 from Model 3a. The change in chi square was not statistically si gnificant at the .01 level indica ting that invariance of the covariance between the two factors was te nable. When the factor variances of emotionality and worry were set equal across the male and female groups, the resulting chi square was 744.61 representing a change of 7.73, which was not statistically significant at the .01 level indi cating that invariance of thes e parameters was tenable. College of Business. Table 23 provides descriptive statistics by gender for the 16 observed variables used in the confirmatory factor analysis. The responses for each item ranged from one to four and the means for the 16 items for males ranged from 1.64 to 2.20 (median = 1.87) with standard deviati ons ranging from 0.70 to 0.98 (median = 0.83), and for females ranged from 1.77 to 2.23 (med ian = 1.98) with standard deviations ranging from 0.77 to 1.11 (median = 0.90). For the emotionality scale, the means for four of the eight items were signifi cantly lower for males vs. females. For the worry scale, the means for two of the eight items were significantly lower for males vs. females. Effect sizes for the individual items were low and (see Table 23) and ranged from -0.17 to -0.21 with a low median effect si ze of -0.18. There was no significant difference between males and females for the overall scale or the worry subscale and the effect size for the emotionality scale was low at -0.18.

PAGE 115

108 Table 23 Descriptive Statistics for the 16 Obser ved Variables Used in the Two-Factor Confirmatory Factor Analyses of the Study Anxiety Inventory for the College of Business by Gender (nM=216, nF=193) Males Females Males Females Males Females Mean SD Mea n SD t es Skewness Kurtosis 1e 1.95 0.87 1.93 0.82 0.24 0.02 0.52 0.77 -0.35 0.30 2w 1.93 0.70 2.03 0.77 -1.37 -0.10 0.43 0.57 0.10 0.24 3w 2.20 0.84 2.23 0.87 -0.35 -0.02 0.20 0.30 0.61 -0.56 4e 1.82 0.80 2.00 0.95 -2.06* -0.14 0.72 0.63 -0.04 -0.53 5e 1.65 0.76 1.79 0.87 -1.72* -0.12 0.94 0.90 0.18 0.09 6w 2.18 0.87 2.19 0.90 -0.11 -0.01 0.43 0.36 -0.41 -0.61 7w 1.90 0.87 2.06 0.86 -1.87* -0.13 0.67 0.53 -0.13 -0.30 8e 1.88 0.98 2.08 1.11 -1.92* -0.14 0.84 0.62 -0.36 -0.89 9e 1.86 0.87 1.97 0.99 -1.19 -0.08 0.68 0.67 -0.43 -0.54 10w 1.89 0.83 2.05 0.98 -1.77* -0.12 0.65 0.52 -0.17 -0.65 11w 1.92 0.92 1.99 0.98 -0.74 -0.05 0.74 0.58 -0.33 -0.64 12e 1.75 0.79 1.91 0.95 -1.84* -0.13 0.87 0.78 0.28 -0.19 13e 1.64 0.80 1.77 0.91 -1.53 -0.11 1.08 0.85 0.40 -0.19 14w 1.85 0.83 1.91 0.86 -0.72 -0.05 0.84 0.58 0.23 -0.28 15w 1.79 0.77 1.87 0.83 -1.01 -0.13 0.76 0.53 0.21 -0.38 16e 1.80 0.85 1.91 0.93 -1.24 -0.09 0.81 0.69 -0.11 -0.33 SAI 1.89 0.63 1.99 0.71 -1.50 -0.11 0.48 0.63 -0.39 -0.22 SAe 1.80 0.69 1.93 0.78 -1.78* -0.12 0.61 0.68 -0.30 -0.42 SAw 1.97 0.64 2.04 0.71 -1.04 -0.07 0.34 0.57 -0.56 -0.16 Note : p <.05, ** p < .01. es = effect size, Emotionality (e) items are 1, 4, 5, 8, 9, 12, 13, 16; Worry (w) items are 2, 3, 6, 7, 10, 11, 14, 15 Examination of the kurtosis and skewness va lues showed that the value of each of the items was close to zero for both males a nd females, which suggests that the peak and tails of the distribution are similar to the normal curve and reflected acceptable symmetry. The 2 fit statistics for the CFA of the two-factor (worry, emotionality) model for male and female students from the College of Business indicated that the fit of the model was not acceptable, 2 (103, NM=210) = 273.61 and 2 (103, NF=189) = 434.36, respectively. The same alternative measur es of fit were included because the 2 is

PAGE 116

109 sensitive to sample size. The SRMRs for the male and female groups of .052 and .059 respectively, each indicated similar and accepta ble fit. The CFI of .921 for the males and .869 for the females indicated less than accepta ble fit. The RMSEAs of .089 and .089 for both indicated less than acceptable fit. The unstandardized factor loadings for th e responses, excluding the one fixed to 1.0, ranged from 1.04 to 1.32 with a mean of 1.18 ( SD = 0.10) for the males and from 1.23 to 1.47 with a mean of 1.36 ( SD = 0.08) for the females for the factor of emotionality. These loadings ranged fr om 1.09 to 1.56 with a mean of 1.29 ( SD = 0.16) for the males and from 1.05 to 1.44 with a mean of 1.26 ( SD = 0.14) for the females for the factor of worry. An examination of the standardized factor loadings showed that the loadings of all items on the worry scale were statistically si gnificant as hypothesized (>.60 for males and >.67 for females). The correlation between em otionality and worry was .87 for males and .85 for females. Sources of misfit of the models for males and females were further explored by comparing the modification indices that indicat ed the expected decrease in model fit chi square ( 2) that would result when a specific para meter, constrained initially to zero, subsequently was freely estimated. As with the combined sample of business students, the largest expected chi square change was be tween covariances of the errors for items 3 and 6 ( 2=45.95 for males and 31.13 for females). Of the eight pairs of items with the highest chi-square change for the females, two of the correlated errors were also found in the list of the highest eight for males (6 and 3, and 15 and 14).

PAGE 117

110 Table 24 Modification Indices for Error Covarian ces for the Study Anxiety Inventory for Responses from Male Business Students (n = 210) Items with errors covarying Chi-Square Difference Pair 1: 3. I can’t keep my mind on the subj ect when studying for an exam because other thoughts interfere (Worry) 6. While studying for tests, other thoughts interfere with my learning (Worry) 45.95 Pair 2: 13. I freeze up while studying for an important test (Emotionality) 6. While studying for tests, other though ts interfere with my learning (Worry) 15.45 Pair 3: 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 7. While studying for a test, I worry about not being able to learn the material (Worry) 13.76 Pair 4: 12. While studying for exams, I feel very tense (Emotionality) 2. While I am studying for an exam I often think “I’m not getting this” (Worry) 12.37 Pair 5: 1. I feel very uneasy just before starting to study for an exam (Emotionality) 2. While I am studying for an exam I often think “I’m not getting this” (Worry) 10.38 Pair 6: 10. While studying for exams, I am stressed with thoughts like “I can’t absorb the material properly (Worry) 6. While studying for tests, other though ts interfere with my learning (Worry) 8.19 Pair 7: 15. When I am studying for a test, I can’t get my brain to organize the information (Worry) 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 9.91 Pair 8: 10. While studying for exams, I am stressed with thoughts like “I can’t absorb the material properly” (Worry) 5. While studying for exams, I have an uneasy, upset feeling (Emotionality) 8.19

PAGE 118

111 Table 25 Modification Indices for Error Covarian ces for the Study Anxiety Inventory for Responses from Female Business Students (n = 189) Items with errors covarying Chi-Square Difference Pair 1: 4. Even when I have plenty of time, I feel nervous when I tr y to study for an exam (Emotionality) 6. While studying for tests, other thoughts interfere with my learning (Worry) 32.83 Pair 2: 3. I can’t keep my mind on the subj ect when studying for an exam because other thoughts interfere (Worry) 6. While studying for tests, other though ts interfere with my learning (Worry) 31.13 Pair 3: 1. I feel very uneasy just before starting to study for an exam (Emotionality) 6. While studying for tests, other though ts interfere with my learning (Worry) 21.81 Pair 4: 4. Even when I have plenty of time, I feel nervous when I tr y to study for an exam (Emotionality) 3. I can’t keep my mind on the subject wh en studying for an exam because other thoughts interfere (Worry) 20.45 Pair 5: 5. While studying for exams, I have an uneasy, upset feeling (Emotionality) 7. While studying for a test, I worry about not being able to learn the material (Worry) 20.43 Pair 6: 4. Even when I have plenty of time, I feel nervous when I tr y to study for an exam (Emotionality) 1. I feel very uneasy just before starting to study for an exam (Emotionality) 20.36 Pair 7: 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 15. When I am studying for a test, I can’t get my brain to organize the information (Worry) 19.92 Pair 8: 12. While studying for exams, I feel very tense (Emotionality) 6. While studying for tests, other though ts interfere with my learning (Worry) 18.21 10 with 9 5 with 4 12 with 11 12 with 4 14 with 13 9 with 3 9 with 8 17.67 15.71 15.50 13.52 13.13 13.01 12.87 9 with 6 16 with 12 15 with 3 11 with 7 3 with 2 12 with 2 9 with 5 12.82 12.12 10.48 9.95 9.80 9.54 9.04 10 with 3 16 with 11 5 with 1 16 with 7 15 with 1 10 with 5 9 with 4 8.24 7.93 7.44 7.39 7.36 7.28 7.05 After looking at the CFA model separa tely for males and females, multigroup CFA was conducted to compare the paramete r estimates (factor loadings, residual variances, covariance between factors, and factor variances) for males and females.

PAGE 119

112 Models were tested sequentially beginning with the least restrictive model and continuing with the addition of specific constraints. Table 26 contains the fit indice s corresponding to each of the models that were tested in th e confirmatory factor analysis to evaluate factorial invariance of the scores on the Study Anxiety Inventory by gender. Model 1 is the baseline model in which there are no equali ty constraints across the male and female groups. For this model, factor loadings, resi dual variances (i.e., uni quenesses), and factor variances and covariance are freely estimated in each group (males and females). Model 2 is a more restrictive model that imposes equality constraints on the loadings by gender. Because two factors were hypothesized, one load ing from each factor was fixed to 1.0 to identify the model. This left 14 pairs (seven from each factor) of loadings free to vary. This establishes an increase in the degrees of freedom for Model 2 by 14 which increases the change in the critical value of chi square by 29.121 ( p = .01). Model 3 adds additional restrictions by imposing equality constrai nts on the residual variances for males and females. Model 4 adds an additional equality constraint, restricting the factor covariance to be equal across gender, and Model 5 im poses equality constr aints on the factor variances across males and females. Table 26 shows the different models tested to determine invariance. The constraint of equal loadings by gender is th e more restrictive Model 2 compared with Model 1 (loadings freely estimated in each grou p). The change in the chi square value of 7.51 relative to the change in degrees of freedom ( df = 14) suggested th at invariance of the factor loadings was tenable.

PAGE 120

113 Table 26 Goodness-of-Fit Indices for Models Tested for Invariance of Scores on the Study Anxiety Inventory by Gender (n = 399 Business Students, nM=210, nF=189) Model Model # 2 df 2 df p 1. Baseline 1 707.97206 2. Equal Loadings 2 715.48220 7.51 14>.01 3. Equal Residual Variances 3 761.74236 46.26 16<.01 3a. Equal Residual Variance for all but 4 & 9 3a 736.27234 20.79 13>.01 4. Equal Factor Covariances 4 741.41235 5.14 1>.01 5. Equal Factor Variances 5 741.85237 0.44 2>.01 The next model that was tested (Model 3) imposed equality constraints on the error variances. The 2 was 42.16 relative to a change in degrees of freedom of 16 indicating that inva riance of errors was not tenable.

PAGE 121

114 Table 27 Goodness-of-Fit Indices for In variance of Error Variances on the Study Anxiety Inventory by Gender (n = 399 Business Students, nM=210, nF=189) Item # 22p 1e 717.72 2.24 .1345 2w 715.72 0.24 .6242 3w 715.48 0.00 .9999 4e 730.7 15.22 .0001 5e 715.7 0.22 .6390 6w 719.48 4.00 .0455 7w 716.61 1.13 .2878 8e 717.16 1.68 .1949 9e 724.59 9.11 .0025 10w 718.89 3.41 .0648 11w 715.54 0.06 .8065 12e 718.67 3.19 .0741 13e 715.54 0.06 .8065 14w 718.19 2.71 .0997 15w 715.86 0.38 .5376 16e 715.92 0.44 .5071 Note : degrees of freedom (df) = 235; the change in df = 1; p -value shows 4 decimal places to compare with .05/16 = .0031; 2 for the baseline model was 761.74, df =206. Emotionality (e) items are 1, 4, 5, 8, 9, 12, 13, 16; Worry (w) items are 2, 3, 6, 7, 10, 11, 14, 15 Because the overall hypothesis of eq ual error variances was rejected ( p < .01), follow-up testing of each item error variance was done to identify the source of the difference. A .0031 (.05/16 = .0031) level of stat istical significance was used to control the type I error rate. Results are displayed in Table 27. These results indicated that item error variance was statistically significantl y different across gender for items 4 and 9, both emotionality items. Model 3a removes the restrictions on the two items that demonstrated an inequality of residual varian ce. Models 4 and 5 were then run while still disallowing the restrictions on the three residual variances.

PAGE 122

115 When the covariance between the emotiona lity and worry factors was set equal, the resulting chi square was 741.41, which re presented a change of 5.14 from Model 3a. The change in chi square was not statistically significant at the .01 level indicating that invariance of the covariance between the tw o factors was tenable. When the factor variances of emotionality and worry were se t equal across the male and female groups, the resulting chi square was 741.85 repres enting a change of 0.44, which was not statistically significant at the .01 level indicating that invari ance of these parameters was tenable. College of Education. Table 28 provides descriptive statistics by gender for the 16 observed variables used in the confirmatory factor analysis. The responses for each item ranged from one to four and the means for the 16 items for males ranged from 1.63 to 2.12 (median = 1.82) with standard deviati ons ranging from 0.68 to 1.06 (median = 0.83), and for females ranged from 1.81 to 2.41 (med ian = 2.15) with standard deviations ranging from 0.84 to 1.18 (median = 0.94). Kurtosis and skewness values for each of the items were close to zero for both males and fe male, which suggest that the peak and tails of the distribution were similar to the normal curve and reflected acceptable symmetry. The mean of every item score for the ma les was statistically significantly lower than the mean item score for the females. Effect sizes for the individual items are displayed in Table 28 and ranged from fairly low at -0.15 to moderate at -0.34 with a low median effect size of -0.24. The effect size of the overall scale was low at -0.29 and the effect sizes of the subscales were low at -0.24 for emotionality and low at 0.31 for worry.

PAGE 123

116 Table 28 Descriptive Statistics for the 16 Obser ved Variables Used in the Two-Factor Confirmatory Factor Analyses of the St udy Anxiety Inventory for the College of Education by Gender (nM=210, nF=203) Males Females Males Females Males Females Mea n SD Mean SD t es Skewness Kurtosis 1e 1.79 0.81 2.08 0.95 -3.33** -0.23 0.72 0.61 0.03 -0.50 2w 1.87 0.68 2.24 0.86 -4.84** -0.34 0.35 0.47 0.47 -0.29 3w 2.12 0.83 2.41 0.89 -3.42** -0.24 0.18 0.19 -0.26 -0.68 4e 1.80 0.85 2.15 1.01 -3.80** -0.27 0.72 0.43 -0.14 -0.91 5e 1.67 0.81 1.96 0.95 -3.33** -0.23 1.00 0.75 0.51 -0.35 6w 2.11 0.83 2.37 0.94 -2.98** -0.21 0.38 0.19 -0.15 -0.84 7w 1.92 0.82 2.29 0.92 -4.31** -0.30 0.51 0.32 -0.19 -0.69 8e 1.96 1.06 2.21 1.18 -2.26* -0.16 0.70 0.45 -0.74 -1.31 9e 1.80 0.90 2.15 1.00 -3.73** -0.26 0.78 0.50 -0.30 -0.81 10w 1.84 0.79 2.23 0.94 -4.56** -0.32 0.52 0.43 -0.28 -0.65 11w 1.87 0.87 2.15 0.99 -3.05** -0.21 0.65 0.44 -0.26 -0.84 12e 1.71 0.82 2.03 0.95 -3.66** -0.25 0.90 0.60 0.27 -0.59 13e 1.63 0.80 1.81 0.92 -2.12* -0.15 0.94 0.81 0.01 -0.26 14w 1.83 0.86 2.13 0.90 -3.46** -0.24 0.71 0.39 -0.17 -0.62 15w 1.78 0.76 2.02 0.84 -3.04** -0.21 0.53 0.42 0.06 -0.49 16e 1.81 0.87 1.97 0.92 -1.81* -0.13 0.89 0.59 0.27 -0.58 SAI 1.86 0.61 2.14 0.77 -4.09** -0.29 0.59 0.56 -0.27 -0.49 SAe 1.79 0.71 2.05 0.84 -3.39** -0.24 0.74 0.59 -0.24 -0.53 SAw 1.93 0.59 2.23 0.76 -4.47** -0.31 0.44 0.46 -0.50 -0.42 Note : p <.05, ** p < .01. es = effect size, Emotionality (e) items are 1, 4, 5, 8, 9, 12, 13, 16; Worry (w) items are 2, 3, 6, 7, 10, 11, 14, 15 The 2 fit statistics for the CFA of the two-factor (worry, emotionality) model for male and female students from the College of Educa tion indicated that the fit of the model was not acceptable, 2 (103, NM=208) = 283.07 and 2 (103, NF=202) = 388.97 respectively. The same alternative measur es of fit were included because the 2 is sensitive to sample size. The SRMRs for the male and female groups of .053 and .046 respectively, indicated sim ilar and acceptable fit, while the CFIs of .916 and .905 respectively, and the RMSEAs of .092 a nd .117 respectively, indicated less than acceptable fit.

PAGE 124

117 The unstandardized factor loadings for th e responses, excluding the one fixed to 1.0, ranged from 1.08 to 1.32 with a mean of 1.21 ( SD = 0.08) for the males and from 1.03 to 1.23 with a mean of 1.10 ( SD = 0.09) for the females for the factor of emotionality. These loadings ranged fr om 1.04 to 1.57 with a mean of 1.33 ( SD = 0.21) for the males and from 1.03 to 1.23 with a mean of 1.12 ( SD = 0.07) for the females for the factor of worry. Standardized factor loadi ngs of all items on the emo tionality and worry scales were statistically significant as hypothesized (>.54 for males and >.77 for females). The correlation between emotionality and worry was .91 for males and .92 for females. Sources of misfit of the models for males and females were further explored by comparing the modification indices that indicat ed the expected decrease in model fit chi square ( 2) that would result when a specific para meter, constrained initially to zero, subsequently was freely estimated. As with th e combined sample of education students, the largest expected chi square change was be tween covariances of the errors for items 3 and 6 ( 2=35.08 for males and 63.59 for females). Of the eight pairs of items with the highest chi-square change for the females, th ree of the covariances between errors were also found in the list of the highest eight for males (6 and 3, 14 and 13, and 15 and 14).

PAGE 125

118 Table 29 Modification Indices for Error Covarian ces for the Study Anxiety Inventory for Responses from Male Education Students (n = 208) Items with errors covarying Chi-Square Difference Pair 1: 3. I can’t keep my mind on the subj ect when studying for an exam because other thoughts interfere (Worry) 6. While studying for tests, other thoughts interfere with my learning (Worry) 35.08 Pair 2: 10. While studying for exams, I am stressed with thoughts like “I can’t absorb the material properly”(Worry) 3. I can’t keep my mind on the subject wh en studying for an exam because other thoughts interfere(Worry) 14.21 Pair 3: 13. I freeze up while studying for an important test (Emotionality) 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 13.90 Pair 4: 9. I feel panicky when studying for an important exam (Emotionality) 5. While studying for exams, I have an uneasy, upset feeling (Emotionality) 11.98 Pair 5: 9. I feel panicky when studying for an important exam (Emotionality) 7. While studying for a test, I worry about not being able to learn the material (Worry) 11.92 Pair 6: 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 5. While studying for exams, I have an uneasy, upset feeling (Emotionality) 11.26 Pair 7: 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 3. I can’t keep my mind on the subject wh en studying for an exam because other thoughts interfere (Worry) 10.76 Pair 8: 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 15. When I am studying for a test, I can’t get my brain to organize the information (Worry) 9.94 12 with 7 7 with 5 10 with 9 9.17 8.82 8.49 10 with 7 7 with 3 4 with 3 8.26 7.96 7.26 12 with 10 7 with 1 9 with 3 11 with 9 7.02 6.95 6.82 6.71

PAGE 126

119 Table 30 Modification Indices for Error Covarian ces for the Study Anxiety Inventory for Responses from Female E ducation Students (n = 202) Items with errors covarying Chi-Square Difference Pair 1: 3. I can’t keep my mind on the subj ect when studying for an exam because other thoughts interfere (Worry) 6. While studying for tests, other thoughts interfere with my learning (Worry) 63.59 Pair 3: 9. I feel panicky when studying for an important exam (Emotionality) 8. I wish studying for tests did not upset me so much (Emotionality) 23.32 Pair 2: 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 15. When I am studying for a test, I can’t get my brain to organize the information (Worry) 20.75 Pair 2: 10. While studying for exams, I am stressed with thoughts like “I can’t absorb the material properly”(Worry) 7. While studying for a test, I worry about not being able to learn the material (Worry) 17.78 Pair 2: 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 7. While studying for a test, I worry about not being able to learn the material (Worry) 16.70 Pair 5: 13. I freeze up while studying for an important test (Emotionality) 7. While studying for a test, I worry about not being able to learn the material (Worry) 16.02 Pair 7: 13. I freeze up while studying for an important test (Emotionality) 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material (Worry) 15.70 Pair 8: 4. Even when I have plenty of time, I feel nervous when I tr y to study for an exam (Emotionality) 2. While I am studying for an exam I often think “I’m not getting this” (Worry) 14.72 10 with 6 9 with 6 11 with 2 15 with 13 14.37 13.41 12.88 11.83 7 with 5 10 with 3 2 with 1 11 with 9 11.15 11.13 11.12 9.95 5 with 4 13 with 4 15 with 7 5 with 3 9.61 9.42 8.67 7.06 After looking at the CFA model separa tely for males and females, multigroup CFA was conducted to compare the paramete r estimates (factor loadings, residual variances, covariance between factors, and factor variances) for males and females. Models were tested sequentially beginning with the least restrictive model and continuing

PAGE 127

120 with the addition of specific constraints. Table 31 contains the fit indice s corresponding to each of the models that were tested in th e confirmatory factor analysis to evaluate factorial invariance of th e scores on the Study Anxi ety Inventory by gender. The constraint of equal loadings by ge nder is the more restrictive Model 2 compared with Model 1 (loadings freely estimated in each group). The change in the chi square value of 24.68 relative to the change in degrees of freedom ( df = 14) suggested that invariance of the fact or loadings was tenable.

PAGE 128

121 Table 31 Goodness-of-Fit Indices for Models Tested for Invariance of Scores on the Study Anxiety Inventory by Gender (n = 410 Education Students, nM=208, nF=202) Model Model # 2 df 2 df p 1. Baseline 1 672.04206 2. Equal Loadings 2 696.72220 24.68 14>.01 3. Equal Residual Variances 3 735.59236 38.87 16<.01 3a. Equal Residual Variance for all except 3 3a 726.05235 29.33 15>.01 4. Equal Factor Covariances 4 735.12236 9.07 1<.01 5. Equal Factor Variances with Covariances and Residual Variance for item 3 free to vary 5 737.24237 2.12 2>.01 The next model that was tested (Model 3) imposed equality constraints on the error variances. The 2 was 38.87 relative to a change in degrees of freedom of 16 indicating that invarian ce of errors was not tenable. Because the overall hypothesis of equal error variances was rejected ( p < .01) follow-up testing of each item error variance was done to identify the source of the difference. A .0031 (.05/16 = .0031) level of statistical significance was used to control the type I error ra te. Results are displayed in Table 32. These results indicated that item error variance was stat istically significantly different across gender for item 3, a worry item. Model 3a removed the equality restrictions on this one item that demonstrated an inequality of residual variance. Model 4 was then run while still disallowing the restriction on the residual variance of item 3. The 2 was 9.07 relative to a change in degrees of freedom of 1 indicating that invari ance of covariances was not tenable. Model

PAGE 129

122 5 was then run while still disallowing the restriction on the residual variance of item 3 and disallowing the restricti on on the covariance between wo rry and emotionality. The 2 was 2.12 relative to a change in degrees of free dom of 2 indicating that invariance of the factor variances was tenable. Table 32 Goodness-of-Fit Indices for In variance of Error Variances on the Study Anxiety Inventory by Gender (n = 410 Education Students, nM=208, nF=202) Item # 22p 1e 696.73 0.01 .9203 2w 697.63 0.91 .3401 3w 706.27 9.55 .0020 4e 702.67 5.95 .0147 5e 698.91 2.19 .1389 6w 700.97 4.25 .0393 7w 698.56 1.84 .1750 8e 698.8 2.08 .1492 9e 697.25 0.53 .4666 10w 698.07 1.35 .2453 11w 697.86 1.14 .2857 12e 702.03 5.31 .0212 13e 698.57 1.85 .1738 14w 696.89 0.17 .6801 15w 696.99 0.27 .6033 16e 697.4 0.68 .4096 Note : degrees of freedom (df) = 221; change in df = 1; p -value shows 4 decimal places to compare with .05/16=.0031; 2 for the Equal Loadings model was 696.72, df =2 20. Emotionality (e) items are 1, 4, 5, 8, 9, 12, 13, 16; Worry (w) items are 2, 3, 6, 7, 10, 11, 14, 15 Summary of Results for Research Question 2 Table 33 presents fit indices for the confirma tory factor analysis for the two-factor model for both males and females for each college. Although for each college, chi square values for males and females indicated less than acceptable fit, the SRMRs indicated acceptable fit for both in each college with SRMRs ranging from .050 to 053 for males

PAGE 130

123 and from .046 to .060 for females. Both males and females for all four colleges had CFIs that indicated less than acceptable fit for the two-factor model (CFIs for males ranged from .907 to .921 and for females from .869 to .921). Both males and females for each college had RMSEAs that indicated less than acceptable fit for th e two-factor model ranging from .089 to .103 for males and from .089 to .112 for females. Table 33 Fit Indices for the Confirmatory Factor Anal ysis of the Hypothesized Two-Factor Model by Gender for the Study Anxiety Inventory Across Four Colleges Arts and Sciences Business Education Engineering Males Females Males Females Males Females Males Females 2 337.64 617.76 273.61 434.36283.07 388.97 359.11 345.52SRMR .050 .053 .052 .059 .053 .046 .053 .060CFI .915 .904 .921 .869 .916 .905 .907 .882RMSEA .103 .106 .089 .089 .092 .117 .102 .112 Standardized loadings of the items for the emotionality and worry factors were consistent across the four colleges and were for both males and females above .60 on emotionality and above .62 on worry. The standardized loadings were above .54 for males and females on both factors for each college. Correlated errors for item pa irs 6 and 3 and 15 and 14 were significant sources of misfit indices for all four colleges and for both males and females. The pair 14 and 13 also had a significant modification index for all four colleges and for each college sample of females (exception was for the sample of males in business). Item pairs 10 and 9, 9 and 6 and 9 and 3 also had significant modifica tion indices for all four colleges but when checked for males and females within the colleges produced inconsistent results.

PAGE 131

124 Correlated errors for pairs 4 and 3, 4 and 2 and 2 and 1 were also significant for all four colleges. Although there were a number of othe r pairs of items that showed modification indices that were significant, none was significant for all four colleges. Table 34 presents a list of the pairs that showed significant s ources of misfit for thr ee or four colleges.

PAGE 132

125 Table 34 Item Pairs that Showed Signi ficant Chi Squares for Modification Indices on the Study Anxiety Inventory by Gender Across Four Colleges Item Pair Colleges Males Females 15 with 14 4 4 4 14 with 13 4 3 4 10 with 9 4 2 2 9 with 6 4 1 2 9 with 3 4 2 2 6 with 3 4 4 4 4 with 3 4 3 1 4 with 2 4 1 2 2 with 1 4 3 1 15 with 1 3 1 1 14 with 7 3 3 1 12 with 2 3 1 1 10 with 3 3 2 2 7 with 5 3 1 2 6 with 4 3 1 1 5 with 4 3 0 4 5 with 1 3 1 1 Note: Numbers in table represent the number of times out of four that the correlated error for the pair of items was a significant source of misfit. Invariance testing by gender indicated that loadings were equal for each of the four colleges; however some differences in item residual variances identified in the four colleges (see Table 35). Factor covariances were invariant across all four colleges and factor variances were equal for all except the College of Arts and Sciences where the factor of worry was shown to vary by gender with females having more varied responses.

PAGE 133

126 Table 35 Summary of Results for Models for the Study Anxiety Inventory Tested for Invariance by Gender by Colleges Arts and Sciences Engineering Business Education Equal Loadings >.01 >.01 >.01 >.01 Equal Residual Variances <.01 <.01 <.01 <.01 Equal Residual Variance with corrections >.01 when 4,7,8 &16 were freed >.01 when 9 & 13 were freed >.01 when 4 & 9 were freed >.01 when 3 was freed Equal Factor Covariances >.01 >.01 >.01 <.01 Equal Factor Variances >.01 for emotionality but <.01 for worry >.01 >.01 >.01 Research Question 3: Relationships between SAI and Related Measures Because the Study Anxiety Inventory (SA I) and the two components of worry and emotionality are based on items from gold standard anxiety measures, it was predicted that scores from the SAI (overall, worry a nd emotionality) would have a strong positive correlation with scores from the Test Anxiet y Inventory (overall, worry and emotionality) and scores from the Trait Anxiety scale. Ba sed on the Optimal Stimulation Dual Process Theory, a moderate to low negative relati onship was predicted be tween the SAI, Study Anxiety/Emotionality (SA/e) and Study Anxi ety/Worry (SA/w) with scores from the Trait Curiosity scale. Based on previous fi ndings, it was predicted that the SAI and the subscales (SA/e and SA/w) would show a weak relationship with the measure of study skills and habits. As a part of the construc t validation process, the nomological network was extended by examining the relationship be tween study anxiety and passive and active procrastination. Because the procrastination measures and the study skills and habits

PAGE 134

127 measure are fairly new, prior to presenting the correlational results as part of the construct validation process, a table of Cronbach ’s alphas for all measures is presented. Reliability of Scores Internal consistency reliability for the scales was computed for each college (see Table 36). Cronbach alphas for the Study A nxiety Inventory and the Test Anxiety Inventory for the four colle ges ranged from .91 to .96. Cronbach alphas for Spielberger’s trait anxiety and trait curiosity measures for the four co lleges ranged from .77 to .85. For the study skills and habits measure (Study fo r Examinations), Cronbach’s alphas were .78 for each college except for Arts and Sciences which was .79. For the passive procrastination scale, Cronbach alphas for the four colleg es ranged from .81 to .85, and for the active procrastination scale from .64 to .66. All of the scales and subscales demonstrate that responses to the items of each scale are highly related. The lowest internal consistency index at .66 for the Activ e Procrastination scale was not so low that the scale was not useful for the purposes of this study. It was c oncluded that scores obtained from these measures were sufficien tly reliable to be used in the construct validation process.

PAGE 135

128 Table 36 Cronbach’s Alpha for Internal Consistency for C onstructs of Interest for Four Colleges Constructs of Interest # items Arts & Sciences Engineering Business Education Study Anxiety Overall Scale 16 .95 .95 .95 .96 Study Anxiety/Worry 8 .92 .91 .91 .92 Study Anxiety/Emotionality 8 .94 .93 .93 .94 Test Anxiety Overall Scale 16 .96 .95 .96 .96 Test Anxiety/Worry 8 .93 .91 .92 .92 Test Anxiety/Emotionality 8 .94 .92 .93 .93 Trait Anxiety 10 .85 .83 .84 .85 Trait Curiosity 10 .80 .77 .78 .82 Study for Examinations 25 .79 .78 .78 .78 Passive Procrastination Scale 6 .85 .81 .82 .82 Active Procrastination Scale 12 .66 .64 .66 .66 Correlational Results for SAI and Constructs of Interest Findings for the correlations between the Study Anxiety Inventory (SAI) scale and the emotionality (SA/e) and worry (S A/w) subscale scores and other anxiety measures are presented in Table 37. These ot her anxiety measures include test anxiety (TAI) with subscale scores reflecting emoti onality (TA/e) and worry (TA/w), and trait anxiety (T-Anx). Table 38 contains the corr elations between the SAI (SA/e and SA/w) with two measures reflecting constructs that are commonly related to academic performance, curiosity (T-Cy) and study skills and habits (SH). Finally, Table 39 provides two measures of differe nt types of procrastination, active (AP) and passive (PP). Anxiety Measures. Although there is no overlap between the content of the emotionality and worry subscales of the SAI and the Test Anxiety Inventory (TAI), because the items of the SAI are based largely on the TAI, the correlations between the

PAGE 136

129 two measures were expected to be large and statistically significant. As expected, there was a strong correlation between the SAI scale and subscale scores with the Test Anxiety Inventory (TAI) scale and subscale scores for a ll four colleges. These correlations ranged from .64 to .83 (median r =.74). As hypothesized, the correlations were m oderate to high for the SAI and scores from the trait anxiety (T-ANX) scale. Correlatio ns for the four colleges for these scores ranged from .40 to .48 (median r =.46).

PAGE 137

130 Table 37 Pearson Product Moment Correlations Between the SAI, SA/e and SA/w with Anxiety Constructs of Interest for Students from Four Colleges SAI SA/E SA/W A & S ( n = 662) TAI .78** .79** .74** TA/E .76** .78** .72** TA/W .76** .77** .73** T-Anx .42** .42** .41** Engineering ( n = 434) TAI .79** .80** .76** TAI/W .78** .78** .76** TAI/E .77** .79** .73** T-Anx .48** .48** .47** Business ( n = 434) TAI .80** .81** .77** TAI/W .77** .77** .74** TAI/E .80** .81** .76** T-Anx .45** .45** .44** Education ( n = 409) TAI .79** .80** .76** TAI/W .78** .78** .75** TAI/E .77** .79** .74** T-Anx .61** .63** .58** SAI=Study Anxiety Overall Scale SA/W=Study Anxiety/Worry SA/E =Study Anxiety/Emotionality TAI=Test Anxiety Overall Scale TA/W=Test Anxi ety/Worry TA/E=Test Anxiety/Emotionality T-Anx=Trait Anxiety *<.05, **<.01 Curiosity and Study Skills and Habits. As hypothesized, the correlations were negative between the SAI, and both the SA/e a nd SA/w subscales, with trait curiosity (TCY). The strength of the relationships was m oderate for Arts and Sciences, Engineering and Business students ( r s ranged from -.13 to -.22) and strong for Education students ( rs =-.39, -.41, and -.42, respectively). For these st udents, higher scores on curiosity were related to lower scores on anxiety while studying. There was no statistically significant re lationship between the study skills and habits (SH) scale and the SA I or the SA/e and SA/w subs cales except in the responses

PAGE 138

131 from the students from the College of Business who showed a small relationship between the emotionality and worry subscales with SH (rSA/e to SH=.-.12 and rSA/w to SH=.-10). Table 38 Pearson Product Moment Correlation for Other Constructs of Intere st for Students from Four Colleges SAI SA/E SA/W A & S ( n = 662) T-CY -.14** -.12** -.15** SH .04 .07 .02 Engineering ( n =434) T-CY -.22** -.23** -.21** SH -.07 -.05 -.09 Business ( n = 409) T-CY -.16** -.15** -.16** SH -.09 -.06 -.12* Education ( n = 413) T-CY -.38** -.38** -.38** SH .01 -.02 .03 SAI=Study Anxiety Overall Scale SA/W=Study Anxiety/Worry SA/E=S tudy Anxiety/Emotionality T-CY = Trait Curiosity SH=Study for Examin ations *<.05, **<.01 Procrastination. A positive correlation was predic ted between scores from the SAI, SA/e and SA/w with scores from the m easures of active and pa ssive procrastination. The correlations of the SAI, SA/e and SA/w scales with each procrastination scale for each college are reported in Table 37. The relationships between the SAI total score and Active Procrastination Scale scores were pos itive and moderate a nd ranged between .22 and .30 (median = .23). For the Passive Procras tination Scale scores, correlations were positive and moderate and ranged from .26 to .32 (median = .28). Except for the College of Engineering, the picture is different for the subscale scores of Emotionality and Worry with correlations ranging between .21 and .29 for active procrastination (median = .24) and .23 and .36 for passive procrastination (median = .26). Correlations for passive

PAGE 139

132 procrastination and SAI (Worry and Emoti onality) were generally higher than the correlations for active procrastination and SAI (Worry and Emotionality). It is, perhaps, worth noting, in case the reader should believ e that these two procrastination measures are merely the same, that Chu and Choi (2005) posited that the constructs of active and passive procrastination as measured with th ese two scales were not related. This was supported by the findings in this study in which the correlations between active and passive procrastination ra nged from -.15 to -.01. Table 39 Pearson Product Moment Correlations Between Each Study Anxiety Variable and Each Procrastination Scale for Arts and Scie nces, Engineering, Business and Education Students SAI SA/E SA/W Arts & Sciences ( n = 500) APS .23** .25** .21** PPS .29** .25** .32** Engineering ( n = 386) APS .30** .29** .30** PPS .26** .23** .27** Business ( n = 362) APS .23** .21** .24** PPS .32** .30** .32** Education ( n = 212) APS .22** .23** .21** PPS .28** .27** .28** Note : SAI = Study Anxiety Inventory SA/E = SA Emotionality SA/W = SA Worry APS=Active Procrastinatio n Scale PPS=Passive Procrastination Scale p <.05 ** p <.01

PAGE 140

133 Summary of Results for Research Question 2 As was expected, for all four colleges, the relationships between the SAI and its two factors with the Test A nxiety Inventory and its two f actors were high while the relationships between the SAI a nd its two factors with trait an xiety were lower. For each of the colleges, the relationships between SAI and its two factors with trait curiosity were moderate and negative. For a ll four colleges, the relationships between the SAI and its two factors with the measure of study skill were, with two exceptions, not significantly different from zero. For three of the college s (exception was Engin eering), correlations for passive procrastination a nd SAI (Worry and Emotionality) were generally higher than the correlations for active procrastinati on and SAI (Worry and Emotionality).

PAGE 141

134 Chapter 5 Discussion The purpose of this study was to collect va rious types of eviden ce to evaluate the construct validity of the inferences deri ved from the Study Anxiety Inventory (SAI; Lunsford, 2001) from students from four colleges (Arts and Sc iences, Engineering, Business, and Education) at a large southeastern state university. This chapter contains six sections. The first section discusses the construct of study anxiety and the development and validation process used fo r the Study Anxiety Inventory. The second section discusses the results related to the first research question, which focused on the factor structure of the SAI. The third and f ourth sections discuss the results related to research questions two and three (invarian ce of the SAI by gender and relations of the SAI to other variables respect ively). The fifth section pres ents the significance of the study with conclusions concer ning the CFA and the relatio nship discovered between study anxiety and two measures of procrastina tion. The sixth section id entifies limitations of the study and provides recommen dations for future research. Background Study anxiety was conceptualized as a situ ation-specific anxiety with the same worry and emotionality components found in test anxiety. Lunsford (2001) used an expansion of Lazarus’s Transactional Process Theory discussed in Chapter 2 as the basis for the development of items for the Study A nxiety Inventory. Lunsford (2001) provided several types of evidence to support the validity of the infe rences from the SAI including an analysis of item content (c ontent validity), internal stru cture of the responses to the

PAGE 142

135 items (exploratory factor analysis), and re lationships between th e construct and other variables (concurrent, predictive, and construct validity). T he purpose of the current study was to collect further evidence of the validity and reliability of the scores from the SA I. More specifically this study had three purposes: (a) evaluate the tw o-factor model underlying the Study Anxiety Inventory, (b) evaluate the factorial equi valence by gender of the twofactor measurement model underlying the SAI, and (c) examine the constr uct validity of the SAI by examining its relationship to test anxiety, trait anxiety, trait curiosity, study habits and skills, and passive and active procrastination. To address these purposes, data were co llected from 2,002 undergraduates at one southeastern state research university. Participants included 664 students from the College of Arts and Sciences, 456 from the College of Engineering, 431 from the College of Business, and 413 from the College of Education. Paper and pencil measures were handed out to 2,002 undergraduate university students during normal class periods. The measures included the Study Anxiety Inventor y, the Test Anxiety Inventory, the Trait Anxiety scale and the Trait Curiosity scale fr om the State Trait Pers onality Inventory, the Study for Exam (SH) scale from the Study Habits Evaluation and Instruction Kit (SHEIK), the Active Procrastination Scale, and the Passive Proc rastination Scale. Research Question One: Evidence of Two-Factor Structure The Study Anxiety Inventory was hypothe sized to consist of two underlying factors or dimensions: worry and emotionalit y. Confirmatory factor analysis (CFA) was used to evaluate the two-factor model (worry and emotionality). Resu lts of these analyses

PAGE 143

136 indicated that the fit of the two-factor mode l was marginally acceptabl e with most of the measures of fit below the guidelines for acceptable fit proposed by Hu and Bentler (1999). The SRMR was the only measure of fit th at suggested an acceptable level of fit. These results, along with the strong correlation between emo tionality and worry (ranging from .87 to .92 across the four colleges), led to the consideration of an alternative model that consisted of one-factor. Fit of the one -factor model, evalua ted using chi-square, SRMR, CFI, and RMSEA, indicated that the one-factor model for all four colleges was less acceptable than the two-f actor model. These results provide some support for the two-factor model and the underlying theo ry that guided the development of the instrument. Although the fit of the two-f actor model was statistica lly better than the onefactor model, support for the two-factor model was not overwhelming. The finding of strong correlations between worry and emo tionality makes it reasonable to question whether viewing study anxiety as having two factors might be unnecessary and that one overall score would give as much informa tion as two. Correlation coefficients between worry and emotionality ranged from .87 to .92 indicating that from 76% to 85% of the variance in one factor can be explained by the other factor. Although there is some unique variance that is captured by the two f actors, some researchers may decide that there is not enough unique variance and ther efore choose to use an overall score for research purposes. Further inves tigation is needed to determin e if these high correlations replicate in other settings and if the factors of worry and emo tionality differentially relate to student outcomes (e.g., GPA).

PAGE 144

137 Examination of the modification indices (i.e ., indicators of where there is misfit in the model) from the confirmatory factor anal yses showed that there may be a degree of redundancy in the items. Modifi cation indices that were si gnificant and large involved correlated errors between two worry items. Item 3 (“I can’t keep my mind on the subject when studying for an exam because other thoughts interfere”) and item 6 (“While studying for tests, other thoughts interfere with my learning” ) both seem to be focusing on the inability to keep thought s from interfering with learning. Two other worry items, item 14 (“When I study for exams, I seem to get a mental block that keeps me from absorbing the material”) and item 15 (“When I am studying for a test, I can’t get my brain to organize the information”), also showed c onsistently large correlated errors across all four colleges possibly because the phrases “m ental block” and “can’t get my brain to organize” could be viewed as the same by many people. These same pairs of items had modification indices that we re significant and large acro ss all four colleges for both males and females except for the item pair 14 and 13, which was only significant for three of the colleges for males. Because the essential idea of including items on a questionnaire is to learn more about the construct rather than having items that are merely repetitions of the same question, this might i ndicate that certain items could be removed without decreasing the information obtained by the measure. Further research would be needed to establish this as the best course of action. Research Question Two: Eviden ce of Invariance by Gender Research has consistently found that self-re port scores of anxiety for females are higher than for males (Hewitt & Norton, 1993; Sp ielberger, 1975; Spielberger & Wasala,

PAGE 145

138 1995). Females have also reported more conc ern about emotional, physical and mental symptoms related to anxiety. Mean differen ces between male and female respondents, then, have been fairly well established, but the factor structure underlying the measures of anxiety have not been compared to determine whether males and females view the meaning of the items in the SAI in a similar manner. Therefore, a second purpose of the st udy was to determine whether there was factorial invariance of the SAI by gender. Invarian ce testing involved carrying out comparisons of the factor pattern coefficients (loadings), uniquenesse s (error variances), and factor variances and covariances acro ss the male and female groups. Factorial invariance of the SAI for males and females was examined using multigroup confirmatory factor analysis. The key element in invariance testing is in establishing that the same items load on the same factor to the same degree acro ss groups. Factor load ings are similar to regression coefficients. They reflect the st rength of the relations hips between each item and its underlying construct and represent the change in observed scores that occurs for every unit change on the latent construct (Vandenberg, 2002). If these loadings are statistically different between groups, it indi cates that the responde rs in the different groups view the items as having different m eanings. The construct is defined by how the items load and if they do not load on the same factor for males and females then the invariance of the residuals, factor varian ces and the covariance between the factors (worry and emotionality) is irrelevant. As Vandenberg and Lance (2000) stated, if there is a difference between groups in the relations of items to the latent variable, then

PAGE 146

139 comparing scores between those groups “may be tantamount to comparing apples and sparkplugs” (p. 9). Because gender differences are ofte n referred to in reports about different types of anxiety, fi ndings of invariance across gender is an important part of the construct validation process for study anxiety, as findings of invariance may indicate that mean differences found could be spurious. At first look, the factor loadings for th e College of Arts and Sciences did not appear to be invariant acro ss gender, but further analysis showed that no item loading was significantly different across gender. F actor loadings by gender for the Engineering, Education, and Business students’ responses we re not found to be significantly different. These findings indicate that there was no evid ence that males and females in each of the colleges view the meaning of the items in the SAI in a different manner. Any differences in observed mean scores between males and fe males on identical items or scales are not due to measurement bias but, rather, are due to true differences on the factor mean. It is therefore reasonable for a researcher to feel comfortable making mean comparisons between males and females for this measure. Further investigation showed that inva riance of the residu als for the observed variables was not supported. If i nvariance of the factor loadings is established, invariance in residuals can be consider ed a test of the invariance of scale reliability by gender (Schmitt & Kuljanin, 2008; Steenkamp & Baumgart ner, 1998). This is the most stringent of the invariance tests and non-significance is not necessary in order to be able to make meaningful cross-group interpre tations of mean differences. These findings state that the items carry an unequal amount of error which suggests that there is a difference between

PAGE 147

140 the reliability of the scales for males vs. fe males. The lack of invariance for the item residuals is a common finding with psychologi cal measures (Steenkamp & Baumgartner, 1998). The covariance between the factors of worry and emotionality was not significantly different between males and fema les for three of the colleges (Education was the exception), which indicated that the two subscales were related in the same way for males and females. Invari ance of the factor variances by gender was also supported for all except the variance for worry for the College of Arts and Sciences where that factor was shown to differ by gender. Female s used a wider range of responses than the males on the factor of worry which suggests that they have a wi der range of worry cognitions than males. This suggests that, should means be compared between males and females for these students, it would be prudent to precede that test with a comparison of the variances to determine whether or not it would be appropriate to carry out an independent t -test (i.e., one assumption underlying an independent t -test is homogeneity of variance). Research Question Three: Validity Evidence Based on Relations to Other Variables Evidence of the validity of the psychol ogical construct of study anxiety was provided by expanding the framework of the nomological net (Cronba ch & Meehl, 1955). Deeper insight into the construct validity of the scores from the SAI was provided by examining the scores from the instrument with other theoretically meaningful constructs. Construct validation using th is framework (AERA et al., 1999) involved carrying out tests of the relationship between study anxiety and the relate d latent variables of test

PAGE 148

141 anxiety, trait anxiety, trait cu riosity, study skills and habits active procrastination, and passive procrastination. The Test Anxiety Inventory (TAI) is based on Lazarus’s Transaction Process Theory and measures situation-specific a nxiety that occurs during an exam. The Study Anxiety Inventory is also based on an exte nsion of this same theory and measures situation-specific anxiety that occurs wh ile studying for that exam. Based on these similarities, it was predicted that a modera te positive correlation would be found between scores from the SAI and the TAI. Because situation-specific anxiety falls unde r the umbrella of tr ait anxiety, it was also predicted that the correlation between scor es from the SAI with scores from the trait anxiety measure would be moderate but lower than the correlation wi th test anxiety. As in a previous study (Lunsford, 2001), findings in the current study showed a positive relationship between study anxi ety and test anxiety (median r = .76), and, as expected, a weaker relationship between study a nxiety and trait anxiety (median r = .44). As scores on the SAI increase, so do scores on the TAI and the trait anxiety measure. This supports previous findings and provides evidence to support the extended theory upon which the items of the SAI were created. The correlation between test and study anxiety is high enough that one might question whether they are separa te constructs. As presented in the first chapter, the two constructs are similar in that the student is responding to the same perceived threat in similar ways (worry and emotionality), but study anxiety and test anxiety are separated by a number of conceptual issues. The anxious thoughts and feelings occur in different

PAGE 149

142 situations (while studying vs. while taking an exam) and have different effects (hinders pre-attention and attentional processes vs. hinders memory retrieval). The environment for studying is set by the student while the environment of an exam is set by the instructor or proctor. The measures specify these different times, situations and effects and, while the correlations indicate that wh en one experiences one construct, one also experiences the other, there will be those who feel anxiety while studying but calm down when they start the exam or those who feel calm until the exam starts and then feel the anxiety symptoms. Based on the Optimal Stimulation/Dual Pr ocess Theory presented by Spielberger and Starr (1994), curiosity w ould be inhibited by anxiety and this, combined with previous findings that SAI scores correla ted negatively with trait curiosity scores (Lunsford, 2001), prompted the prediction that the same negative correlation would be found in this study. As in the previous study, findings in this study showed a negative relationship between study a nxiety and the construct of trait curiosity (median r = -.19), which supports the validity of the construct and adds evidence to the theory upon which the items were based. Lazarus’s Transaction Process Theory sugge sts that deficits in study skills and habits will influence a student to believe that failure on an exam is imminent which would lead to test anxiety. The expanded theory suggests that deficits in study skills and habits will lead directly to worry and emoti onality while studying. This theory led to the hypothesis that scores on the SAI would correl ate negatively with scores from the study skills and habits measure. Contrary to what was hypothesized, study anxiety showed no

PAGE 150

143 significant relationship with study skills and habits across all four colleges (median r = .04). This suggests that people with or without good study skills and habits will experience symptoms of study anxiety. Th ese results were determined using a correlational design and therefore future re search may examine if an experimental intervention designed to impact students’ know ledge and practice of study techniques would impact their study anxiety. Chu and Choi (2005) have suggested that there are two major types of procrastination: passive (a response to stre ssors) and active (a planned behavior to improve performance). Atkinson (1974) proposed that those who tend to be more anxious about failing will avoid tasks that will bring on that anxiety. McCown and Johnson (1991) stated that anxiety is a motivating fact or in dilatory behavior. This implies that study anxiety would correlate positively with passive proc rastination. Chu and Choi (2005) further suggested that active procrastin ators are less like pa ssive procrastinators than they are like non-procras tinators in terms of anxi ety which suggests that the relationship of scores from the SAI would not be as strongly positive with active procrastination. This study f ound that the measures of passi ve and active pr ocrastination showed a positive relationship (median r = of .28 and .23, respectively) with the scores from the SAI, which indicates that thos e who experience study anxiety may also experience either passive or active procra stination. Those who put off studying because they find the task stressful, or because they believe they work bette r under stress and so put off tasks until the last minute, may also experience some degree of anxiety while studying. Not much variance in the scores of the SAI can be accounted for by the scores

PAGE 151

144 from the passive or active procrastination measure. Because people passively respond to situations considerably more than they actively plan behaviors, a higher positive correlation should appear with passive procra stination than with active procrastination. Although the difference was not statistically significant, the difference was in the direction one would expect fo r the students from the College of Arts and Sciences (rSA to PP=.29, rSA to AP=.23), the College of Business (rSA to PP=.32, rSA to AP=.23), College of Education (rSA to PP=.28, rSA to AP=.22) but in the opposite di rection for the College of Engineering (rSA to PP=.26, rSA to AP=.30). In summary, as was hypothesized, the SAI sc ores were positively correlated with scores on measures of test anxiety, trait anxiety, active procras tination and passive procrastination but negatively correlated with trait curiosity. Contrary to what was hypothesized, no relationship was demonstrated between study anxiety and study skills and habits. The nomological network wa s extended in this study by examining relationships between scores obtained from students on the SAI and measures of active and passive procrastination. It should be kept in mind that the participants completed these measures at the same sitting (common time) and that these measures were all of the same type (paper and pencil) with a sim ilar response format (common method), which could possibly account, in part, for the observed relationships. Significance of the Study This study was designed to examine sy stematically the two-factor model underlying the SAI. Part of this objective was achieved by testing the two-factor model of the SAI by college and then separately by gender. The current study has provided some

PAGE 152

145 support for the factorial validity of the Study Anxiety Inventory, so that, at least for research purposes, this measure can be used to continue investiga ting the construct of study anxiety. The correlations between the tw o factors ranged fr om .85 to .92 within each college and for males and females, which led the researcher to consider an alternative one-factor model. The one-factor model of the SAI provided an inadequate fit to the data, and while the two-factor model is not ideal, it appears th at the SAI is better represented by a two-factor model. Further res earch evaluating the fact or structure of the SAI is warranted. Another part of the objective was ac hieved by addressing potential gender differences in the factorial structure of the SAI. This is the first study that has systematically examined the factorial invari ance of the SAI by gender, which is important because previous research using the SAI has s hown men’s mean scores to be consistently lower than women’s scores. This difference could have been due to noninvariance in SAI items rather than gender differences in leve l of reported study anxi ety. Unless the factor loadings are invariant, it is not mean ingful to make mean comparisons. The results obtained in the current study provide support for gender invariance in a nonclinical population in the situation-specific level of anxiety while studying. The factor structure for both males and female s was not significantly different, providing further evidence that men and women are in terpreting the items in a similar way but endorsing them differently. Females may have el evated anxiety but the relationships that the items have with the construct are similar. Given this invariance it is appropriate to examine mean differences by gender. This applies to a non-clinical population only,

PAGE 153

146 however, as this research wa s carried out on a non-clinical population. Until research is carried out on a clinical population, this m easure should be used for researching the construct of study anxiety and not for individua l diagnosis or clinical purposes such as deciding treatment for those suffering from anxiety. Comparisons of means in the present study indicated that there were statistically significant gender differences in self-repor t of anxiety by males and females while studying, although the effect sizes were moderate to low. These effect sizes are similar to those reported in studies over the last decade that have compared trait anxiety scores for males and females (Everson, Millsap, & Rodriguez, 1991; Foot & Koszycki, 2004; Marcus, 2001). These results are consistent w ith theory relating gender to anxiety and with findings from other research, thus providing support that the SAI measure is performing as expected. The results of the CFA lead to these c onclusions and the correlational analyses: 1. The fit of the two-factor model is margin al but the model would be acceptable to use in research to investig ate further the relationships between each factor and other variables. Further research might also address the issue of items being somewhat redundant. 2. The two-factor structure of study anxi ety was invariant by gender, but gender differences were detected in the mean s indicating that females reported higher levels of anxiety with low to moderate effect sizes. This supports the theory discussed in Chapter 1 which led to the pr ediction that there would be differences in the means but that the factor struct ure would be invariant. This allows

PAGE 154

147 researchers who want to make gender comparisons to be more comfortable that their findings are due to real differe nces and not a measurement artifact. 3. The finding that there is a relations hip between anxiety while studying and different types of procrastination is a new addition to the literature. 4. Overall, there is sufficient evidence of va lidity and reliability that a researcher should feel confident that the SAI is a reas onable research tool that holds up fairly well across a number of diffe rent types of students. Limitations Although the sample size of undergraduate students was approximately equal by gender and the sample was heterogeneous from four different colle ges with many majors, one limitation of this study was that the students were not sel ected randomly. Instead, convenience sampling was used to recruit th e participants from one southeastern university and hence the sample may not represent students from other types of universities (e.g., private, on-line, “Ivy League,” etc.). A second limitation of this study was that it measured students at one time only with the students being mostly around the age of 21. Also, the sampling was cluster samp ling (i.e., classes) so there may be some violation of the independence of the data wh ich can lead to infl ated Type I error. A third limitation of this study was that al l data were collected utilizing a surveytype methodology. The advantage of a self-report measure of anxiety is that it enables the efficient assessment of the frequency of beha viors, thoughts and feelings across time of a large number of participants. Disadvantages of a self-report measure include: (a) inability of the items to encompass the entire range of anxious symptoms of the responders, (b) the

PAGE 155

148 avoidance or denial of the anxiety expe rienced by the respo nders, (c) responders’ difficulty in revealing weakness (social desirabi lity) or secret feeli ngs of anxiety (selfserving bias), (d) misinterpretation of items by those with low reading ability or low comprehension, (e) non-compliance due to lack of interest or retalia tion, (f) forced-choice categories may not fit the experi ence of the responder, (g) res ponse bias due to inaccurate recall of experience (Sallis & Owen, 1999), and (h ) the responders’ lack of awareness of their thoughts, feelings, and behaviors. It ma y also be that the observed similarities between the measures are due to the similarity of the items and response constraints rather than the perceptions or constructs themselves. An attempt was made to address some of the disadvantages connected with selfreport measures. To make the reading level of the measure sufficiently low to cover even students whose first language was not Englis h, wording on the survey was established at a sixth grade level using the Flesch-Kincaid Gr ade Level test. In order to avoid social desirability, emphasis was placed on the fact that no names were recorded on the measure thus providing complete anonymity. To address self-serving bias, it was pointed out that the information obtained from the measures woul d be reported as group data; to deal with non-compliance due to lack of interest or retaliation, it was announced that that they could opt out of filling in the items at any time. As previously mentioned, re searchers have suggested that test anxiety in the form of emotionality and worry is a stable phe nomenon (Spielberger, 1980). Because of the correlation found between study a nd test anxiety and in view of the theory upon which the measure is based, this statement could be extended to suggest th at study anxiety is

PAGE 156

149 also stable in this way. Because the rela tionship between rumination and mood in both cross-sectional and longitudinal designs ha ve been examined using survey studies (Brinker & Dozois, 2009), it is r easonable to suggest similar st udies for future research on study anxiety. A fourth limitation of this study was that none of these students was screened as needing help due to situation-specific anxiet y (e.g., test anxiety) even though these are the types of students that this measure may eventually be used to assess. Recommendations for Future Research With regard to the first limitation, future studies should expand the sample to include students from differe nt parts of the country and of different ages including graduate students and pre-college students. This would help to determine how well the SAI works with different types of students. Concerning the second limitation, future studies should include both younger and older age groups and investigate whether study anxiety changes with age and influences learning for both younger and older students. Also, further research is needed to determine whether differences in study anxi ety between males and females change with age. If it is demonstrated that study anxiety is stable over specific situations and time, this could show that this t ype of anxiety may be a contributing factor in schooland work-related learning problems. Future rese arch, then, could investigate study anxiety over time by following participants over a peri od of years. In addition to using the SAI to examine anxiety over time, daily diary logs ove r that same time would extend the validity

PAGE 157

150 of the construct of study anxi ety and the inferences made based on the responses to the SAI. The third limitation could be addressed by using the diary entry approach. The method used by the Study Anxiety Inventory of asking individuals, in a retrospective way, whether they experience anxiety while studying, is only one way to obtain this information. An alternative approach to meas uring study anxiety that could be used in a multitrait-multimethod design to provide construct validity evidence of the SAI would involve using a diary method dur ing a period in which student s had important exams. For example, if students were asked to record on a provided paper or elect ric diary what they felt at the time they approached the time of studying, or were studying for exams, this information could then be examined and compar ed with responses to the scores from the Study Anxiety Inventory. Requi ring an individual to selfobserve and systematically record his or her anxiety at the time it o ccurs would be an effective way to collect evidence concerning the frequency of study anxiety as well as its consequences (Shiffman, Stone, & Hufford, 2008). With a di ary approach to examine study anxiety, other disadvantages of self-report measur es are addressed. Th e range of anxious symptoms experienced by the responder coul d be reported as the technique would not restrict the person to a preset list of symp toms. This approach would also have the responses at the time of studying so an exam ination of current feelings would help the responder from denying his or her experience of anxiety. This technique of data collection would not be so reliant on me mory, would not rely on the participant understanding the language of the items, and, with appropriate cues, would cause the

PAGE 158

151 responder to increase his or her awareness of thoughts, feelings and behaviors. If the reliability and validity of the scores obtain ed from the diary method were demonstrated, this method in conjunction with the SAI could increase our u nderstanding of study anxiety. Diaries have been demonstrated to be an effective assessment tool with externalizing behavior disorders (N elson, Hay, Devany, & Koslow-Green, 1980), although it is not an approach frequently us ed to assess anxiety. This addition to the literature would establish a di fferent method to approach th e validity of the inferences from the scores from the study anxiety measure. The fourth limitation could be addressed in future studies by including students who had applied to the university counseli ng center for help with problems related to anxiety. Although important similarities on responses to the SA I may exist between students who have and who have not been asse ssed for clinical leve ls of anxiety, there have not been any studies that have exam ined invariance across gender among those who experience anxiety at this hi gher level, and so conclusions concerning these types of students are premature. It is im portant therefore to replicate th e research in other settings with non-clinical and clinical samples. Future studies need to be carried out usi ng methods that are similar to those used here to assess gender invariance. A possible starting point would be to use a group being treated for test anxiety in uni versity counseling centers to a ssess the fit of the two-factor model for males and females and the invari ance across gender in a clinical sample. A study of this kind would add to the construct validity of this instrument. Assuming results were similar among groups that were obtaining treatment, a pot ential use of this measure

PAGE 159

152 would be for screening and treatment evalua tion of those who suffer from anxiety while studying. Also, classes designed to encour age appropriate beha viors and attitudes towards college (University Experience, Lear n to Learn, etc.) would be another way to find students who had high worry and hi gh emotionality, low worry and low emotionality, or who were high in one factor but low in another to determine the characteristics of these differe nt types of characters. Once studies like these have been carried out, the potential uses of this measure are as varied as those for test anxiety in terms of research (e.g., techniques to alleviate anxiety) although ultimately use of the measure in a clinical setting would be most useful. Those students who are suffering from study anxi ety can become aware of the fact that study anxiety is affecting their learning and d eal with it using met hods researched using this measure. The mean item and scale scores for the engineering students were statistically significantly lower than the other three colleg es. Invariance testing across colleges also is needed to determine whether the lower scor es of the students in the College of Engineering were due to variation in the way students from that college interpreted the items rather than an actual diffe rence in their level of anxiety. Although construct validity requires evidence from diffe rent sources, similar studies to this one could be carried out and examination of the item order effect could be carried out by introducing the pairs of items showing significant correlated error in different places on the questionnaire. Furt her expansion of the nomological network by including measures relating to facilitative and debilitative anxi ety, individual coping

PAGE 160

153 styles, social desirability, a nd life-styles may introduce furthe r explanations for the worry and emotionality differences observed in the current research. In conclusion, the current study provides evidence that the two-factor solution using the 16 items of the SAI is an accepta ble conceptualization of this scale for both men and women. Tests of invarian ce revealed that the factor ial structure of the SAI was invariant across gender, thus providing good s upport for the validity of inferences made from responses to this instrument. As pred icted, scores from the Study Anxiety Inventory were related to measures of test anxiety, tr ait anxiety, curiosity, pa ssive procrastination and active procrastination. The SAI was not show n to be related to scores from the study skills and habits measure. Overall, the result s from this study provide support for the use of the SAI as a research tool for exam ining study anxiety in male and female undergraduate college students.

PAGE 161

154 References Ackerman, D., & Gross, L. (2005). My instruct or made me do it: Task characteristics of Procrastination. Journal of Marketing Education, 27 (2), 5-13. Alpert, R., & Haber, R. N. (1960). Anxiet y in academic achievement situations Abnormal Social Psychology, 61, 207-215 American Educational Research Associat ion, American Psychological Association & National Council on Measurement in Education (1999). Standards for educational and psychological testing. Washington D.C.: American Education Research Association. American Psychiatri c Association (1994). The diagnostic and statistical manual of mental disorders (4th ed.) Washington, D.C.: Author. Anton, W. D., & Reed, J. R. (1991). College adjustment scales professional manual Odessa, Florida: Psychological Assessment Resources, Inc. Anxiety disorders (2001). Retrieved Oct. 21, 2005, from Mental Help Net Web site: http://www.mentalhelp.net/poc /center_index.php?id=1&cn=1. Armstrong, K. A., & Khawaja, N. G. (2002) Gender differences in anxiety: An investigation of the symptoms, cognitions and sensitivity towards anxiety in a nonclinical population. Behavioural and Cognitive Psychotherapy, 30, 227–231. Atkinson, J. (1974). The mainsprings of achieve ment-oriented activity. In J. W. Atkinson & J. Raynor (Eds.), Motivation and achievement (pp. 13-41). New York: Wiley. Berlyn, D. E. (1960). Conflict, arousal, and curiosity New York: McGraw-Hill.

PAGE 162

155 Bernstein, D. A., Roy, E. J., Srull T. K., & Wickens, C. D. (1988). Psychology. Boston, MA: Houghton Mifflin Company. Beswick, G., Rothblum, E. D., & Mann, L. ( 1988). Psychological ante cedents to student procrastination. Australian Psychologist, 23, 207–217. Bolger, N. (1990). Coping as a persona lity process: A prospective study. Journal of Personality and Social Psychology, 59 525-537. Bollen K. (1989). Structural equations w ith latent variables Oxford, England: John Wiley & Sons. Breslau, N., Schultz, L., & Peterson, E. (1995). Sex differences in depression: A role for preexisting anxiety. Psychiatry Research, 58 1-12. Brinker, J., & Dozois, J. (2009). Rumi native thought style and depressed mood. Journal of Clinical Psychology, 65, 1-19. Brown, W. R., & Holtzman, W. H. (1984). The manual of the survey of study habits and attitudes New York: Psychological Corporation. Byrne, B. (1989). A primer of LISREL: Basic applications and programming for confirmatory factor analytic models New York, NY, US: Springer-Verlag Publishing. Campbell, D., & Fiske, D. (1959). Converg ent and discriminant validation by the multitrait-multimethod matrix. Psychological Bulletin, 56, 81-105. Cannon, W. B. (1929). Bodily change in pain, hunger, fear, and rage. New York: Appleton.

PAGE 163

156 Carlsson, M., & Carlsson, A. (1988). In vivo evidence for a greater brain tryptophanhydroxylase capacity in female than in male rates. Naunyn-Schmiedeberg’s Archives of Pharmacology, 338, 345-349. Chu, A., & Choi, J. (2005). Rethinking procrastination: Positive effects of “active” procrastination behavior on attitudes and performance. Journal of Social Psychology, 145 (3), 245-264. Covington, M. V., & Omelich, C. L. (1985). Ab ility and effort valuation among failureavoiding and failure-accepting students. Journal of Educational Psychology, 77, 446-459. Cronbach, L. J. (1949). Essentials of psyc hological testing. New York: Harper & Row. Cronbach, L., & Meehl, P. (1955). Constr uct validity in psychological tests, Psychological Bulletin, 52 (4), 281-302. Darwin, C. (1965 version, or iginal publication 1872). The expression of emotions in man and animals. Chicago, IL: University of Chicago Press. Diamond, D. M., Park, C. R., Heman, K. L., & Rose, G. M. (1999). Exposing rats to a predator impairs spatial working memory in the radial arm water maze. Hippocampus, 9, 542–552. DiStefano, C., & Hess, B. (2005). Using conf irmatory factor analysis for construct validation: An empirical review. Journal of Psychoeducational Assessment, 23, 225-241.

PAGE 164

157 Everson, H. T., Millsap, R., & Rodriguez, C. (1991). Isolating gender differences in test anxiety: A confirmatory factor analysis of the Test Anxiety Inventory. Educational and Psychological Measurement, 51, 243-251. Eysenck, H. J. (1979). Anxiety, learni ng and memory: A reconceptualization. Journal of Research in Personality, 13, 363-385. Eysenck, H. J. (1991). Anxiety: The cognitive perspective Hove, England: Lawrence Erlbaum associates Ltd. Eysenck, M. W., & Byrne, A. (1992). Anxi ety and susceptibility to distraction. Personality & Individua l Differences, 13, 793-798. Eysenck, M. W, MacLeod, C., & Mathews, A. (1987). Cognitive functioning in anxiety. Psychological Reports, 49, 189-195. Ferrari, J. R., Johnson, J. L ., & McCown, W. G. (1995). Procrastination and task avoidance: Theory, research, and treatment. New York: Plenum Press. Fiely, D. (2007, January 20). Caffeine crutch : Young people's growing need for ‘fix’ a concern among medical experts. The Columbus Dispatch, p. News 1A. Fontana, D. (1986). Students’ attributes of academic success. Assessment and Evaluation in Higher Education, 20, 345-365. Foot, M., & Koszycki, D. (2004). Gender differen ces in anxiety-related traits in patients with panic disorder. Depression and Anxiety, 20, 123-130. Freud, S. (1936). The problem of anxiety. New York, NY: W. W. Norton & Company.

PAGE 165

158 Gadzella, B., Goldston, J., & Zimmerman, M. (1976, April). Interrelationships of study habits and attitudes, locus of contro l, motivation achievement and academic achievement. (ERIC Document Reproduction Service No. ED 143928). Gonzalez, H. P. (1978). Effect of syst ematic desensitization, study counseling, and anxiety coping training in the treatment of test anxious st udents with good and poor study habits (Doctoral dissertation, University of South Florida, 1978). Dissertation Abstracts International, 39 1955B. (University Microfilms No. 7818997). Gonzalez, H. P. (1995). Systematic desensi tization, study skills c ounseling, and anxietycoping training in the treatment of test a nxiety. In C. D. Spielberger and P. Vagg (Eds.), Test anxiety: Theory, assessment, and treatment (pp. 117-132) Washington D. C.: Taylor and Francis. Graziano, A., & Raulin, M. (2007). Research methods: A process of inquiry (6th Ed.). New York, NY: Pearson Education, Inc. Hewitt, P. L., & Norton, G. N. (1993). The Beck Anxiety Inventory: A psychometric analysis. Psychological Assessment, 5, 408-412. Hong, P. (2007, January 21). How do you react when the person around the corner has been convicted of abusing his daughter? Fort Worth Star-Telegram, p. E1. Hopkins, K., & Chambers, A. (1966). Anxiety, physiologically and psychologically measured, and its consequences on mental test performances (HEW Contract No. OE-5-10-324). Los Angeles: Univer sity of Southern California.

PAGE 166

159 Howard, J. R. (1993). Study behaviors of undergraduate and graduate students. (ERIC Document Reproduction Service No. ED360581). Hoyle, R., & Panter, A. (1995). Writing about st ructural equation models. In R. H. Hoyle (Ed.), Structural equation modeling: C oncepts, issues, and applications (pp. 158176). Thousand Oaks, CA, US: Sage Publications, Inc. Hu, L., & Bentler, P. M. (1999). Cutoff crite ria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55. Huck, S. W., & Jacko, E. J. (1974). Effect of varying the response format of the AlperHaber Achievement Anxiety Test. Journal of Counseling Psychology, 21 (2), 159163. Kalechstein, P., Hocevar, D., Zimmer, J. W., & Klechstein, M. ( 1989). Procrastination over test preparation and test anxiety. In R. Schwarzer, H. M. Van der Pavlov, & C. D. Spielberger (Eds.), Advances in test anxi ety research: Vol. 6 (pp. 63-76). Lisse, The Netherlands: Swets & Zeitlinger. Keiffer, K., Reese, R. J., & Cronin, C. (2004). A psychometric evaluation of the Test and Study Attitudes Inventory. Paper presented that the APA Conference. Washington, D. C. Lay, C., Edwards, J., Parker, J., & Endler, N. (1989). An assessment of appraisal, anxiety, coping and procrastination dur ing an examination period. European Journal of Personality, 3 195-208.

PAGE 167

160 Lay, C. H. (1986). At last my res earch article on procrastination. Journal of Research in Personality, 20, 474–495. Lee, D., Kelly, K., & Edwards, J. (2005). A closer look at the relationships among trait procrastination, neuroticism and conscientiousness. Personality and Individual Differences, 40 27-37. Levitt, E. (1967). The psychology of anxiety Indianapolis, IN: Bobbs-Merrill. Liebert, R. M., & Morris, L. W. (1967). C ognitive and emotional components of test anxiety: A distinction and some initial data. Psychological Reports, 20, 975-978. Lunsford, G. (2001). Effect of test anxiety, study anxiety, study hab its, test taking skills and academic problems on GPA Unpublished Master’s Thesis, University of South Florida, Tampa. Marcus, D. (2001). Gender differences in tr eatment-seeking chronic headache sufferers. Headache: The Journal of Head and Face Pain, 41, 698-703. May, R. (1950). The meaning of anxiety New York: W. W. Norton & Company. McCown, W. G., & Johnson, J. L. (1991). Pers onality and chronic procrastination by university students during an academic examination period. Personality and Individual Differences, 12 413-415. Messick, S. (1989). Validit y. In R. L. Linn (Ed.), Educational measurement (3rd ed., pp. 13-103). New York: Macmillan. Milgram, N., & Tenne, R. (2000). Personality correlates of decisi onal task avoidant procrastination. European Journal of Personality, 14 141-156.

PAGE 168

161 Muthn, L. K., & Muthn, B. O. (1998). Mplus users guide. Los Angeles, CA, Muthen & Muthen. Naveh-Benjamin, M. (1991). A comparison of training programs intended for different types of test anxious students: Furthe r support for an information processing model. Journal of Educational Psychology, 83, 134-139. Nelson, R., Hay, L., Devany, J., & Koslow-Gree n, L. (1981). The reactivity and accuracy of children's self-monitoring: Three experiments. Child Behavior Therapy, 2 1 24. New Zealand Council for Educational Research (1979). SHEIK professional manual. Wellington, New Zealand: NZCER. Onwuegbuzie, A. (2000). Academic proc rastination and st atistics anxiety. Assessment & Evaluation in Higher Education, 29, 3-19. Owens, A., & Newbegin, I. (1997). Procrastin ation in high school achievement: A causal structural model. Journal of Social Behavior and Personality, 12, 869-887. Pace, A., Walters, K., & Sherk, J. (1990). Wh at determines course achievement? An investigation of several possible influences on academic outcomes. National Reading Conference Yearbook, 39 347-351. Pitts, F. (1971). Biochemical factors in anxiety neurosis. Behavioral Science, 16 (1), 8291. Prevatt, F., Pretscher, Y., Proctor, B., Hurst, A., & Adams, K. (2006). The Revised Learning and Study Strategies Inventory: An Evaluati on of Competing Models. Educational and Psychological Measurement, 66, 448-458.

PAGE 169

162 Rau, W., & Durand, A. (2000). The academic et hic and college grades: Does hard work help students to "make the grade"? Sociology of Education, 73, 19-38. Razavi, T. (2001). Self-report measures: An overview of concerns and limitations of questionnaire use in occupational stress research. In Papers. University of Southampton, England: Dept of Accounting and Management. Roysdon, K. (2006, September 10). Business ex po will offer networking, and cure for networking fears. The Star Press, p. Business 1F. Rudner, L. M. (1992). Pre-employmen t testing and employee productivity. Public Personnel Management, 21, 124-136. Sallis, J., & Owen, N. (1999). Physical activity and behavioral medicine. Thousand Oaks, CA Sage Publications. Sarason, I. G., & Smith, R. E. (1971). Personality. Annual Review of Psychology, 22, 393-446. Sarason, S. B., & Mandler, G. (1952). Some correlates of test anxiety. Journal of Abnormal and Social Psychology, 47, 810-817. Sarason, S. B., Hill, K. T., & Zimbardo, P. ( 1964). A longitudinal study of the relation of test anxiety to performance on in telligence and achievement tests. Monographs of the Society for Research in Child Development, 29 1-51. Schmitt N., & Kuljanin G. (2008). Measuremen t invariance: Review of practice and implications. Human Resource Management Review, 18, 210-222. Seeman, M. V. (1997). Psychopathology in wo men and men: Focus on female hormones. American Journal of Psychiatry, 154, 1641-1647.

PAGE 170

163 Shiffman, S., Stone, A., & Hufford, M. (2008).Ecological momentary assessment. Annual Review of Clinical Psychology, 4, 1-32. Silverman, M. S., & Riordan, K. (1974). Study skills training for "high risk" freshmen. Journal of College Student Personnel, 13, 63. Solomon, L. J., & Rothblum, E. D. (1984). Academic procrastination: Frequency and cognitive behavioral correlates. Journal of Counseling Psychology, 31, 503–509. Spielberger, C. D. (1966). The effects of anxiety on complex learning and academic achievement. In C. Spielberger (Ed.), Anxiety and Behavior (pp. 361–398). New York: Academic Press. Spielberger, C. D. (1975). The measurement or state and trait anxi ety: Conceptual and methodological issues. In L. Levi (Ed.), Emotions—Their Parameters and Measurement (pp. 713-725). New York: Raven Press. Spielberger, C. D. (1980). Test anxiety inventory: Pr eliminary professional manual Palo Alto, CA: Consulting Psychologists Press. Spielberger. C. D., Jacobs. G. A., Crane. R. S., Russell. S. F., We stberrv, L., Barker, L., Johnson, E. H., Knight, J., & Marks, E. (1979). State-trait Personality Inventory Tampa: Human Resources Institute, University of South Florida. Spielberger, C. D., & Starr, L. M. (1994) Curiosity and exploratory behavior. In H. O'Neil, Jr. & M. Drillings, (Eds.), Motivation: Theory and research (pp. 221243). Hillsdale, NJ: Lawrence Erlbaum Associates. Spielberger, C. D., & Wasala, C. F., (1995). Preliminary manual for the state trait personality inventory Tampa, FL: Psychological Assessment Resources, Inc.

PAGE 171

164 Spielberger, C. D., Gorusch, R. L., & Lushene, R. E. (1970). State-trait anxiety inventory: Preliminary professional manual Palo Alto, CA: Consulting Psychologists Press. Steenkamp, J., & Baumgartner, H. (1998). A ssessing measurement invariance in crossnational consumer research. Journal of Consumer Research, 25, 78-90. Stevens, J. (1996). Applied multivariate statistics for the social sciences (3rd Ed.). Mahwah, NJ: Lawrence Erlbaum Associates, Inc. Stone, M. H. (1996). Healing the mind: A history of ps ychiatry from antiquity to the present. New York, NY: W. W. Norton & Company. Taylor, J. A. (1953). A personal ity scale of manifest anxiety. Journal of Abnormal and Social Psychology, 48, 285-290. Tobias, S. (1985). Test anxiety: Interferen ce, defective skills and cognitive capacity. Educational Psychologist, 3, 135-142. Trapey, E. R., & Harris, J. B. (1979). Study skills courses make a difference. Journal of College Student Personnel, 20, 62-66. Vagg, P., & Papsdorf, J. D. (1995). Cognitive therapy, study skills training, and biofeedback in the treatmen t of test anxiety. In C. D. Spielberger & P. Vagg (Eds.), Test anxiety: Theory, assessment, and treatment. Washington D. C.: Taylor and Francis. Vandenberg, R. J. (2002). Toward a furthe r understanding of an improvement in measurement invariance methods and procedures. Organizational Research Methods, 5 139-158.

PAGE 172

165 Vandenberg, R. J., & Lance, C. E. (2000). A review and synthesis of the measurement invariance literature: Suggestions, practices, and recommendations for organizational research. Organizational Research Methods, 3, 4 70. Wachtel, P. L. (1967). Conceptions of broad and narrow attention Psychological Bulletin, 68, 417-429. Ware, W. B., Galassi, J., & Dew, K. (1990). The test anxiety inventory: A confirmatory factor analysis Anxiety Research, 3 (3), 205-212. Weinstein, C. E., & Palmer, D. R. (2002). Learning and study strategies inventory (LASSI) user’s manual (2nd ed.). Clearwater, FL: H & H Publishing. Welsh, M., Bachelor, P., & Wright, C. (1990). The exploratory and confirmatory factor analyses of the latent structure of th e Study Attitudes and Methods Survey for a sample of 176 eighth-grade students. Educational and Psychological Measurement, 50 369-376. Williams, J., Watts, F., MacLeod, C., & Mathews, C. (1988). Cognitive psychology and emotional disorders Oxford, England: John Wiley & Sons. Wilson, M., & Biscardi, R. ( 1994). Sex-differences in gaba benzodiazepine receptor changes in corticosterone rel ease after acute stress in rats. Experimental Brain Research, 101, 297-306. Zeidner, M. (1998). Test anxiety: The state of the art New York, NY: Plenum Press.

PAGE 173

166 Appendix A: Summary of findings from studies using the SAI

PAGE 174

167 Appendix A: Summary of findings from studies using the SAI Authors Alphas Retest Reliabilities Other Statistically Significant Correlates Lunsford (2001) Study using students from two colleges in a Florida state university .96 overall .94 for each subscale For 2 days 81, .82 and .82 For 2 months .84, .82, & .83 for the Overall, Worry and Emotionality scales, respectively EFA using n=536 Showed 2 factors Test Anxiety TA-Worry TA-Emotionality Trait Anxiety Trait Depression Trait Curiosity Trait Anger Testwiseness Academic Problems Self-Esteem Intelligence Kieffer, Reese, & Cronin (2004) Study using students from three university locations .976 overall and .92-.94 for subscales For 10 weeks .88 overall and .67-.81 for subscales Corrected item to total corr .56-.84 EFA using n=512 Showed 2 factors CFA using n=1025 Showed 2 factors Kieffer, Cronin & Gawet, (2006) Overall not reported .83-.87 for subscales Examined the relationship between SAI, TAI and Reasons for Drinking for 365 students. Social Camaraderie Mood enhancement Tension reduction Draper (2001) Study using student from a dorm in a Florida state university Examined the relationship between SAI, TAI and GPA for 200 college students living on campus. Test Anxiety TA-Worry TA-Emotionality Trait Anxiety ACT SAT Note: GPA = Grade Point Average SAT = Scholastic Aptitude Test ACT = American College Test TAI = Test Anxiety Inventory SAI = Stud y Anxiety Inventory TA = Test Anxiety

PAGE 175

168 Appendix B: Preamble to data collection Explanation of the study and wh at the consent form says Educational Debriefing

PAGE 176

169 Appendix B continued Explanation of the study and w hat the consent form says. Hello. My name is Douglas Lunsford and I am a graduate student in the Research and Measurement Department. I am here to ask you to fill out a questionnaire designed to find out how you describe yourself regardi ng your behavior, thoughts and sensations. You will find that some of the questions ask you to consider these while studying and very similar questions will ask for these while taking tests. Please keep these in mind while answering questions about your view of yourself generally. Analysis of your answers will help to find the relationship be tween these different items and will help education majors and psychologists devel op better programs for understanding them. You do not need to complete any consent form s as your name will not be taken so there are no risks associated with participation in this study. If you are in any way concerned and do not wish to participate, merely turn the blank form back in. No record will be kept to show that you did not participate. On th e questionnaire, which is set up like a scantron, blacken in the circles in the co lumn that most applies to you. (For participating, you will rece ive one extra credit point that can be put toward your grade in this class.) Remember this is entire ly voluntary so you may withdraw at any time without fear of reprisal. (The re is no other compensation th an the extra credit point for completing the whole measure.) Afterwards you will be given a sheet explaining the items that we expect are associated and how you can contact me to find what the overall results of the study show us. Both my te lephone number and the number of the Division of Compliance Services are on your copy of the consent form. If you have any questions, I will be here to answer them

PAGE 177

170 Appendix B Continued EFFECT OF STUDY ANXIETY ON ACADEMIC ACHIEVEMENT Educational Debriefing The goals of this research are to evaluate the responses of the Study Anxiety Inventory and to relate those responses to the res ponses to other measures. The other anxiety measures that have been used in this study are the Test Anxiety I nventory (TAI) and the trait anxiety scale of the State-Trait Pers onality Inventory (STPI). Study habits were assessed by the Study for Exam (SH) scale from the Study Habits Evaluation and Instruction Kit (SHEIK). Procrastination was measured by the Ac tive Procrastination Scale and the Passive Procrastination Scale. We expect to find that anxiety during studying and during test taking is negatively correlated with active procrastination and positively correlated with passive procrastina tion. If you would like to find out what the results are for this study, you may ca ll Douglas Lunsford at ________________or attend our debriefing meeting which will be held at the offices of Dr. Dedrick on Monday, August 8, 2007 at 4 p.m. If you would like to re ad more about this subject, you will find that the below references are exceptional wo rks which give a very in-depth background. Thank you for participating in this study. Spielberger, C. D. (1976). The effect of anxiety on complex learning and academic achievement. In C. D. Spielberger (Ed.), Anxiety and behavior NY Academic Press. Zeidner, M (1998). Test Anxiety: The state of the art NY: Plenum Press.

PAGE 178

171 Appendix C: Experimental Measures This Appendix includes the Study Anxiety Inve ntory, Test Anxiety Inventory, the trait anxiety and trait curios ity scales from the State-Trait Anxiety Inventory Form Y-2, the Study Habits and Test-Taking Skills scal es from the Study Habits Evaluation and Instruction Kit, the Active Procrastination Sc ale and the Passive Procrastination Scale. A scoring guide for these measures is also provided in this appendix.

PAGE 179

172 Appendix C Continued Code number: __ __ __ __ __ Age: ____ Sex: __M, ___F, Today’s Date ___________ Ethnic Code: African American or Black, Asian American/Asian/Pacific Islander, Hispanic or Latino, White, American Indian/Alaska Native Two or More Races Other ________ College_______________ Department _________________ Major_________________ Have you attended either the University E xperience or the Counse ling center to gain learning or studying skills? ______________________ Directions: The Study Attitudes Inventory (SAI) presents a number of statements which people have used to describe themselves while studying for tests are given below. Read each statement and then blacken the appropriate circle to the right of the statement to indicate how often it generally applies to you “while you are studying for an exam.” There are no right or wrong answers. Do not spend too much time on any on e item. Give the answer which seems best to describe your thoughts and feelings while studying for an exam. Almost Almost Never Sometimes Often Always 1. I feel very uneasy just before starting to study for an exam 2. While I am studying for an exam I often think “I’m not getting this” 3. I can’t keep my mind on the subject when studying for an exam because other thoughts interfere 4. Even when I have plenty of time, I feel nervous when I try to study for an exam 5. While studying for exams, I have an uneasy, upset feeling 6. While studying for tests, othe r thoughts interfere with my learning 7. While studying for a test, I wo rry about not being able to learn the material 8. I wish studying for tests did not upset me so much 9. I feel panicky when studyi ng for an important exam 10. While studying for exams, I am stressed with thoughts like “I can’t absorb the material properly” 11. I worry so much when I study for a test that I do things that distract me 12. While studying for exams, I feel very tense 13. I freeze up while studying for an important test 14. When I study for exams, I seem to get a mental block that keeps me from absorbing the material 15. When I am studying for a test, I can’t get my brain to organize the information 16. I feel jittery while study ing for important exams

PAGE 180

173 The Test Attitudes Inventory (TAI) evaluates thoughts and f eelings that are experienced by students when taking or studying for exam inations. A number of statements which people have used to describe themselves wh ile taking tests are given below. Read each statement and then blacken the appropriate circ le to the right of the statement to indicate how often it generally applies to you “while you are taking an exam.” There are no right or wrong answers. Do not spend too much time on any one item. Give the answer which seems best to describe your thoughts and feeling while taking an exam. Almost Almost Never Sometimes Often Always 1. While taking examinations I have an uneasy, upset feeling 2. Thinking about my grade in a course interferes with my work on tests 3. I worry and freeze up on important exams 4. During exam, I find myself thin king about whether I’ll get through school 5. The harder I work at taking a test, the more confused I get 6. Thoughts of doing poorly interfere with my concentration on tests 7. I feel jittery when taki ng an important test 8. Even when I’m well prepared for a test, I feel very nervous about it 9. During an exam, I start feeli ng uneasy about not doing well 10. During tests I feel very tense 11. I wish examinations did not upset me so much 12. I seem to defeat myself while working on important tests 13. I feel very panicky when I take an important test 14. During test, I find myself thinking about the consequences of failing 15. I feel my heart beating very fast during important tests 16. During examinations, I get so ne rvous that I forget facts I really know

PAGE 181

174 A number of statements that people have used to describe themselves are given below. Read each statement and then darken the appropriate valu e to the right of the statement to indicate how you generally feel There are no right or wrong answers. Do not spend too much time on any one statement but give the answer which seems to describe how you generally feel. Almost Almost Never Sometimes Often Always 1. I am a steady person 2. I feel like exploring my environment 3. I feel satisfied with myself 4. I am curious 5. I get in a state of tension or turmoil as I think over my recent concerns and interests 6. I feel interested 7. I wish I could be as happy as others seem to be 8. I feel inquisitive 9. I feel like a failure 10. I feel eager 11. I feel nervous and restless 12. I am in a questioning mood 13. I feel secure 14. I feel stimulated 15. I lack self-confidence 16. I feel disinterested 17. I feel inadequate 18. I feel mentally active 19. I worry too much over someth ing that really does not matter 20. I feel bored

PAGE 182

175 The following statements refer to how you study for an ex amination such as a midterm or final exams. In these statements, the term ‘multiple-choice exam’ includes exams w ith true-false or multiple choice questions, which require picking the correct answer out of four or five alternatives The term 'essay exam' refers to exams where you have to write an extended answer, e.g., an essay or paragraph. If the statement does not specify an essay or multiple-choice exam, then consider it to be about both types. There ar e no right or wrong responses for the statements in this inventory. Please read each statement and indicate how often these statements generally apply to by blackening in the circle that most applies to you. Darken for NEVER or ALMOST NEVER. Darken for about of the time. Darken for about of the time. Darken for about of the time Darken for ALWAYS or ALMOST ALWAYS Almost About About About Almost Never of the time of the time of the time Always 1. When an exam is near I spend more time doing homework and studying than I do normally 2. I start to study for the exam at least two days before it…...… 3. I do not read my notes over at all………………………………… 4. When preparing for an exam I study for it on at least two separate occasions…..……………… 5. I think up questions which might be asked in the exam and see if I can answer them……… 6. I rewrite at least part of my notes…………………………… 7. I do not study for an exam at all……………………..……… 8. I use memory aids such as rhymes and mnemonics to help me remember things 9. I try to find out as much as I can beforehand about the exam 10. If I do any study for an exam it is only on the day of the exam 11. Before an exam I try to find out how many questions will be asked, what kinds they will be, etc. 12. I study by asking questions of other students and by answering their questions………..……… 13. If appropriate old exam papers are available, then I look to see if I can answer the questions 14. I do the same amount of study for a multiplechoice exam as I would for an essay exam 15. If I read over my notes at all, I do it only once………………… 16. I memorize rules, definitions and formulae…………………… 17. I concentrate on specific tasks rather than main ideas……….

PAGE 183

176 18. I read over my notes several times……………………………… 19. I do less than one hour’s study for an exam………………… 20. I rewrite my notes in the form of a summary……………… 21. I skim read the parts of the textbooks which cover what the exam will be on….……………. 22. When I am studying for an exam I concentrate on those parts I already know….……………... 23. I try to guess what questions are likely to be asked…………… 24. I read through the impor tant facts more than once……………. 25. I make sure I know what topics the exam will be on…………

PAGE 184

177 Instructions: A number of statements are listed below which people have used to describe themselves. Read each statement and then blacken the appropriate circle to the right of the statement to indicate how true you generally feel or react in the manner described. There are no right or wrong answers. Do not spend too much time one any one statement but give the answer that seems best to describe how you generally feel or react. s Not at all true Very true 1 I tend to work better under pressure + 2 Even though I tend to work on papers or study for exams at the last moment, I am still motivated to do my best 3 Since I often start working on things at the last moment, I have trouble finishing assigned tasks most of the time + 4 It is hard to keep myself motivated while working against impending deadline. + 5 I feel like giving up the task when I know there is no way that I can finish it on time + 6 I intentionally put off work to maximize my motivation 7 To use my time more efficiently, I deliberately postpone some tasks 8 I am unsatisfied with the outcome of my work when I put it off until the last moment + 9 I am more focused and motivated while I am working against the impending deadline 10 I find the return for working under deadline is great 11 I tend to do things at the last minute and often find it difficult to complete them on time + 12 I feel that putting work off until the last minute does not do me any good + 13 I tend to finish tasks well ahead of deadlines 14 Even after I make a decision I delay acting upon it + 15 I prepare to study at some point of time but don’t get any further + 16 I tend to leave things until the last minute + 17 I often find myself performing tasks I intended to do days earlier + 18 I generally delay before starting on work I have to do + Copyright Jin Nam Choi Contact: jinnam.choi@mcgill.ca

PAGE 185

178 Scoring for the measures Measure Positively Scored Items Negatively Scored Items SAI Emotionality 1,4,5,8,9,12,13,16 SAI Worry 2,3,6,7,10,11,14,15 TAI Emotionality 1,7,8,9,10,11,13,15 TAI Worry 2,3,4,5,6,12,14,16 T-Anx 5+,7+,9+,11+,15+,17+,19+ 1-,3-,13-, T-CY 2+,4+,6+,8+,10+,12+,14+,18+, 16-,20Study for Examinations 1+,2+,4+,5+,6+,8+,9+,11+,12+, 13+,16+,18+,20+,21+,23+,24+,25+ 3-,7-,10-,14-,15-,17-,19-,22Active Procrastination 3,4,5,8,11,12 1-,2-,6-,7-,9-,10Passive Procrastination 14,15,16,17,18 13

PAGE 186

179 About the Author George Douglas Lunsford returned to unive rsity after a 35 year break to receive a B.A. in Psychology from the University of Sout h Florida in 1997 and an M.A. in Clinical Psychology from the same university with fa ithful support from his wife, Nancy. While working for his master’s degree, he worked as both a research and t eaching assistant after which he became an adjunct professor of psychology, social science statistics, and research methods and entered USF’s College of Education to obtain his Ph.D. in Educational Measurement and Research from the Measurement and Evaluation Department. My education was enriched by Major Professor, Robert F. Dedrick, and committee members Bruce W. Hall, Jeffr ey D. Kromrey, and James A. Eison.