SENIOR PROJECT AS A REMEDY FOR â€œSENIORITISâ€: WHAT ARE THE ACTIVE INGREDIENTS? By BRYAN P. DUFF A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2006
Copyright 2006 by Bryan P. Duff
iii ACKNOWLEDGMENTS First I wish to thank the members of my committee: Elizabeth Bondy, Tracy Linderholm, David Miller, and Elizabeth Yeag er. I could not have asked for a more supportive and congenial group of mentors. I am deeply grateful to James Algina for his assistance in the early phases of my learning about structural equation modeli ng (SEM); and to Jason Cole and other SEMNET folks for their assistance with tec hnical and conceptual aspects of SEM. I also wish to thank the University of Fl orida College of Education for its generous financial support during my te nure as a graduate student. Fran Vandiver, Randy Scott, and Brian Marchman at P.K. Yonge Developmental Research School deserve special thanks for their support of my Senior Project work. My parents deserve more appreciation than I can convey in one lifetime. They set the bar high as professionals and human be ings. Finally, and most importantly, I acknowledge my wife, Kerin, and our little man, Liam, for inspiring me to do this work and do it efficiently.
iv TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iii LIST OF TABLES...........................................................................................................viii LIST OF FIGURES...........................................................................................................ix ABSTRACT.......................................................................................................................xi CHAPTER 1 INTRODUCTION TO THE PROBLEM.....................................................................1 Senior Year in the Educational Spotlight.....................................................................1 Symptoms of Senioritis.................................................................................................4 Attitudes of Senioritis............................................................................................5 Attendance Problems.............................................................................................8 Minimal Time on Homework................................................................................9 Course Enrollment Patterns.................................................................................10 Sequelae of Senior-Year Disengagement...................................................................10 Effects on College Performance..........................................................................10 Effects on Standardized Test Performance..........................................................12 Potential Remedies for Senioritis...............................................................................15 Senior Project..............................................................................................................18 Rationale for the Study...............................................................................................24 Prospectus...................................................................................................................25 2 REVIEW OF LITERATURE.....................................................................................27 Engagement: The Aim of the Remedy.......................................................................27 Definition of Academic Engagement..................................................................27 Educational Value of Academic Engagement.....................................................29 Academic Engagement on the Decline................................................................33 Promoting Engagement: An Overview of Ingredients...............................................35 Definitions of Major Variables............................................................................35 Hypotheses..........................................................................................................36 General Conceptual Framework..........................................................................37 Newmann and authentic pedagogy..............................................................38 Self-efficacy theory......................................................................................40
v Expectancy value theories.........................................................................42 Self-determination theory.............................................................................45 Section Summary and Prospectus.......................................................................47 A Baseline for Engagement: Inactive Ingredients......................................................48 Effect of Academic Engagement on Senior Project Engagement.......................48 Effect of Previous Mastery Experi ence on Senior Project Self-Efficacy............50 Promoting Engagement: Active Ingredients...............................................................52 Effect of Senior Project Self-Effi cacy on Senior Project Engagement...............53 Effect of Utility Value on Senior Project Engagement.......................................59 Effect of Autonomy on Senior Proj ect Engagement and Self-Efficacy..............66 Effect of Clarity of Expectations on Senior Project Engagement and SelfEfficacy............................................................................................................77 Effect of Sensitivity on Senior Project Engagement...........................................86 Effect of Novelty on Senior Project Engagement...............................................92 Effect of Advisor Support on Senior Project Engagement and Self-Efficacy.....97 Effect of Peer Support on Senior Project Engagement.....................................106 Effect of Parent Support on Senior Project Engagement and Self-Efficacy.....112 Chapter Summary and Prospectus............................................................................121 3 RESEARCH DESIGN AND METHODS................................................................123 Instrument.................................................................................................................123 Content and Format of Survey..........................................................................123 Pilot-Testing of Survey......................................................................................127 Participants...............................................................................................................128 Recruitment of Schools.....................................................................................128 Characteristics of Schools.................................................................................130 Attrition of Schools...........................................................................................130 Final Sample Information..................................................................................130 Procedure..................................................................................................................131 Administration of Survey..................................................................................131 Screening and Preparation of Data....................................................................132 Data Analysis............................................................................................................133 Overview...........................................................................................................133 Imputation of Missing Data...............................................................................138 Split of Dataset..................................................................................................139 Refinement of Measurement Scales..................................................................140 Assumption Checks...........................................................................................146 Linear relationships....................................................................................146 Multivariate normality................................................................................147 Independence of cases................................................................................150 Primer to Aid Interpretation of Model Analyses...............................................153 Initial Model Testing.........................................................................................156 Cross-Validation of Model................................................................................158 Model Stability with Cluster Adjustment..........................................................160 Limitations of Methods.............................................................................................160 Chapter Summary.....................................................................................................165
vi 4 RESULTS.................................................................................................................166 Descriptive Statistics................................................................................................166 Model Testing and Modification..............................................................................171 Model 1 Testing.................................................................................................172 Model 2 Testing.................................................................................................176 Model 3 Testing.................................................................................................178 Model 4 Testing.................................................................................................181 Model 5 Testing.................................................................................................182 Model 6 Testing.................................................................................................184 Model 6 Interpretation.......................................................................................185 Summary of Model-Refinement........................................................................189 Cross-Validation of Model Results..........................................................................189 Model Stability with Cluster Adjustment Results....................................................191 Chapter Summary.....................................................................................................192 5 CONCLUSION.........................................................................................................194 Interpretation and Implications of Main Findings....................................................194 The Role of Inactive Ingredients in Promoting Engagement............................194 The Role of Active Ingredients in Promoting Engagement..............................197 Support from parents..................................................................................197 Self-efficacy: Its effects and causes...........................................................200 Summary of Implications..................................................................................204 Limitations of the Study...........................................................................................206 Methodological Limitations..............................................................................206 Causal indeterminacy.................................................................................206 Variable contamination..............................................................................206 Conceptual Limitations.....................................................................................210 One-way paths............................................................................................210 Assumption of causal homogeneity...........................................................211 Omitted variables.......................................................................................213 Suggestions for Future Research..............................................................................216 APPENDIX A SURVEY ITEMS AND RESPONSE OPTIONS BY VARIABLE.........................219 B FORMATTED SURVEY.........................................................................................222 C PERMISSION TO USE UNI VERSITY WORDMARK.........................................227 D PILOT STUDY PROTOCOL APPROVE D BY THE UNIVERSITY OF FLORIDA INSTITUTIONAL REVIEW BOARD..................................................228 E DEMOGRAPHIC CHARACTERISTIC S OF RECRUITED SCHOOLS...............230 F SURVEY RETURN DATA BY SCHOOL..............................................................231
vii G MAIN STUDY PROTOCOL APPROVE D BY THE UNIVERSITY OF FLORIDA INSTITUTIONAL REVIEW BOARD..................................................232 H ITEM-LEVEL DESCRIPTIVE STATISTICS.........................................................234 I RELIABILITY STATISTICS FOR MEASURE REFINEMENT...........................235 J INPUT MATRIX FOR INITIAL MODEL TESTING............................................236 K INPUT MATRIX FOR COMPLEX SAMP LE-ADJUSTED MODEL TESTING..238 L MODEL RESULTS FOR ERROR TERMS............................................................240 REFERENCES................................................................................................................241 BIOGRAPHICAL SKETCH...........................................................................................282
viii LIST OF TABLES Table page 3-1 Demographic characteristics of the full sample.....................................................131 3-2 Number of items and reliability estimates for measurement scales.......................145 3-3 Correlations between model variab les in the calibration subsample.....................147 3-4 Skew and kurtosis statistics for model variables in the calibration subsample......149 3-5 Intraclass correlations for model variables.............................................................153 4-1 Descriptive statistics for m odel variables in the full sample..................................167 4-2 Correlations between model variables in the full sample......................................170 4-3 Parameter estimates and standard erro rs for Model 6 based on the calibration subsample...............................................................................................................186 4-4 Summary of the models and the model-refinement process..................................190 4-5 Model-fit and -comparison statistics for multigroup SEMs...................................191 4-6 Parameter estimates and standard errors for Model 6 based on the full sample....192 E-1 Schoolsâ€™ location, size, racial compos ition, and average socioeconomic status....230 F-1 Number of surveys and respons e rates for participating schools...........................231 H-1 Item-level means and standard deviations for the full sample...............................234 I-1 Alpha-if-item-removed and corrected ite m-total correlations for item-level data in the refinement subsample...................................................................................235 J-1 Variance-covariance matrix for the calibration subsample....................................237 K-1 Variance-covariance matrix for the full sample.....................................................239
ix LIST OF FIGURES Figure page 2-1 Model representing the influences on student engagement in Senior Project..........36 2-2 Model diagram highlighting the effect of Academic Engagement on Senior Project Engagement..................................................................................................49 2-3 Model diagram highlighting the effe ct of Previous Mastery Experience on Senior Project Self-Efficacy.....................................................................................50 2-4 Model diagram highlighting the effect of Senior Project Self-Efficacy on Senior Project Engagement..................................................................................................53 2-5 Model diagram highlighting the effect of Utility Value on Senior Project Engagement..............................................................................................................60 2-6 Model diagram highlighting the effect s of Autonomy on Senior Project SelfEfficacy and Senior Project Engagement.................................................................67 2-7 Model diagram highlighting the effects of Clarity of Expectations on Senior Project Self-efficacy and Se nior ProjectEngagement..............................................78 2-8 Model diagram highlighting the effect of Sensitivity on Senior Project Engagement..............................................................................................................88 2-9 Model diagram highlighting the e ffect of Novelty on Senior Project Engagement..............................................................................................................94 2-10 Model diagram highlighting the effect s of Advisor Support on Senior Project Self-Efficacy and Senior Project Engagement.........................................................98 2-11 Model diagram highlighting the effe ct of Peer Support on Senior Project Engagement............................................................................................................107 2-12 Model diagram highlighting the effects of Parent Support on Se nior Project SelfEfficacy and Senior Project Engagement...............................................................113 4-1 Path diagram for the original m odel of Senior Project Engagement......................172 4-2 Path diagram for Senior Project Engage ment model after addi ng a latent variable (â€œMethodâ€).............................................................................................................177
x 4-3 Path diagram for Senior Project Enga gement model after removing Prev Exper SP Efficacy and adding Prev Exper Acad Engage.................................................179 4-4 Path diagram for Senior Project Engage ment model after adding Acad Engage to the General Academic Experien ce (formerly Method) factor................................181 4-5 Path diagram for Senior Project Engagement model after adding Advisor Clarity.....................................................................................................................183 4-6 Path diagram for Senior Project E ngagement model after removing Parent SP Efficacy and Advisor SP Efficacy and eliminati ng Utility and Sensitivity......185
xi Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy SENIOR PROJECT AS A REMEDY FOR â€œSENIORITISâ€: WHAT ARE THE ACTIVE INGREDIENTS? By Bryan P. Duff December 2006 Chair: Elizabeth Bondy Major: Curriculum and Instruction (ISC) Academic disengagement is a common pr oblem for high school seniors. One proposed remedy is Senior Project, in which seniors study a topic of their choosing, write a research paper, develop a practical or cr eative application related to the topic, and present their work to a panel. The goal of this study was to identify the â€œactive ingredientsâ€ in this remedyâ€”the features of a studentâ€™s experience most strongly associated with greater engagement. Po tential ingredients were derived from the theoretical and empiri cal literature on academic engagement and the high school senior year. These variables were or ganized into a path model in which support variables (e.g., quality of support from the project advisor) and task-quality vari ables (e.g., degree of autonomy experienced) were hypothesized to a ffect engagement directly and in some cases indirectly via self-efficacy. To test these hypotheses, st ructural equation modeling wa s applied to survey data from a national sample of students participa ting in Senior Project (N = 1,309). Because
xii this research involved a novel c onfiguration of variables in a little-researched context, model modifications were anticipated. Theref ore, the sample was split into calibration and validation subsamples, an iterative proce ss of data-based model refinement was used on the calibration subsample, and a multigroup i nvariance analysis was used to assess the modelâ€™s stability in the validation subsample. In the model refinement, the major change was adding a latent variable to the path model to represent a general orientation to senioryear academic work. The model proved stable when applied to the full sample with a correction for the statistical non-indepe ndence of students in the same school. According to the model, patterns of varia tion in studentsâ€™ enga gement in Senior Project were best explained by parent interest and encouragement, studentsâ€™ general level of engagement in academic work, and their sense of project self-e fficacy. Efficacy, in turn, was best explained by the complementar y features of autono my and clarity of expectations. Clear expectations were mo st likely when students had strong advisor support. These findings have implications fo r theories of engagement and for the design and administration of Se nior Project programs.
1 CHAPTER 1 INTRODUCTION TO THE PROBLEM This study is grounded in a c oncern and a conviction. The concern relates to the extent of academic disengagement among high school seniors in the United States. The conviction is that alternatives to the â€œbusine ss as usualâ€ senior year have great potential to address this concern. One such alternat ive is Senior Project, an opportunity for students to conduct in-depth rese arch on a topic of their choice; to use the research to write a major paper and provide a basis for cr eating an artifact or participating in an experience related to the topic; and, finally, to present and defend their work orally to a panel of judges. Senior Proj ect is likely to engage students only if it respects general research-tested ideas about student motivation and specific issues t ypical of students in the last year of high school. The purpose of this dissertation is to id entify these potential â€œactive ingredientsâ€ and then test whether the degree to which they are present in studentsâ€™ Senior Project experiences explains their level of engagement. Senior Year in the Educational Spotlight The problem of academic disengagement during the high school senior year has received systematic attention from educa tional researchers and policymakers since the 1970s. For example, during the 1974-75 academic year, a rotating task force of twenty high school students, first assembled in 1969 by New Yorkâ€™s education commissioner, decided to focus on â€œfind[ing] solutions to the persistent problems of the 12th yearâ€ (Student Affairs Task Force, 1975, p. 6). The group conducted interviews with seniors and visited exemplary local programs, co mpiling an extensive list of senior-year
2 curriculum innovations designed to address seni oritis. A similar list emerged from the work of a former teacher and educational c onsultant (Pannwitt, 1979) who contacted over 100 high school principals across the country. A time of rela tive quiescence in research and advocacy related to senior year followe d in the 1980s, perhaps because of the strong focus on basic skills (National Commissi on on Excellence in Education, 1983). However, in the 1990s attention turned again to the issue of senior itis. For example, Hofstra University and the Nassau County Hi gh School Principals Association sponsored a series of conferences on the senior year that attracted national press attention (Kirsch, 1997). Perhaps no one has directed more interest to the senior year than Richard Riley, the U.S. Secretary of Education under President Clinton. In his seventh annual address on the state of American education, Riley (2000) noted: [The senior year] is an important time of transition for young people. Surely we can offer our young people some exciting and meaningful challenges between midterms and the senior prom. This is why I am announcing a new initiative we call the â€˜Senior Year Transitionâ€™ project. We intend to work with the Woodrow Wilson and Mott Foundations to bring toge ther university leaders, educators, parents, and yesâ€”students t ooâ€”to take a new and close l ook at the senior year of high school. The resulting collaboration, the National Co mmission on the High School Senior Year, reviewed the literature, conduc ted eight focus groups with high school graduates, heard expert testimony, held two formal meetings , and supported the development of several major papers. After issuing its first re port in January 2001 (National Commission on the High School Senior Year, 2001a), the Commissi on discussed its findi ngs with teachers, parents, and students nationwide in a series of public hearings . The second report, issued in October 2001 (National Commission on th e High School Senior Year, 2001b), outlined
3 the recommendations of the committee. (See also Conley, 2001, for a description of the Commissionâ€™s work.) What the Commission learned in its research is that even the most academically successful students often spend senior year taking easy classes, skipping class, being bored with schoolwork, and worrying more a bout part-time jobs and extracurricular activities than academics. After using th e memorable and oft-quoted description of senior year as a â€œfarewell tour of a dolescence and schoolâ€ (National Commission on the High School Senior Year, 2001a, pp. 16-17), the Commission punctuated its first report with a call-to-action reminisc ent of the strong language in A Nation at Risk (National Commission on Excellence in E ducation, 1983): â€œThe United Stat es desperately needs to seize the lost opportunity of the senior year . The need is immediate. The goal is important. The time to act has arrivedâ€ (N ational Commission on the High School Senior Year, 2001a, p. 29). The clarion call was heeded. For instan ce, the Commissionâ€™s reports and their attendant publicity paved the way for the publication of a large interview-based qualitative study of the phenomenology of senior year (Sizer, 2002). Al so in the wake of the reports, in 2002 the West Virginia Depart ment of Education sponsored a conference, â€œStrengthening the Senior Year,â€ for ove r 150 educators and business leaders (West Virginia Department of Education, 2002). A similar conference wa s held in 2003 at the University of New Mexico for teachers from 37 schools; among the recommendations emerging from the summit were increasing the stateâ€™s mathematics requirement from three to four years and ending the stateâ€™s ea rly-dismissal policy allo wing seniors to attend school part-time once they have fulfilled graduation requirements (Chmelynski, 2003).
4 Finally, in a recent report on high school reform issued by the Florida Senateâ€™s Committee on Education (2005), the committ ee strongly recommended that Florida schools â€œreinvent the senior yearâ€ (p. 8) through a combination of rigorous academic work, research projects, community service, and other programs. Senior-year reform acquired new momentum with Virginia Governor Mark Warner at the helm of the National Governors A ssociation (NGA) in 2004-05. The theme of Warnerâ€™s tenure was redesigning American high schools, and the February 2005 summit of state governors and educati on experts focused on raising state graduation requirements to match the skills demanded by colleges and employers (Peterson, 2005). A major component of Chairman Warnerâ€™s agenda was the senior year, consistent with his prior work as governor of Virginia, where in 2003 he proposed a number of initiatives designed to make grade 12 more rigorous a nd relevant (Hebel, 2003) . One outcome of the NGA summit was the online â€œRate Your Fu tureâ€ survey for high school students, which contained many questions about the seni or year and which was promoted by state governors via press releases in May 2005 (e.g., â€œGov. Guinn Proclaims,â€ 2005). The significant attention paid to the se nior year raises this question: Why do educators, researchers, and policymakers consid er the issue worth their time and effort? A closer look at the symptoms and sequel ae of senioritis brings the reasons for the attention into clear relief. Symptoms of Senioritis Senioritis has been the subject of much at tention in the national popular press, from major magazines (e.g., Kantrowitz & Wingert , 2000; Lord, 2001) to newspapers (e.g., Jayson, 2005; Kadyszewski, 2003) to radio programs (Conan, 2005). Local press appears to find the topic equally compelling, offe ring commentaries written by seniors (e.g.,
5 Stinchcomb, 2005) in addition to stories by adult reporters (e.g., Davis, 2006; Vaznis, 2005). Parenting guides written by counselors and behavioral scien tists (e.g., Kastner & Wyatt, 2002) warn parents about senioritis and provide advice on how to address it. However, even for those who have seen none of this media coverage, the words of a former high school principal probably ring true : â€œNo person who has occupied either side of a teacherâ€™s desk needs a lengthy explanat ion of what is commonly called the â€˜senior slumpâ€™â€ (Miller, 1972, p. 73; see also Kessler, 1996, and Mayher, 1998). Attitudes of Senioritis Nevertheless, examining expressions of the attitudes of senioritis, as well as the behaviors that accompany them, helps reinforc e the urgency of developing solutions. The following descriptions, in the words of seniors themselves, are typical of the voluminous anecdotal evidenceâ€”often appearing in news and school-based accountsâ€” for the high incidence and seriousness of se nioritis, even among successful students: At times, my only motivation to even go to my classes was the fact that I could only have so many absences and still be ex empt from my exams. (Senior Caroline Doud, recently accepted at the University of North Carolina-Chapel Hill, quoted in Shar, 2005, p. D1) Every single thing about high school is getting annoying. I feel Iâ€™m about to not be controlled, and theyâ€™re contro lling me. (Senior Ashley Cullum, straight-A student, quoted in Dunn, 2001, p. 12) Basically, Iâ€™m just going th rough the motions or whatev er, trying to hang up and get out of high school. (Senior with a 3.9 GPA speaking in the PBS series Senior Year , filmed at Fairfax High School in Los Angeles, quoted in What Kids Can Do, 2002, p. 1) The same disengagement dynamics are revealed by more formal ethnographic research in high schools. Two independent yearlong st udies, both (coinciden tally) conducted at highly-regarded public high schools in Californi a, provide illustrative portraits based on class observations and interviews with students and teachers.
6 Silence. The tooth-pulling session conti nues as Charmack [a teacher] doggedly goes over ground she knows they know, though they stubbornly remain quiet. No matter. Charmack understands the ebb and flow of senioritis, which, she knows, is far milder at Whitney than at many other public high schools, wher e senior year all too often can become a wasteland. Senior itis at Whitney is mostly a matter of attitude. The kids still do the work, and lots of it, although they will try every trick in the book to talk their way out of it if they can. (Humes, 2003, p. 37) Westridgeâ€™s â€œideal pupilsâ€ e xplained that they planned their high school careers, especially the senior year, so that they w ould be able to â€œkick backâ€ and relax. In classes, they were complia nt, if minimally engaged, as they exhibited a try-andmake-me posture that worked surprisingly we ll to keep teachers in check. . . . [T]heir explanation was thick with self-ri ghteousness and a sense of entitlement. Westridge students said they had been â€œgood adolescentsâ€ for a number of years by complying with rigorous course schedules and attaining gentlemanly As, Bs, and Cs. In returnâ€”as kickbackâ€”they should now be allowed to â€œlive a little.â€ (Page, 1999, p. 579) These attitudes are not new. Among the major themes in interviews in the late 1970s with 75 seniors, all in the top 15% of their classes in five schools in suburban Chicago, were acknowledged efforts to â€œbea t the systemâ€ and do a minimum amount of work (Linton & Pollack, 1978). Surveys of high school seniors reinforce the picture of low engagement. For example, the Colorado School-to-Career Part nership (1999) conducted an exit survey of almost 9,000 seniors from 132 high schools in th e state. Asked how often they were bored in class, 65% of the respondents selected frequencies at or above â€œhalf of the time.â€ In a study comparing student attitudes towa rds school across three Pennsylvania high schools, one of them featuring a junior-senior curriculum recently re structured to include more interdisciplinary and independent study, Sandberg (1981) collected longitudinal data using semantic differential scales. Tests comparing students â€™ junior and senior responses on these scales showed that, in the non-restructured schools, seniors found school in general, and their co urses in particular, less intere sting, easier, less meaningful, and less useful than they had the previous year. Seniors also described their classmates
7 as enjoying school less, working less diligentl y, and being less mature than they had the previous year. These attitudes about the senior year app ear to have worsened over time. One piece of evidence comes from the Monitoring the Future survey. Sponsored by the Institute for Survey Research at the Univ ersity of Michigan since 1976, the survey includes approximately 15,000 seniors annually. Several major attitudi nal indicators of engagement declined between 1980 and 1999: the proportion of seniors who rated their coursework as â€œoften or alwa ys meaningfulâ€ fell from 42% to 30%; the percentage of seniors indicating that their co urses were â€œquite or very in terestingâ€ declined from 38% to 26%; and the proportion of seni ors saying that they â€œliked school very much or a lotâ€ fell from 46% to 34% over that time peri od (Boesel, 2001). These declines, moreover, have been largely independent of the t ype of high school program (e.g., academic or vocational) (Kadyszewski, 2003). The same downward trend was documented in a study of 250,000 college and university students, statis tically adjusted to be representative of the 1.6 million students entering college as fi rst-time, full-time freshmen in 1997 (Sax, Astin, Korn, & Mahoney, 1997). The fall 1997 survey, conducted by the Higher Education Research Institute at UCLAâ€™s Gra duate School of Educa tion, revealed that a record high of 36 percent of freshmen reporte d being â€œfrequently boredâ€ in class during their high school senior year, compared w ith the all-time low of 26.4 percent in 1985. Accompanying these negative attitudes toward s the high school senior year is a set of behaviors indicative of low academic enga gement. These behaviors were humorously described in a recent New York Times editorial (Herring, 2001):
8 A friend recalls that when he was in hi gh school, one teacher said: â€œWhen I was a kid, if you cut classes, showed up late, and didnâ€™t do any work, they called you a bum. Now they call you a senior.â€ And that was then. (p. B9) Attendance Problems Cross-sectional survey data show signif icant attendance problems in the senior year. In their study of students from over 1,000 high schools, Coleman, Hoffer, and Kilgore (1982) found that seniors reported si gnificantly higher frequencies of skipping class, being absent from school entirely, a nd arriving at school late compared to sophomores. Apparently little has changed in 20 years. Data for the year 2000 from the National Center for Education Statistics (NCES) showed that, whereas 16% of sophomores admitted skipping school, 26% of seniors did. In addition, while only 14% of sophomores indicated that they had been ab sent five or more days within a specified month-long period, the corresponding percen tage for seniors was over 20% (NCES, 2002). Even if they do not skip school entire ly, seniors seem especi ally likely to skip individual classes. In the previously described exit survey of 9,000 seniors in Colorado (Colorado School-to-Career Pa rtnership, 1999), 69% of respondents admitted skipping classes at least â€œ once in a while.â€ These student-reported data are reinforced by information collected from the supervising adults in their lives. For example, a survey of nearly a ll secondary principals in Tennessee found a significant inverse relatio nship between attendance and grade level of the students (Brimm, Forgety, & Sadler, 1978). Educators and administrators who have spent their entire professional lives in schools (e.g., Miller, 1972) have commented on the increase in absences and tardies in the senior year, and it is not uncommon for seniors to ask how many days they can mi ss before disciplinary consequences ensue (Davis, 2006). If these in-school attendance i ssues were offset by greater engagement in
9 school-related work at home, then perhaps the issues would be less im portant. However, no such compensation has been observed. Minimal Time on Homework On the contrary, a series of studies ha s shown that seniors spend less time on homework than other high school students. For example, Natriello and McDill (1986) analyzed data collected in 1964-65 fr om over 12,000 students from 20 high schools around the country. Asked to estimate time sp ent daily on homework (on a 6-point scale ranging from â€œnone or almost none each dayâ€ to or more hours each dayâ€), seniors reported spending significantly le ss time than juniors. These findings are consistent with those of Coleman et al. (1982), whose 1980 da ta suggested that seniors spend less time weekly on homework than sophomores in a ll schools except â€œhighperformanceâ€ public and private schools, defined as those with the highest proportions of seniors qualifying as semifinalists in the National Merit Scholar ship competition. In addition, the 2002 version of the UCLA study that found high rates of boredom among seniors (Sax et al., 1997) showed that a record low 33.4% of entering fi rst-year college student s reported spending six or more hours per week on homework as seniors (cf. a corresponding figure of 47% when the question was first asked in 1987). The proportion of senior s reporting less than one hour per week on that survey nearly doubl ed over that same period (Engle, 2003). Further evidence of such declines emerge d from an analysis comparing National Assessement of Educational Progress (NAEP) data from 1984 to 2004 (Perie & Moran, 2005). Several comparisons provide some telling perspective on these numbers. First, setting data from the 2004 High School Survey of Student Engagement alongside those from its college-level count erpart, the 2004 National Survey of Student Engagement,
10 indicated that first-year co llege students spend approximately twice as many hours per week preparing for their cl asses as high school seniors do (McCarthy & Kuh, 2005). In addition, data from the Third Internationa l Mathematics and Science Study (TIMSS) showed that the daily average of 1.7 hours of studying and homework among U.S. high school seniors was significantly lower than the international mean of 2.6 hours (Miller, 2001; Mullis, 1998). Just as these numbers are little surprise given the attendance issues described earlier, so too is it fairly predicta ble that this minimal investment in homework is joined by a tendency among seniors to lighten their course loads. Course Enrollment Patterns It is not uncommon for stude nts to enroll in fewer rigor ous courses in the senior year, especially in the second semester (Weldy, 1984). In light of evidence to be reviewed later, one of the most worrisome trends is for students to discontinue their course-taking in mathematics. A number of large-sample surveys in the United States have indicated that only about two-thirds of seniors are enro lled in a mathematics course, well below the percentages for other grade levels and for students in most other countries (Kelly, 2001; Mullis, 1998). Even among U.S. seniors on college-p reparatory tracks, nearly 20% do not take a mathematics class (High School Survey of Student Engagement, 2005a). The percentage of se niors enrolling in science courses is comparably low (Kelly, 2001; Miller, 2001). Sequelae of Senior-Year Disengagement Effects on College Performance Recall that senior enrollment in rigorous academic courses, particularly mathematics and science classes, is relatively low. A series of large-sample regression studies sponsored by the U.S. Department of Education have shown the potential danger
11 of this pattern (Adelman, 1999, 2006). Adelmanâ€™ s analysis in both studies showed that the strongest predictor of college-degree co mpletionâ€”stronger even than family income, standardized test scores, and high school gradesâ€”was â€œacademic intensity,â€ measured by the number of Carnegie units earned in co re academic classes in high school. The predictive power of academic intensity was espe cially strong for low-income students. For example, for students in the lowest quint ile of socioeconomic st atus, moving into the top 40% of the academic-intensity index a nd entering postsecondary education directly from high school would improve degree comple tion from 36% to 59%. Microanalyses of the data in both studies also indicated that the highest level of mathematics taken in high school was the strongest curriculum-rela ted predictor of degree completion. Even if students complete their colle ge degrees, many take longer and pay more than necessary for the experience because of th e need for remedial courses. According to NCES data, 20% of full-time, first-year studen ts in public universitie s took at least one remedial course in reading, writing, or ma thematics in 2000. The corresponding figures were significantly lower for students at pr ivate colleges and uni versities (12%) and significantly higher for student s in community colleges (42 %) (Parsand & Lewis, 2003). Education researchers (e.g., Adelman, 1999; Kirst, 2001; National Commission on the High School Senior Year, 2001b), district supe rintendents (e.g., Jester & Ismail, 2006), and university chancellors (e.g., Jayson, 2005) have attributed this need for postsecondary remediation in part to a weak ening of skills and a slowing of academic momentum in the senior year. Remediati on not only costs taxpayersâ€”in a sense doublebilling them for skills that should have b een acquired in secondary education (Jayson, 2005; Kirst, 2001)â€”but it also put s students at greater risk of not graduating from
12 college. This increased risk was indicated in data from the National Educational Longitudinal Study (NELS:88), according to whic h the percentage of students earning a 4-year degree dropped from 76% for students requiring no remedial coursework to 45% for students requiring one or more remedial classes in mathematics (Carey, 2004). Current college students often acknowle dge the pernicious effects of their disengagement in the senior year. In one st udy first-year college st udents were asked to write an â€œadvice letterâ€ to thei r former teachers in high school. Often mentioned was that the transition to college would have been smoother had teachers not indulged seniorsâ€™ disengagement (Sandel, 1990). Another st udy conveying a similar point was based on interviews with approximately 900 students en rolled in twoand f our-year colleges who had graduated from public high schools in 2002, 2003, and 2004 (Peter D. Hart Research Associates, 2005). Asked whet her, knowing what they know today as college students, they would approach high school differently if they had a chance to do it again, 65% said that they would have worked harder. Alt hough this question did not specifically target the senior year, it seems reasonable to infer that studentsâ€™ memories of senior-year work habits were most accessible and therefore provided some basis for their responses. Difficulties making the adjustment to college-level work certainly reflect more than just senior-year disengagement. The cumu lative impact of twelve prior years of education cannot be ignored (National Commission on the High School Senior Year, 2001a). However, evidence from senior-year performance on nationa l and international exams suggests that the senior year is in fact not adding much value. Effects on Standardized Test Performance There is evidence of diminished college -readiness momentum as students progress through school. Results from a recent st udy published by the American College Testing
13 program (ACT, Inc., 2006) provide one example. Of all the seniors taking the ACT in 2004-05, only 51% met ACTâ€™s â€œcollege readines s benchmarkâ€ for readingâ€”a level of achievement predicting a high probability of success (75% chance of earning a B or better) in first-year college courses. In c ontrast, in a combined population of four recent cohorts who took â€œpreparatoryâ€ versions of the ACT in grades 8 and 10 followed by the full ACT in grade 12, 62% of eighth-graders scored high enough (adjusted for expected growth) to be considered on-track for colle ge reading. The percentage reaching the benchmark was slightly higher when these stude nts reached grade 10. However, as in the reference population of seniors, the seniors in these special cohorts were less likely to reach the benchmark in their last year of high school. A different perspective on the issue of adde d value in the senior year comes from recent reports of trends in NAEP scores, wh ich are based on representative samples of over 10,000 U.S. students at each of three grade levels (4, 8, and 12). While students in grades 4 and 8 have posted significant gains in reading and mathematics scores since the tests were first administered in the early 1970s, the scores for stude nts in grade 12 have remained flat (Perie & Moran, 2005). The pi cture for NAEP writing scores is similar (Manzo, 2003). Concerned about these re sults, the National A ssessment Governing Board, the group of policymakers, educators, and citizens charged by the U.S. Congress to set NAEP policy, established the National Commission on NAEP 12th Grade Assessment and Reporting. Among the Commis sionâ€™s observations were that seniors have extremely low participation rates on th e exam and that seniors who do sit for the tests tend to omit a large number of questions , especially those requiring greater effort (e.g., essays). The Commission discussed incentives, such as awarding college
14 scholarships to randomly-selected seniors who answered all questions and earned scores above some threshold. The Commission even hired a marketing firm to develop more ideas for enhancing motivation among se niors (Cavanagh, 2004; National Assessment Governing Board, 2004; Nati onal Commission on NAEP 12th-Grade Assessment & Reporting, 2004). Senior test-takers fare little better wh en compared with their peers in other countries. The TIMSS study, conducted in 1995, was an internati onal comparison of knowledge and problem-solving skill in mathema tics and science. Whereas U.S. students scored at or above the intern ational average in grades 4 and 8, American seniors scored near the bottom (Michelau, 2002). Bracey (2000, 2004) articulated some important criticisms of these comparisons, among th em disparate partic ipation rates across countries, age discrepancies (with students in comparison countries generally being older than U.S. seniors), and greater probability of non-uniform curricula in the decentralized U.S. educational system. However, Bracey did not completely dismiss the unflattering comparisons as a statistical mirage; indeed, he argued that motivational differences across countriesâ€”related to features of the senior year such as widespread employmentâ€”were at least partly responsible. While the verdict that the senior year is an â€œeducational wastelandâ€ (as one university chancellor put it, in Jayson, 2005) might be too strong, the conclusion that the last year of high school is a time of ra mpant academic disengagement and relatively limited educational value-added is defensible for the reasons noted above. Furthermore, even if the rhetoric of the National Comm ission on the High School Yearâ€™s (2001a) callto-action seems overblown, there is little doub t that educators and policymakers would do
15 well to heed the following counsel: â€œThough th e seniorsâ€™ minds are elsewhere, their bodies are still in high school, and work must be found for them to do thereâ€ (Sizer, 2002, p. xvii). Potential Remedies for Senioritis One approach to the problem of senior-year disengagement emphasizes the â€œstickâ€ rather than the â€œcarrot,â€ with high standard s of academic and behavioral accountability and consequences for failure to meet those standards. This approach was forcefully articulated by Weldy (1984), a high school s uperintendent who called for educators to â€œtake charge of our studentsâ€™ education and require their pr esence, their a ttention, their cooperation, and their best performanceâ€ ( p. 91). Weldy derided â€œmore exotic and tasteful palliatives . . . for the problem of senioritisâ€ (p. 93), such as electives and internships, advocating instead a return to more traditi onal requirements. Among the recommendations that he and ot hers have advanced are more course requirements in the senior year (e.g., required mathematics; Mich elau, 2002); stricter gr ading practices (e.g., passing every component of a course rather than needing only a passing average ; Whitacre, 1984); and revocation of privileges (e.g., attendance at prom and graduation) for seniors with excessive absences, tardies, and other disciplina ry issues (Lord, 2001). By themselves, stringent policies of this kind are insufficient, perhaps masking symptoms but not actually addressing many of the underlying factor s causing senioritis (Haffey, 1995; Karweit, 1973). Haffey thus studied an incentive program, developed at a suburban Texas high school in the early 1990s, that allowed seniors to be exempt from spring final exams if their prior attendan ce, conduct, and achievement met certain standards. Comparing two graduating classes before the program was instituted with two classes after the program went into effect, Haffey found that the program was effective in
16 reducing absences and disciplinary problems. Unfortunately, incentive programs like this one may convey the message that what happens in school in terms of engaged and highly relevant learning is less important than â€œserving timeâ€ in an orderly way (Sedlak, Wheeler, Pullin, & Cusick, 1986) . Greater potential for d ealing with the underlying malaise of senioritis inheres in th e approach described by Kelly (2001): As educators work to strengthen education fo r seniors, they need to remember that students are going through a time of profound transition, says Frank Sachs, senior program coordinator at the private Blake School in Minneapolis. For that reason, senior year needs to emphasize the opport unity for individual decisionmaking and exploration as much as it doe s accountability. (p. 4) Similarly, the National Commission on the Hi gh School Senior Year (2001b) called for â€œmoving away from a system in which the senior year is just more of the same to one in which it provides the time to explore options and prove knowledge a nd skillsâ€ (p. 22). Although such novel opportuni ties for individual decisi onmaking and exploration are still more the exception than the rule (Conley, 2001; Dunn, 2001; Viadero, 2001), alternative senior-year programs are becoming more common in American high schools. Among the extant programs are the following: cour ses or units that use art, literature, and other disciplines to explore transitions a nd the challenges of independence (Kessler, 1999/2000; Sizer, 2002; Thompson, 1998); leadersh ip opportunities such as planning courses and mentoring younger students (Dre is & Rehage, 2006; Fazio & Ural, 1995; Lott, 1995); community-service experien ces (Arms, 1980; DeWitt & Joyce, 2001; Rosenberg, 1997); and dual-enrollment programs in which seniors spend part or all of their time taking college courses (Andrews, 2004; Bonesteel & Spe rry, 2002; Peterson, 2003). Another opportunity widely recomm ended by educational reformers and policymakers is the completion of a cap stone project (e.g., Ma rsh & Codding, 1998;
17 National Commission on the High School Senior Year, 2001b). Such projects, intended to demonstrate hands-on mastery of academic skills and dispositions, are generally driven by a central question or problem chosen by the student and culminate in some kind of useful product (Blumenfeld, Soloway, Mar x, Krajcik, Guzdial, & Palincsar, 1991; Good & Brophy, 1997). Capstone work of this kind has received special emphasis since the founding in 1984 of the Coalition of Essential Schools, one of whose core principles is â€œgraduation by exhibitionâ€ (Sizer , 1992). For example, culminating projects have been part of high school restructuring plans across the country (e.g., Darling-Hammond, Ancess, & Falk, 1995; Fischetti, Dittmer, & Hohmann, 1993). They also have become part of the high school exit standards for at least five states: Hawaii, North Carolina, Pennsylvania, Rhode Island, and Washi ngton (Hood & Egelson, 2005; McLain, 2002; Wills, 2005). For these reasons, the projects ha ve received extensive attention both in the popular press (Archer, 2005; DeFao, 2005; Mathews, 1998; Tomsho, 2005; Vaznis, 2005; Wills, 2005) and in the descriptive educati on literature (Chadwell, 1991; Kemmery & Cook, 2002; Lorenz, 1999; Nicolini, 1999; Shaunessy, 2004; Sills-Briegel, Fisk, & Dunlop, 1996/1997; Summers, 1989). Not all capstone projects are created e qual, however. For example, in 2001 the Washington State Institute for Public Policy su rveyed principals from nearly all public high schools in the state (McLain, 2002). Fift y-one percent of the schools were requiring a culminating project, and 25% were in the pl anning phases of adding such projects to the curriculum. Among the schools already implem enting a capstone prog ram, the depth of the required projects varied significantly. In some school s, for instance, a research paper in the senior English course was consider ed a culminating project; in others, the
18 requirements were more extensive, includi ng oral presentations and exhibitions of physical products related to the research topi cs. These findings are consistent with the verdict of Dr. Richard Basom, the managing director of the Part nership for Dynamic Learning, Inc., which provides s upport services for senior projects in more than 1,000 high schools nationwide: â€œImplementation va ries widelyâ€”maybe 80% of the programs out there are laxâ€ (Wills, 2005, p. D1). The focu s of this dissertation is the other 20% of programs. Senior Project To understand what Senior Project is, it is he lpful to know its forebears. One of the earliest programs serving as its model wa s the Woodlands Individualized Senior Experience (WISE) program (DeFao, 2005; Hoover, 2003; National Commission on the High School Senior Year, 2001b; Wade, 1999). Established in 1973 at Woodlands High School in New York, WISE was intended to alleviate senioritisâ€”i n the words of the programâ€™s founder, to create an environment in which students were â€œcelebrating senior year instead of leaving with a whimperâ€ (Wade, 1999, p. 764). Now implemented in over 60 schools across the countr y, the WISE model allows se niors to pursue individual learning projects during the s econd semester. These projects can range from internships to intensive academic research to performan ce-based experiences. Regardless of the specific form of the project, however, all seniors in the WISE program present and defend their work to a panel comprising faculty, younger studen ts, and community members. About a decade later, on the opposite coast of the U.S., a group of teachers at South Medford High School in Oregon was equally distressed by senioritis:
19 We thought about the seniors, those placi d, horizontal creatu res who maintain a hazy attention until the homecoming queen is crowned or MacBeth dies, whichever comes first, at which point they pass in to the oblivion known as senioritis. We diagnosed their comatose stat e as evidence of an inadequa te curriculum. (Chadwell, 1991, p. 36) Like WISE, the model developed at South Me dford included a presentation to a group of judges. However, it also clearly stipulated two additional requirement s: a research paper and a practical applica tion or hands-on experien ce related to the topi c. In the late 1980s the programâ€™s founders formed a consulti ng group, called FarWest EDGE (Energetic Designs for Growth in Education), to provi de workshops and support materials for other schools wishing to start or rede sign senior-year projects. FarW est sold its Senior Project trademark and property to SERVE (SouthEas tern Regional Vision for Education). SERVE is a federally-funded research consortium , affiliated with the University of North Carolina-Greensboro, that conducts research and provides support for schools in the southeastern U.S. In early 2005, after admini stering Senior Project support to schools for a few years, SERVE divested its Senior Project operations to an independent, not-forprofit spin-off, the Partnership for Dynamic Learning, Inc. (R. Basom, personal communication, April 25, 2006). The hundreds of schools across the country th at are formally affiliated with the Partnership for Dynamic Learning, as well as hundreds of others not formally affiliated but implementing full-scale Senior Projects, certainly have variati ons in their program designs. For example, some schools restrict th e project to only one semester (Chadwell, 1991), while others stretch the project over mo st of the academic year (Lippard, 2000). However, what all of these Senior Projects have in common is the three components of a research paper, a practical or creative appl ication, and a formal presentation. The kinds of topics and corresponding appl ications that students develo p are well illustrated by the
20 following examples: a research paper on the ba llets of George Bala nchine is followed by the performance of an original choreogr aphed ballet (Chadwe ll, 1991); a paper on childhood obesity is the basis for compiling a cookbook for children and teaching a nutrition lesson to an elementary class (Sha unessy, 2004); and research on the principles of feng shui is followed by shadowing an interior designer, creating a video record of several buildings visited, and analyzing the interiors using feng shui concepts. Regardless of the exact topic chosen by the student, the Senior Project paradigm is designed to provide rigorous ski ll-training in the senior year and address some of the factors that seem to precipitate senioritis. The rigor is related to precisely those sk ill areas of greatest concern to employers and postsecondary faculty, including writing, speaking, research skills, and time management (Conley, 2003; Cushman, 1994; Da rling-Hammond et al ., 1995; Oâ€™Grady, 1999; The Secretaryâ€™s Commission on Achiev ing Necessary Skills, 1991). In a study cited earlier (Peter D. Hart Research Associates, 2005), 1,500 recent graduates of public high schools participated in structured intervie ws. About two-thirds of the interviewees were in college, and the remainder had ente red the workforce. Forty percent of the college students and 45% of the workers acknowledged gaps in their ability to find and synthesize information. Approximately 45% of both groups acknow ledged weaknesses in their preparation for oral communicatio n/public speaking, with slightly smaller percentages admitting gaps in their writing ability. College students who said that they were expected to write a lot in high school, including longer papers such as research reports, were far more likely to feel somewhat or very well prepared for college writing. This study also included interviews with 400 employers and 300 colle ge instructors (in
21 twoand four-year institutions). Thirty-nine percent of the em ployers said that they were dissatisfied with graduatesâ€™ public-speaking skills, 62% of the college instructors were dissatisfied with graduatesâ€™ writing skills , and 59% of the college teachers were dissatisfied with the research skills of incoming students. That Senior Project addresses these concerns not just on pa per, but also in practice, is supported by a small but growing number of studies. For example, Egelson, Harman, and Bond (2002) reported a survey study conducted at 16 established Senior Project sites. Of the 1,800 senior respondents, 75% agreed or strongly agreed that their writing, research, speaking, and time-management skills had improved because of their participation in Senior Project. Graduates of these schools also were surveyed. These students were significantly more likely th an a similar number of graduates from comparable schools (matched on size, location, SES, and average performance in a state testing program) to indicate that high school had taught them the skills needed to write a research paper and to prepare and present a speech. Additional data on the positive effects of Senior Project emerged from a small experimental intervention trial reported by Lopez (2004). Two South Carolina schools agreed to pilot-test a Se nior Project program, and th e researcher found two schools matched on location to comprise the cont rol group. Among othe r instruments, a standardized research-skills test was administ ered to seniors in all four schools as a pretest in the first month of classes and as a posttest in th e last month. After controlling for individual studentsâ€™ gende r, race, parental education, intellectual confidence, and pretest scores, students in the Senior Project schools showed greater improvement in their research-skills test scores than did students in the control schools.
22 These advances in skills are attributab leâ€”according to the theory behind Senior Project (e.g., Tsuzuki, 1995)â€”in large part to the projectâ€™s abi lity to capitalize on intrinsic motivation. Allowing students to choos e the content of their study, rare in most formal educational settings, should activ ate that motivation (e.g., Brophy, 1987b; Good & Brophy, 1997). Brophy expresse d the point well: The intrinsic motivation approach calls for teachers to select or design academic activities that students will engage in w illingly because these activities incorporate content that the students are already inte rested in. . . . However, teachersâ€™ opportunities to capitaliz e on studentsâ€™ existing intrinsic motivation are limited by several features inherent to the nature of schooling, [am ong them the fact that] the curriculum is prescribed ex ternally rather than chos en by the student. (p. 220) Indeed, the kind of independent study represented by Senior Project is something that many high school students say they want. To wit, when asked to comment on a series of ideas to make the senior year â€œmore mean ingful,â€ 78% of the 10,000 respondents to the online Rate Your Future survey (National G overnors Association, 2005 ) said that â€œtaking independent study with a favorite teacherâ€ would work somewhat or very well. The in-theory popularity of independent st udy would seem to be contradicted by the publicized negative reacti ons in schools and di stricts when Senior Project is first implemented. For instance, the initial studen t response to Senior Projects at Wakefield High School near Washington, D.C., was a peti tion suggesting that the school cancel the program (Mathews, 1998). In fall 2004 student s in Jerome, Idaho, staged a walkout when Senior Project became a graduation requirement (Tomsho, 2005). Students at a school in Springfield, Missouri, that had just implemented Senior Proj ect ordered protest t-shirts reading â€œClass of 1995: Guinea Pigs for the Futureâ€ (Sills-Briegel et al., 1996/1997). Students at schools described by Darling-Hammond et al. (1995) and Lorenz (1999) reacted similarly:
23 What was the studentsâ€™ reaction to this opportunity to direct their own learning, explore new fields of knoweldge, raise thei r cognition to heights before unknown? They mutinied. They wept. They raged. They trembled. They murmured. They pleaded. They threatened to transfer. . . . I was genuinely surp rised at the fury of their response, [especially among the] group of high achievers . (Lorenz, 1999, p. 80) These initial reactions, however, often yi eld to engagement among Senior Project participants. Some evidence comes from rela tively informal, single-school evaluations. For example, in dissertation research conduc ted at one school that recently implemented Senior Project, Winters (2000) interviewe d English teachers. One theme in the interviews was that students complained abou t the work but nevertheless worked harder than previous senior classes that did not ha ve Senior Project. One teacher expressed the point as follows: â€œI had heard that Senior Pr oject helped control se nioritis, but I really didnâ€™t believe it until that first yearâ€ (p. 226) . Focusing on other signs of engagement, faculty at a restructured vocational-techni cal school described by Darling-Hammond et al. (1995) and Godowsky, Scarbrough, and St einwedel (1992) reported better attendance and greater homework completion among senior s once Senior Project was instituted. The principal at North Providen ce High School in Rhode Island was interviewed in 2005 and noted that the drop-out rate among seniors ha d dropped significantly since Senior Project was started in 2000 (Archer, 2005). Such testimonials from researchers and faculty are joined by comments from students. In fact, laudatory quotations from seniors are co mmon in stories about Senior Project that appear in the popular and education press. Collectively, at least, these comments are suggestive: By this time, none of my classes are rea lly interesting anymore. But the senior project is still fun to work on, especially if youâ€™re doing something creative that you like. (Student at Hamilton High School Humanities Magnet, in Kirsch, 1997, p. 15)
24 Throughout my project I kept thinking to myself that I didnâ€™t understand why we had to do this and that it was a waste of my time. I am happy to say that I was seriously wrong. I have learned so many thi ngs, not just about my topic but of the real world. (Student at Franklin Road Christian School, in Lorenz, 1999, p. 84) The most frustrating thing was how impractic al slacking off became. In fact, I had to work harder, and with more dedica tion and serious thought, those final weeks than I did all school year, because the presur e was really on. (Student at Francis W. Parker Charter Essential Sc hool, in Lippard, 2000, p. 14) More systematic empirical research dem onstrating that Senior Project enhances student engagement was reporte d by Riedel (2001) in his di ssertation. Riedel used surveys and interviews to study two cohorts of students, both at Delaware high schools that were restructuring accord ing to Coalition of Essentia l School Common Principles (Sizer, 1992). As indicated earlier, â€œgraduati on by exhibitionâ€ is one of those principles, and both schools therefore had recently starte d Senior Project program s. Riedel followed the students longitudinally over four years of high school. Analysis of the data indicated a decrease in days skipped and an increase in GPA in the senior year. Semi-structured interviews shed more light on these results.. Most of the seniors at tributed their greater engagement in school to Seni or Project, along with increa sed expectations and caring from their teachers. One remark, from a st udent who had claimed ear ly in the interview that he did not care much about his educati on, illustrates the point well: â€œI was more engaged because of senior pr oject. I worked more and harder. I was done before everyone else. I didnâ€™t put it offâ€ (pp. 182-183). Rationale for the Study The preceding description of Senior Projec t suggests that the program has great potential to keep students e ngaged and enhance their postsecondary readiness. However, translating this potential into reality is not ea sy, as Blumenfeld et al. (1991) pointed out:
25 Unfortunately, evidence indicates that st udents do not necessarily respond to highlevel tasks with increased use of learni ng strategies (C.W. Anderson & Roth, 1989; Blumenfeld & Meece, 1988; Winne & Marx, 1982). Students often are resistant to tasks that involve hi gh-level cognitive processing and try to simplify the demands of the situation through negotiation (D oyle, 1983; Starke & Easley, 1978). Although students may be interested in th e topic and possess relevant knowledge and procedures for solving problems or mastering new material, they do not necessarily use these strategies (Par is, Lipson, & Wixson, 1983; Winne & Marx, 1982). It is insufficient merely to provi de students with oppo rtunities designed to promote knowledge that is integrated, dynami c, and generative, if students will not invest the effort necessary. (pp. 374-375) Identifying what is likely to promote this inve stment of effort is the goal of the literature review. The original study that follows th e review is meant to assess whether these ingredients truly are â€œactiveâ€ in determining how engaged students are in their Senior Project work. Answering these questions has both practic al and theoretical significance. In practical terms, academic disengagement in the senior year is costly both educationally and financially. More and more high schools are adopting Senior Project as part of a solution to this problem. Thus, any inform ation that might help school personnel design engaging Senior Project program s would be useful. Theoreti cally, this study is grounded in models of academic motivation and achieve ment that have been described in the educational literature. As such, it provides a te st of some of the major constructs in these theories in the novel context of Senior Projec t. Such a test has potential to provide further support for these theo ries, illuminate possible boundary conditions for them, and provide new directions for general acc ounts of academic motivation and school achievement. Prospectus In Chapter 2 I present theo retical and empirical support for the model I developed to explain engagement in Senior Project work. I begin that chapter by introducing a
26 diagram that represents the model and then briefly explaining each component in the diagram, as well as the general nature of th e interconnections between these components. The remainder of the chapter specifically defe nds each path in the model. In Chapter 3 I describe the methods used to test the model of Senior Project enga gement, as well as the limitations of these methods. Chapter 4 contai ns the results of the analyses, and Chapter 5 is a discussion of the theoretical and practic al implications of th e study and directions for future research.
27 CHAPTER 2 REVIEW OF LITERATURE The goal of this study is to understand what promotes student engagement in Senior Project work. An important first step towa rds this goal is defining engagement more explicitly rather than depending on the va gue meaning assumed thus far. Once a definition is in place, the following question arises: Why focus on engagement rather than some other outcome? The answer that engageme nt is the key to ensuring that seniors will achieve more and be better prepared for postsecondary life must be supported with evidence. Following the presentation of this evidence is a review of the literature on student motivation and engagement. The aim of this review is to develop a structured list of ingredients hypothesized to promote enga gement in Senior Project and any other proposed remedy for senioritis. Engagement: The Aim of the Remedy Definition of Academic Engagement So it is no news that many students put in li ttle effort. Some actively resist school learning and drop out either physically or mentally. Others go through the motions to receive high school diplomas or college acceptances but display bloodless, passive, and cynical minimalism toward their studies. (Powell, 1996, p. 21) Powellâ€™s description is the antithesis of the academic engagement that educators wish to see in their students. The co ncept of academic engagement has received extensive theoretical and empirical atte ntion from educational researchers and policymakers in the last two decades (Fredr icks, Blumenfeld, & Paris, 2004; National Research Council & Institute of Medicine , 2004). Although definitions of academic
28 engagement abound (e.g., Finn, 1989; Newmann, Wehlage, & Lamborn, 1992; Schaufeli, Martnez, Pinto, Salanova, & Bakker, 2002; Stevenson, 1990), the spirit of most is captured well in the following gene ral description of engagement: Engagement refers to active, goal-directe d, flexible, constructi ve, persistent, and focused interactions with the social a nd physical environments. In contrast, patterns of disaffection, in which individua ls are alienated, apathetic, rebellious, frightened, or burned out, turn people aw ay from opportunities fo r learning. (Furrer & Skinner, 2003, p. 149) Signs that a student is engaged in academic work generally fall into one of three categories: behavioral, emoti onal, or cognitive (Fredricks et al., 2004). Among the oftenused behavioral indicators of engagement are preparati on for the task, time on task, proactive initiation of activity relevant to the task, and persistence in the face of challenges (Connell & Wellborn, 1991; Geocar is, 1996/1997; Skinner & Belmont, 1993). Countervailing behaviors, such as tardiness for or absence from the situation, often are used as negative measures of behavioral engagement (e.g., Lee & Smith, 1995). As Dewey (1904/1965) pointed out over a century a go, students can offer â€œouter attentionâ€ to a task while withholding their â€œinner attentionâ€ (see also Goffman, 1967). For evidence of such â€œinner atten tion,â€ researchers look for signs of emotional and cognitive investment. Emotional indicators of engage ment include enjoymen t and pride in oneâ€™s work, optimism, and curiosity (Connell & Wellborn, 1991; Csikszentmihalyi, 1997; Newmann et al., 1992). Emotional engagement also is inferred from the frequency of negative emotional states, such as boredom (Kanevsky & Keighley, 2003; Marks, 1995). Ultimately, the quality of the learning that results depends on the degree of cognitive engagement (National Research Council & In stitute of Medicine, 2004; Shulman, 1986; Yair, 2000). Researchers have assessed cognitive engageme nt by asking about studentsâ€™ level of concentration and attention, as well as their use of cognitiv e, metacognitive, and
29 self-regulatory strategies during learning (Pintrich & DeGroot, 1990; Shernoff, Csikszentmihalyi, Schneider, & Shernoff, 2003). Educational Value of Academic Engagement The goal of increasing student engagement in academic work has emerged as a major target of educational reform, alongsideâ€” in fact, usually cons idered prerequisite toâ€”higher achievement (Epstein & McPa rtland, 1976; Lee & Smith, 1995; McLain, 2002; Murphy, 1991; Newmann et al., 1992; Wehlage, Rutter, Smith, Lesko, & Fernandez, 1989). A major reason that engage ment is seen as a productive target for research and intervention is that researcher s view the behaviors, emotions, and cognitions described in the previous section as pr oximal and manipulable (Connell, Spencer, & Aber, 1994; Finn & Rock, 1997; Finn & Voelkl, 1993). Indeed, this belief in the utility and malleability of engagement spurred th e recent development of the national High School Survey of Student Engagement (HSSSE ) by researchers at I ndiana University. Administered to over 80,000 students across the U.S. each year since 2004, this survey is predicated on the idea that standardized test results alone cannot identify issues with students, teachers, classrooms , and schools that might be a ddressed in order to improve student attitudes, behaviors, and learning (McCarthy & Kuh, 2005). Engagement thus has assumed a central place in theories of academic achievement, being used to explain the connection between achievement and such factors as school size, teacher support, and student self-efficac y. In educational psychology, for example, Skinner, Connell, and their co lleagues have developed theori es of achievement in which teacher and classroom features affect a studentâ€™s feelings of competence, autonomy, and belonging; these feelings promote behavioral and emotional engagement, which in turn support achievement (e.g., Connell, 1990; Skin ner, 1995; Skinner, Wellborn, & Connell,
30 1990). Focusing on cognitive engagement, Pint rich and colleagues (e.g., Pintrich & De Groot, 1990) have defended a model in wh ich greater cognitive and self-regulatory engagement (e.g., monitoring oneâ€™s understa nding) mediates th e strong relationship between task performance and psychological inputs such as se lf-efficacy. Similar models interpolating engagement between classr oom-context variables (e.g., instructional techniques and quality of pe rsonal relationships in the cl ass) and achievement outcomes are found in diverse educa tional domains, including liter acy research (Guthrie & Wigfield, 2000) and school reform (New mann et al., 1992; Wehlage et al., 1989). These engagement-centered models of academic achievement are grounded in, and in turn provide impetus for, a range of empi rical studies linking the behaviors, emotions, and cognitions of engagement with positive st udent outcomes. The association between academic engagement and school achievement (measured both by grades and achievement tests) has been amply demonstr ated in studies with elementary students (Finn & Cox, 1992; Finn, Pannozzo, & Voel kl, 1995; Furrer & Skinner, 2003) and middle school students (Finn, 1993). Indeed, there is some evidence that engagement is a stronger predictor of school grades for th e older students (Skinner, Zimmer-Gembeck, & Connell, 1998). Studies with high school stud ents tell a similar st ory. For example, Miller, Greene, Montalvo, Ravindran, and Ni chols (1996) asked 300 hi gh school students about their self-regulation, use of cognitive strategies, persistence, and effort in their mathematics classes. Scores on the self-report measures of effort, persistence, and selfregulation together explained 24% of the variance in studentsâ€™ semester grades in the class.
31 The association between engagement and ach ievement in high school appears to be especially strong for students traditionally labeled â€œat risk.â€ Whereas students from advantaged backgrounds learn less than th ey could when they become disengaged, students in high-poverty schools are unlikely even to finish high school when they become disengaged (Connell, Spencer, & Aber, 1994; National Research Council & Institute of Medicine, 2004). Support for th e special importance of engagement for lowincome students was provided in a study by Finn and Rock (1997). They analyzed longitudinal data from a national sample of 1,800 low-income students in grades 8, 10, and 12. Even after controlling for characteri stics such as self-esteem and locus of control, engagement measures collected from teachers (e.g., reports of student effort and attentiveness) and from student s themselves (e.g., reports of da ily preparation for classes) reliably distinguished between dropouts and grad uates, as well as between graduates who were academically successful and those who were unsuccessful. A much smaller study along these lines (L affey, 1982) considered engagement in social studies class for a sample of 88 Af rican-American sophomores in an inner-city high school. The outcome of interest was pe rformance on a researcher-designed test of the material covered during the course of the study. The predictors were various measures of engagement, including teacher rati ngs of class participation, number of days absent from the class, proportion of class assignments completed, and student-reported feelings of involvement when cued by a ra ndom signal during class. In a stepwise regression these four indicators were highly significant predictors of test score, collectively accounting for one-t hird of its variance and in dividually explaining more variance than a measure of reading aptitude.
32 Promoting higher achievement is a signi ficant reason for paying attention to academic engagement, but it is not the sole r eason. Current engagement also predicts long-term continuing motivation, which Maehr ( 1976) defined as â€œthe tendency to return to and continue working on tasks away from th e instructional context in which they were initially confrontedâ€ (p. 443). For example, Shernoff and Hoogstra (2001) investigated whether high school seniorsâ€™ engagement in science classe s predicted their choice of a college science major two years later. Self -report engagement measures of interest, enjoyment, and concentration were collected at randomly-cued times during class over the course of a week. Logistic regression controlling for both de mographic factors (e.g., gender, ethnicity, family type) and high sc hool science grades showed that these engagement measures were significant predic tors of the choice of a college science major. An additional benefit of engagement is that it can elicit reciprocal reactions from teachers, thereby raising the quality of instruction (Newmann, 1998; Skinner & Belmont, 1993). Early and late in an academic year, Skinner and Belmont collected student and teacher data on behavioral and emotional enga gement in class at the elementary school level. They also collected data on â€œc ontextual support,â€ including the amount of structure and degree of interpersonal warmth offered by the teacher. Using a time-lagged path analysis, Skinner and Belmont found th at children who were more behaviorally engaged early in the year received more c ontextual support from teachers by the end of the year. Given the educational benefits of greater engagement, the low levels documented in the senior year are cause for concern. The disengagement associated with senioritis is
33 part of a downward motivationa l trend that typically starts before the last year of high school. Briefly examining this negative trend provides further insight into what a remedy might look like. Academic Engagement on the Decline Children often come to school eager to lear n but, as this report suggests, many lose their academic motivation as they move through elementary school into high school. In fact, by the time many student s enter high school, disengagement from course work and serious study is common. (National Research Council & Institute of Medicine, 2004, p. ix) Academic engagement drops especially precipitously between elementary school and middle or junior high school (see, e.g., Eccles et al., 1993; McDermott, Mordell, & Stolzfus, 2001). However, the downward trend continues well past this transition. For example, Epstein and McPartland (1976) surv eyed over 4,000 students, once when each student was in grade 4, 5, 6, 8, or 11 and th en a second time the following year. Among the scales on the survey was one measuring interest in academic work (e.g., â€œIn class, I often count the minutes until it endsâ€) and anot her measuring satisfaction with school as a social environment. Ratings of social satisfaction were fairly steady both crosssectionally and longitudinally, but the measure of academic engagement generally declined from one grade to the next. A dow nward trend in engagement from middle to high school also was found in two large crosssectional studies (Mar ks, 2000; Yair, 2000) that used student-reported effort and lack of boredom as measures of engagement. Tellingly, this trend was attenuated after accounting for factors including the perceived relevance of instruction and th e amount of parental involvement in educational matters. Low levels of engagement in high school have been the t opic of much research and commentary. Large-sample observational a nd survey studies of high schools around the nation in the 1980s (Goodlad, 1984; Powell, Farr ar, & Cohen, 1985; Sedlak et al., 1986;
34 Sizer, 1984) were fairly uniform in their critic al portrayal of engage ment in high school. The authors of a three-year study of nine â€œaverageâ€ suburban hi gh schools (Steinberg, Brown, & Dornbusch, 1996) cap tured the zeitgeist well: True, most students report th at they attend classes regul arlyâ€”only about 10 percent cut classes routinelyâ€”and well over 80 percen t say that they would stay in school even if they were able to secure a good full-time job. But at the same time, it is clear that when they are in school, a large proportion of students are physically present but psychologically ab sent. According to their own reports, between onethird and 40 percent of student s say that when they are in class, they are neither trying very hard nor payi ng attention. (p. 67) That academic engagement in high schools is generally low is reinforced by the results of the 2004 High School Survey of Student E ngagement, completed by over 90,000 students across the country. Only 48% of the high sc hool respondents agreed or agreed strongly that their school work made them curious to learn more, and only 30% agreed or agreed strongly that they were excited about their classes (McCarthy & Kuh, 2005). These negative indicators of engagement a ppear to worsen over the course of high school. In an analysis of cross-sectiona l data from sophomores and seniors attending over 1,000 public, Catholic, and independent high schools, Coleman et al. (1982) found that seniors were less interested in school and expressed less enthusiasm for hard work than the sophomores. A more compelling st udy showing such decline was based on data collected longitudinally from nearly 700 high school students for three consecutive years (Crosnoe, 2001). Latent growth modeling reve aled sharp declines over time in studentsâ€™ self-reported academic involvement (sample items: â€œI feel satisfied with school because I am learning a lotâ€ and â€œMost of my classe s are boringâ€). Perhaps surprisingly, this downward trend was driven primarily by att itude change among students in the more advanced academic tracks.
35 Promoting Engagement: An Overview of Ingredients What factors help to explain the declin e in engagement? The answer to this question lies in the literature on what st udents find engaging. The literature in educational psychology and school reform offers a rich substrate of theoretical models and empirical evidence related to this issue. It is from this literature that I derived a list of potential active ingredient s for any remedy targeting motivational problems in education. Additional support fo r these ingredients comes from the descriptive, mostly atheoretical literature on the senior year in particular and high school in general. Definitions of Major Variables A list of these engagement-promoting ingredients, with a brief definition of each, appears below: Advisor Support : Level of encouragement and in strumental support for project work provided by the seniorâ€™s project advisor. Peer Support : Perceived level of engagement in Senior Project among oneâ€™s friends and peers. Parent Support : Level of encouragement and in strumental support for project work provided by the seni orâ€™s adult guardian(s). Clarity of Expectations : Extent to which the project deadlines and other expectations are clear to the student. Autonomy : Degree to which the student percei ves freedom of choice in pursuing the project work. Utility Value : Extent to which the senior views Se nior Project as useful to his or her real-world concerns and aspirations. Sensitivity : Degree to which the student perceive s deadlines and other expectations as fair and appropriate in light of other demands on time. Novelty : Extent to which Senior Project work feels different from the studentâ€™s previous academic experiences.
36 Senior Project Self-Efficacy : How confident the student feels in his or her ability to complete the project successfully. Hypotheses Although it is customary to present hypothe ses after the review of literature, I reverse this sequence as an organizational ai d. Because of the variety in the variables identified in the previous list, the review of literature that follows is lengthy and broadranging. The schema provided here is m eant as an advance organizer for this information. Figure 2-1 summarizes how th e ingredients previ ously identified are believed to fit together to explain a st udentâ€™s engagement in Senior Project. Senior Project Engagement Senior Project Self-Efficacy Advisor Support Peer Support Parent Support Clarity of Expectations Autonomy Utility Value Sensitivity Novelty Previous Mastery Experience Academic Engagement Figure 2-1.Model representing the influences on student engagement in Senior Project. The set of hypotheses represented by this diagram, in which each arrow denotes a positive relationship, can be described as follows: Advisor Support contributes to a studentâ€™ s engagement in Senior Project both directly1 and indirectly via its effect on Senior Project Self-Efficacy. Peer Support has a direct effect on Senior Project Engagement. 1 In this and all subsequent hypotheses, a â€œdirectâ€ effect is one in which an intervening psychological process or state is not explicitly modeled. For exampl e, the â€œdirectâ€ effect of Advisor Support on Senior Project Engagement may be mediated by a sense of quid pro quo ; that is, students may work harder because they believe that they owe it to an advisor who clearl y is working hard for them.
37 Parent Support contributes to Senior Project Engagement both directly and indirectly via its influence on Senior Project Self-Efficacy. Clarity of Expectations has a direct effect on Senior Project Engagement and an indirect effect via Seni or Project Self-Efficacy. Autonomy contributes to Se nior Project Engagement bot h directly and indirectly via Senior Project Self-Efficacy. Utility Value has a direct effect on Senior Project Engagement. Sensitivity contributes directly to Senior Project Engagement. Novelty has a direct effect on Senior Project Engagement. Additional variables in Fi gure 2-1 require explanati on. Academic Engagement refers to a studentâ€™s level of behavioral, emotional, and c ognitive investment in senioryear classes. Previous Mast ery Experience refers to a st udentâ€™s prior level of success with project-related skills. These variables are theoretically and spatially separated from the others because Senior Proj ect coordinators can do little to control them. Nonetheless, the logical and empirical support for the im portance of these vari ables in explaining engagement is strong enough that they shoul d be included in any model that does not wish to overstate the effects of the more immediately manipulable variables. The remaining paths in Figure 2-1 now can be explained: Students who are more generally engaged in school (Academic Engagement) will tend to have greater Senior Project Engagement. Studentsâ€™ Senior Project Se lf-Efficacy depends on the level of previous success they experienced with project-related skills (Previous Mastery Experience). General Conceptual Framework How to stimulate engagement is the firs t question every good teacher asks. The answers? Get the incentives right. . . . Virtually all of learning comes down to incentives. I learn what I want to learn. I want to learn what I value or am convinced by people whom I trust that I will ev entually value. . . . I want to learn . . . what I am good at. (Sizer, 1984, p. 164)
38 Sizerâ€™s answer to the question of how to stimulate engagement encompasses many of the variablesâ€”including a sense of competence, autonomy, utility, and connection with trusted othersâ€”that have been identif ied across different theories of academic motivation and engagement (Nat ional Research Council & Ins titute of Medicine, 2004). Each theory emphasizes different component s of Sizerâ€™s answer or, in some cases, underscores components that Sizer did not lis t. However, the similarities among the theories described below are more robust th an the differences, thus pointing the way towards the active ingredients identifie d at the beginning of this chapter. Newmann and authentic pedagogy Fred Newmann, a major proponent of school restructuring and performance-based assessment, has written extensively on how to increase student engagement. In one of his earliest theoretical papers (Newmann, 1981), he argued that engaging schools are those that reverse the typically alienating features of most high schools: the lack of strong personal connections with peers and teachers and the decontextualized nature of the academic work. Thus the major foundations of engagement in the model that Newmann and his colleagues have articulated over two decades are studentsâ€™ sense of personal membership in the school and their sense of the authenticity of their academic work. To the extent that feelings of belonging a nd perceptions of authenticity are missing, a studentâ€™s underlying need for competence will not be channeled into academic work (Marks, 1995; Newmann, 1989, 1998; Newmann et al., 1992). Membership entails a sense of being supported by teachers and peers. This support is important partly because deep learning of ten requires risk taking, and people are more apt to take risks when they feel cushione d by supportive others. Support also contributes less directly to academic engagement by enhancing student bonding to staff, which in
39 turn strengthens studentsâ€™ commitment to th e schoolâ€™s ends and means (Newmann et al., 1992). The other major component of Newmannâ€™ s model of engagement, authenticity, is a broad construct. Tasks are authentic to the extent that they have the following features: aesthetic, utilitarian, or personal value ap art from documenting the competence of the learner; parameters that allow students signi ficant input into the conception, execution, and evaluation of the work; and resemblance to the kind of work that people do and value in the world outside school (Newman n, 1989; Newmann, Marks, & Gamoran, 1995; Newmann & Wehlage, 1995; Newmann et al., 1992). The suggestions for school restructuring based on these ideas are unsurprising. Small class and school size, as well as mo re opportunities for non-academic contact between students and teachers (e.g., advisory gro ups that stay together for several years), are proposed to create a stronger sense of support for students (Lee & Smith, 1994, 1995; Newmann, 1981). A school with more au thentic tasks would be one in which performance-based assessments and individually -designed research projects are common (Lee & Smith, 1994, 1995; Newmann & Wehlage, 1995). The model proposed in this dissertation de rives support for several variables from Newmannâ€™s work. One, related to his em phasis on student bonding to staff as a key motivator, is Advisor Support. From Newma nnâ€™s multifaceted constr uct of authenticity comes Utility Valueâ€”related to the notion of work that people do and value in the â€œreal worldâ€”and Autonomyâ€”related to the idea that students wo uld have significant input into the design, execution, and assessment of their own work. Recall that Newmannâ€™s theory suggests that the presence of thes e factors determines the extent to which a personâ€™s fundamental need to feel and demonstrate competence is channeled into
40 academic work. The idea that a sense of competence is a powerful motivator is reinforced by each of the theories revi ewed here, especially the following one. Self-efficacy theory Self-efficacy is a judgment of how capable one is to perform a specific task successfully (Bandura, 1977, 1986, 1997; Pajares, 1996; Schunk, 1985). Before proceeding it is important to distinguish se lf-efficacy from related constructs. Selfconcept is one. Self-concept is domain-specif ic but not task-specifi c, and it incorporates self-evaluative feelings that are not in cluded in strict self-efficacy judgments (Alvermann, 2002; Schunk & Pajares, 2002; Zimmerman, 2000). Asking someone â€œHow good are you at English?â€ is a request for a domain-specific self-concept judgment; in contrast, asking â€œHow confident are you th at you can diagram thes e sentences?â€ is a question about self-efficacy (Zimmerman, 2000). Another concept sometimes co nflated with self-efficacy is perceived control. Central to the work of Skinner and her colleagues (e.g., Skinner, Connell, & Wellborn, 1990; Skinner et al., 1998), perc eived control does include beliefs about the self as capable of specific actions; however, it also in cludes beliefs about the extent to which luck, effort, and ability are required for su ccess and the degree to which a person feels lucky, effortful, and able. In other word s, perceived control includes not only selfefficacy, but also locus of control (Rotter, 1966) and attributional st yle (Weiner, 1974). An example may help to differentiate self -efficacy and perceived control: A student may have high self-efficacy (â€œI am very confiden t in my ability to learn this material on stoichiometryâ€) but low percei ved control (â€œUnfortunately, th e professor grades so hard andâ€”just my bad luckâ€”my classmates are so smart that Iâ€™m not sure I will do well on the testâ€) (Pintrich, 2003).
41 Self-efficacy for a particular task is belie ved to promote achievement via its effects on engagement (e.g., Bandura, 1986, 1997; Linne nbrink & Pintrich, 2003; Pintrich & De Groot, 1990; Schunk, 1995). Evidence that self-efficacy is indeed positively linked with achievement is abundant (see, e.g., revi ews by Bandura, 1997; Pajares, 1996; and Schunk, 1991). For example, in mathematics, Pajares and colleagues have shown a strong link between self-efficacy and subsequent performance. In a sample of college students, self-efficacy for solv ing specific problems was a stronger predictor of their performance than were gender, prior experien ce in mathematics, anxiety and self-concept related to mathematics, and perceived usefulne ss of the subject (Paj ares & Miller, 1994). Pajares and Kranzler (1995) added a measur e of general mental ability (the nonverbal Ravenâ€™s Advanced Progressive Matrices test) to this picture and found, in a sample of high school students, that se lf-efficacy was as strong a pr edictor of performance as aptitude, with both measures be ing significantly better predicto rs than the others in the model. The link between self-efficacy and achievement is equally clear in a very different academic domain such as writing. For example, college undergraduatesâ€™ ratings of their confidence in such writing tasks as using parts of speech correctly and organizing an argument accounted for significant varian ce in their performa nce on holisticallyscored writing samples (Shell, Murphy, & Br uning, 1989). This well-established link between self-efficacy and performance, combined with the evidence connecting engagement with performance (reviewed earli er) and self-efficacy with engagement (to be reviewed), is consistent with the aforementioned models that interpose engagement between self-efficacy and achievement.
42 In addition to contributing directly to engagement, self-efficacy is a potential mediator for other variables linked to e ngagement (Bandura, 1986). Schunkâ€™s (1981) work with elementary students provides one example. Participan ts in this study, all struggling with school mathematics, took part in one of two interventions: one modeled appropriate problem-solving skills and the ot her provided direct in struction in those skills. The interventions had an effect on the outcome of interest (p ersistence on a set of mathematics problems) only through their e ffect on student self-efficacy. This study demonstrated the operation of two of the four major sources of self-efficacy identified by Bandura (1986, 1997): in the first interventi on, exposure to efficacious models; and, in the second intervention, verbal feedback and persuasion from significant others. The other sources of information that people us e to form self-efficacy judgments are past experiences of success (or failure) with tasks like the one in question and the experience of affective and physiological arousal (e.g., anxiety) dur ing such performance. The self-efficacy perspective reinforces the central place of competence in the present engagement model. It also builds upon Newmannâ€™s emphasis on support by broadening the scope of social supports for learning beyond teachers to include parents, who can promote a studentâ€™s self-efficacy th rough verbal and instrumental support. Hence the addition of Parent Support to the model. Finally, the idea that feedback enhances self-efficacy partly by making e xpectations clearer (Bandura, 1997; Schunk, 1983, 1984; Schunk & Pajares, 2002) ju stifies the addition of Clar ity of Expectations to the model. Expectancy value theories Self-efficacy is a necessary but not sufficient condition for academic engagement. Even if people are confident that they can perform a task successfully, they generally will
43 not invest much effort in the task if they do not value its outcomes (Good & Brophy, 1997). This premise is at the heart of exp ectancy value theori es, which propose that motivation and engagement are a joint func tion of expectancies of success and the subjective value that the individual attaches to the task itself or its outcomes. First articulated by Atkinson (1964) in the contex t of general achievement motivation and by Porter and Lawler (1968) in the context of employee motivation, expectancy value theories have been most recently applied to education in the work of Eccles, Wigfield, and their colleagues (e.g., Eccles et al., 1983; Wigfield, 1994; Wigf ield & Eccles, 2000; Wigfield & Tonks, 2002). Kellerâ€™s (1987) mode l of instructional design, widely used by educators and applied researchers to build motivational principles into lesson plans, is predicated on these theories. The way that Eccles, Wigfield, and thei r colleagues have op erationalized the expectancy component of expectancy valu e theory is different from the way Bandura and his colleagues have measured self-effi cacy. Whereas self-efficacy is generally measured by asking people how confident they ar e that they can complete particular tasks successfully (Bandura, 1997; Pajares, 1996), the expectancy construct is broader, resembling domain-specific self-concept. For example, Wigfield and Eccles (2000) listed the following survey items used in their research on expectancies for success in mathematics: How good in math are you? If you we re to list all the st udents in your class from the worst to the best, where would you put yourself? Compared to most of your other school subjects, how good are you in math ? How well do you expect to do in math this year? How good would you be at learni ng something new in math? These items
44 consistently have cohered in factor analys es conducted with samples of children and adolescents (e.g., Eccles & Wigfield, 1995). Like expectancy, value is a multifaceted construct. Its major components are attainment value, defined as the importance of doing well on the task for oneâ€™s sense of self; intrinsic value, the extent to which one derives pleasure from doing the task; and utility value, the degree to which the task ha s instrumental value in reaching oneâ€™s shortand long-term goals (Eccles, 1994; Eccles et al., 1983). Though positively correlated, these three facets of value ar e in fact distinguished by st udents, as demonstrated by confirmatory factor analyses with 700 student s in grades 5-12 (Eccl es & Wigfield, 1995). While the factor structures for the younger (g rades 5-7) and older (g rades 8-12) students were the same in this study, the authors neve rtheless suggested that age groups may vary in which of the facets better pr edicts engagement. Consistent with this conjecture, Harter (1981) found in a sample of 3,000 students in grades 3-9 that motivation for academic achievement was more extrinsic for the older students; specifically, students were motivated more by teacher approval than by curiosity and preferred easier work that made them look good over harder work that w ould challenge and improve their skills. Along similar lines, Wigfield and Eccles ( 1989) found age differences in the components of value that predicted studentsâ€™ intentions to continue taking mathematics classes. For junior high students, the only si gnificant predictor was the intrinsic value of mathematics, and for high school students, both intrinsic value and utility value were significant predictors. As do the motivational theories devel oped by Newmann and Bandura, expectancy value theories highlight the importance of a studentâ€™s sens e of efficacy. Also, in its
45 emphasis on utility value, the work of Eccles , Wigfield, and their colleagues reinforces Newmannâ€™s focus on authentic academic work. In sum, the components of my model that are firmly rooted in expectancy value concepts are Senior Project Self-Efficacy and Utility Value. Self-determination theory Self-Determination Theory (Deci & Ry an, 1985; Ryan & Deci, 2000) emerged from a very different paradigm than did the preceding theories. The â€œorganismicâ€ perspective, perhaps best exemplified by Ma ria Montessori and Ca rl Rogers, assumes that motivation for learning is natural and intr insic and that environments must be altered to support rather than thwart that potential (Ryan & Powelso n, 1991). Firmly within this tradition, Ryan, Deci, and their colleagues have proposed that the level of engagement in an activity is increased to the extent that the activity and the environment in which it occurs satisfy a personâ€™s basic psychological needs. These fundamental human needs are competence, relatedness, and autonomy. When satisfied, these needs transform into inner motivational resources upon which people can draw. This theory has been applied specifically to education in the work of Sk inner, Connell, Wellborn, and their associates (e.g., Connell, 1990; Connell & Wellborn, 1991; Skinner et al., 1990). By now a familiar component of engagement theories, competence generally has been operationalized within the SDT paradigm as â€œperceived controlâ€â€”the composite of self-efficacy, locus of control, and attribut ional style described previously (Connell & Wellborn, 1991; Skinner et al., 1990; Sk inner et al., 1998). Among the proposed supports for perceived control are clear expe ctations for performance, positive feedback, reliable assistance, and consistent cons equences (Connell & Well born, 1991; Skinner & Belmont, 1993).
46 The second basic need, relatedness, is si milar to Newmannâ€™s concept of school membership, reflecting a general striving towards support from and community with significant others (Ryan & Deci, 2000). The major focus of research on relatedness in academic settings has been studentsâ€™ feelings of support and connection with teachers (e.g., Skinner et al., 1990); these feelings ge nerally have been measured by questions about teachersâ€™ expression of affection and encouragement, their provision of instrumental help, and their sensitivity to student needs (Connell & Wellborn, 1991). Similar questions about peers in school and parents also ha ve been used to expand the scope of relatedness to othe r potential sources of support and belonging (e.g., Furrer & Skinner, 2003). Autonomy, the third fundamental psychologi cal need according to SDT, refers to the experience of choice in in itiating and regulating oneâ€™s behavior (Deci & Ryan, 1985; Ryan & Deci, 2000). As the name of th e theory suggests, autonomy is the sine qua non of SDT. Much of what people do, accord ing to the theory, is not intrinsically motivatedâ€”appealing and enjoyable in its own right. However, extrinsically motivated behavior can vary in the degr ee to which it is perceived as autonomous or controlled. Behavior strictly dictated by another person and rigorously bound by a system of rewards and punishments is generally experienced as high ly controlling. In contrast, tasks that are not entirely freely chosen but that allow opportunities for sel f-direction and expression of feelings are generally perceived as mo re autonomous. The amount of autonomy experienced by a person during a task is hypothesi zed to be the most important factor in explaining the individualâ€™s level of engage ment in the task (Connell & Wellborn, 1991; Deci & Ryan, 1987; Ryan, Connell, & Deci, 1985; Ryan & Deci, 2000).
47 From SDT the most significant contri bution to the proposed model is the Autonomy variable. In additi on, the importance of Clarity of Expectations is reinforced by its proposed link to the fundamental need for a feeling of competence. Finally, the emphasis on relatedness reinforces the thre e support variables in the model: Advisor Support, Peer Support, and Parent Support. Section Summary and Prospectus All four theoretical frameworks include student feelings of competence as a key motivational variable. Self-efficacy therefore occupies a central position in the present model, contributing directly to engagement and serving as a potential mediator for many of the other variables. The variables acco mpanying self-efficacy under the umbrella of â€œperceived controlâ€ (e.g., the studentâ€™s locus of cont rol) are beyond the scope of program-manipulable variables of interest in this study. Another common denominator across several theories is the dependence of self-efficacy and engagement on support from significant others in the learning environm ent. Teachers, peers, and parents are seen as a key motivational resource, especially when tasks are challenging (Darling-Hammond & Bransford, 2005; Martin, 2005; National Re search Council & Institute of Medicine, 2004). The perceived real-world value of the task to the individual also appears in multiple theories and therefore merits a place in the current model of Senior Project engagement. Autonomy, though prominent in just one of the four theories, is included not only because it occupies the central position in that well-tested theory, but also because it permeates the descriptive literature on high school in general and the senior year in particular. The two variables in the model that are not clea rly grounded in these theories,
48 Sensitivity and Novelty, derive most of their s upport from that same literature, as well as other, atheoretical empirical research. The goal of the next two sections is to defend each path (hypothesis) in Figure 2-1. As indicated, two of the paths in the model lie beyond the immediate influence of educators who design and administer Senior Pr oject programs. I refer to the variables emitting these paths as â€œinactive ingredientsâ€ only to differentiate them from the â€œactive ingredientsâ€ over which Senior Project coordinators have mo re control. I preface each description with a high lighted copy of the full model diag ram to remind the reader how each path fits into the larg er context of the model. A Baseline for Engagement: Inactive Ingredients As indicated earlier, two path s in the model of Senior Project engagement mostly lie beyond the immediate influence of edu cators who design and administer Senior Project programs. One path suggests the infl uence of studentsâ€™ general engagement in school (Academic Engagement) on their investme nt in the specific school task of Senior Project (Senior Project Engagement). The other path indicates the contribution of studentsâ€™ previous level of success with project-related skills (Previous Mastery Experience) to their self-efficacy for project work (Senior Project Self-Efficacy). Failing to include these links, for which both ra tional arguments and empirical support are compelling, would increase the probability of overestimating the effects of the more controllable variables. However, because th ese links are not the primary focus of this study, the defense of each will be brief. Effect of Academic Engagement on Senior Project Engagement The path highlighted in Figure 2-2 suggests that seniors who are more behaviorally, emotionally, and cognitively engaged in their cl asses will tend to be more engaged in
49 Senior Project. This hypothesi s arises not from any particular theory or set of empirical findings but, instead, from a l ogical argument. Studentsâ€™ ge neral engagement in their coursework reflects both their basic dis position towards academic matters and the influence of variables that are beyond the control of a Senior Project coordinator, such as the amount of time they spend in part-time employment and the volume of stress and work they experience in their family situations. These external factors seem likely to influence a studentâ€™s engagement in Senior Pr oject work primarily through their effect on a studentâ€™s general commitment of time and energy to academic matters. Senior Project Engagement Senior Project Self-Efficacy Advisor Support Peer Support Parent Support Clarity of Expectations Autonomy Utility Value Sensitivity Novelty Previous Mastery Experience Academic Engagement Figure 2-2.Model diagram highlighting the e ffect of Academic Engagement on Senior Project Engagement. Academic Engagement should not, however, be a perfect predictor of Senior Project Engagement. Indeed, one of the argumen ts made in favor of Senior Project is its potential as an â€œequalizerâ€ th at provides opportunities for en gagement and success to all students, not just the usual academic st andouts (Darling-Hammond et al., 1995; Parizek & Kevasan, 2000). Therefore, it is possible th at more than a few students whose general academic engagement is low nevertheless woul d agree with the following student whose Senior Project was writing, directing, a nd performing in a 90-minute comedy routine:
50 â€œ[Senior Project] is one of the best things Iâ€™ve ever done in my life. Iâ€™ve never been so successful in a class before. Itâ€™s funny it ha ppened second semester senior yearâ€ (DeFao, 2005, p. B5). Effect of Previous Mastery Experien ce on Senior Project Self-Efficacy According to Bandura (1986, 1997; see also Schunk, 1984), the most reliable source of information that people use in j udging their self-efficacy for a task is their previous performances on tasks like the curren t one. The path highlighted in Figure 2-3 represents the hypothesis that, ot her things being equal, more prior success with skills related to Senior Project will support greater self-efficacy for Senior Project. Senior Project Engagement Senior Project Self-Efficacy Advisor Support Peer Support Parent Support Clarity of Expectations Autonomy Utility Value Sensitivity Novelty Previous Mastery Experience Academic Engagement Figure 2-3.Model diagram highlighting the e ffect of Previous Mastery Experience on Senior Project Self-Efficacy. The ceteris paribus qualifier in the previous sentence is important partly because prior successes and failures do not translate directly into greater or lesser self-efficacy. Rather, there is a mediating inferential process shaped by contextual and personal factors (Bandura, 1997). For example, a dispositiona l optimist may maintain high self-efficacy for a task even in the face of past fail ure. Conversely, and perhaps more commonly,
51 students who have been successful neverthele ss may doubt their abil ities. While it is important to consider these complexities, they will not be explored fu rther in this review. Rather, the focus is on the st rength of previous mastery experiences as a source of self-efficacy. In one relevant study, Lopez a nd Lent (1992) examined how well the four sources of efficacy information identified by Banduraâ€”previous experiences, verbal persuasion, modeling, and physio logical states during task performanceâ€”predicted mathematics self-efficacy in a sample of 50 high school juniors. Self-efficacy was measured as confidence ratings for being able to solve each of twenty course-related problems. An index of general academic self-confidence also was obtained from these students and used in a stepwise multiple regre ssion to predict self-efficacy ratings. Of all the sources of self-efficacy information, only past performance in mathematics (measured by previous grades and other recognition for ma th achievement) led to significant change in the proportion of variance explained ( R2 .40). The importance of previous mastery expe riences as a source of self-efficacy appears to increase as students advance in school. In a sample of elementary and middle school students, Skinner et al. (1998) te sted the following model: teacher-provided structure in the classroom pred icts studentsâ€™ perceived contro l, which in turn predicts their behavioral and emotional engagement in class, which predicts performance in the class, which in turn has a feedback effect on subsequently measured perceived control. The effects of performance (measured by course grade) on one aspect of perceived controlâ€”beliefs about oneâ€™s ability, which are most similar to Banduraâ€™s self-efficacy constructâ€”were stronger for the older st udents. Specifically, there was no evident
52 feedback from achievement to ability beliefs fo r third-graders, a moderate effect for fifthgraders, and a strong eff ect for seventh-graders. Analyses reported by Marks (2000) show a similar grade-level dependency for the link between previous mastery experiences and engagement. The national sample in this study included approximately 4,000 total students in grades 5, 8, and 10 from schools that were determined to have made significant pr ogress in restructuring. Students responded to survey items about their personal bac kground, the degree of support for achievement from peers and teachers, and the authenticity of the work in their classes. The criterion, student engagement in instru ctional activity, was measured by student reports of effort, attentiveness, boredom, and co mpletion of class assignments. Previous success in school, measured by GPA prior to the study, was a st rong positive predictor at all three grade levels but was strongest for the high sc hool students, accounting for more withinclassroom variance in engagement th an any other predictor variable. In sum, prior success in tasks related to Senior Project, such as researching and writing, generally should be associated with greater confidence in oneâ€™s ability to complete Senior Project successfully. However, as in the previously described path in the model, the association should be far fr om perfect. Because of differences in personality and contextual influencesâ€”inc luding support from advi sors, parents, and peersâ€”students with comparable levels of prior success may calibrate their self-efficacy differently. Promoting Engagement: Active Ingredients The remainder of this review focuses on th e paths in the model that may be most useful to developers and admini strators of Senior Project programs. For example, if the study were to reveal that Parent Support is strongly linked to Seni or Project Engagement,
53 then a campaign of parent outreach might be a dvisable. Or, if the paths from Clarity of Expectations to Senior Project Self-Efficacy and from thence to Senior Project Engagement were especially robust, then a project coordinator woul d be well advised to guide students through examples of excellent and poor projects. Effect of Senior Project Self-Effi cacy on Senior Project Engagement Self-efficacy researchers emphasize the c onnection between how confident people feel in their ability to perform a task and how hard they work, how long they persist, how comfortable they feel, how deep ly and flexibly they think, an d how often they return to the taskâ€”in short, how engaged they are (e.g., Bandura, 1977, 1997; Guthrie & Wigfield, 2000; Linnenbrink & Pintrich, 2003; McComb s & Whisler, 1989; Pajares, 1996, 2002; Pintrich & Schrauben, 1992; Schunk, 1989). This idea is highlighted in Figure 2-4. Senior Project Engagement Senior Project Self-Efficacy Advisor Support Peer Support Parent Support Clarity of Expectations Autonomy Utility Value Sensitivity Novelty Previous Mastery Experience Academic Engagement Figure 2-4.Model diagram highlighting the e ffect of Senior Project Self-Efficacy on Senior Project Engagement. This connection certainly applies at the most minimal level of academic engagement, namely, the decision to stay in school. Such was the finding of Vallerand, Fortier, and Guay (1997), who administer ed surveys to approximately 4,500 students attending grades 9 and 10 at seven Canadi an public schools. The researchers obtained
54 school records one year later. Student s who had dropped out by this time had significantly lower perceived academic competen ce than those who were still in school. Similar findings were reported by Hardre a nd Reeve (2003), who studied 500 students in rural U.S. high schools. Self-reported inten tions to drop out of school were predicted most strongly by poor academic performan ce (measured by GPA); however, even after accounting for this predictor, a significant am ount of variance in dr opout intentions was explained by a combination of two motiv ational variables, perceived academic competence and perceived autonomy. Beyond the minimal engagement of staying in school and attending classes, more active forms of engagement also appear to de pend on a studentâ€™s feel ing of self-efficacy. Persistence and the amount of time spent on a ta sk are two such indica tors. Wettersten et al. (2005) used self-reported time spent on wo rk for English and mathematics courses as a measure of engagement in a sample of near ly 700 students in nine rural high schools. In a multiple regression whose predictor variables included measures of parental involvement and of student beliefs about th e relevance of educa tional performance to their career goals, academic self-effi cacy was the strongest predictor ( rpartial = .25) of the engagement measure. Experimental studies using task persistence (time on task) as the measure of engagement also indicate the importance of self-efficacy. In one study (Bouffard-Bouchard, 1990) college students trie d to complete verbal puzzles and then received bogus feedback about their performa nce relative to their p eer group. This selfefficacy manipulation was indeed successful, as indicated by participantsâ€™ subsequent ratings of their efficacy for solving four addi tional puzzles. After providing these ratings, participants had a 20-minute period during whic h they tried to solv e these four puzzles
55 and at the end of which they had the option to work for an additional five minutes. Students whose self-efficacy was high (because of the manipulation) took more time and completed a greater number of problems than pa rticipants in the low self-efficacy group. In a different sample of college stud ents, Brown and Inouye (1978) manipulated participantsâ€™ self-efficacy for anagram-solv ing tasks by varying the success and the purported ability level of a model who first wo rked on the tasks. Students in the higher efficacy groups persisted longer on the task than those in the lower efficacy groups. Moreover, the strength of th is relationship increased over trials, suggesting that the students came to rely more and more over ti me on their judgments of self-efficacy to regulate their effort expenditure. Indicators of engagement other than persis tence also are related to self-efficacy in samples of older students. For example, Wolters (2003) administered a survey to university students that addressed the followi ng: the studentsâ€™ self-efficacy for learning and completing assigned tasks in a particular class; their general orientations toward learning (e.g., relative emphases on learning for its own sake vs . learning because it brings valued extrinsic rewa rds); and their procrastinati on behaviors in the class. Procrastination, manifesting an underlying relu ctance to initiate activity, is a negative indicator of engagement. Regression analys es on two different samples showed that higher self-efficacy was the stronge st and most consistent pred ictor of lower self-reported incidence of procrastin ation. Another measure of engage ment is self-reported effort. MacIver, Stipek, and Daniel s (1991) reported a study that used structural equation modeling to test a model meant to explain ch anges in effort in a specific class over the course of a semester. The independent vari ables were changes in each of the following
56 variables: self-concept of ability in the pa rticular subject, extrinsic pressures for achievement, perceived utility value, and intrin sic interest in the course material. In samples of junior high and high school students, the best-fit ting models were those in which change in self-efficacy was the independe nt variable showing the strongest direct effect on change in effort. Studies with more elaborated measures of engagement have suggested similarly strong links between self-efficacy and engageme nt. For example, Laffey (1982) studied a sample of 88 African-American sophomores in their social studies classes at an urban high school. Measures of engagement incl uded attendance figures, ratings of class participation by teachers, ratings of on-task behavior by a classroom observer who observed each class ten times, and student ratings of their degree of interest when cued by a random signal once during each of six class periods. Regression analyses using these measures as separate criterion variables assessed the relative influence of studentreported efficacy, locus of control, educationa l aspirations, and othe r variables. Across the different engagement measures, self -efficacy was the most consistent strong predictor, with R2 reaching as high as .21 for some of the engagement measures. This finding was reinforced by interviews with te n students, most of whom mentioned that they felt most involved in classes in which they felt successful. Engagement measures that tap into more subtle, self-regulatory f acets of commitment to oneâ€™s work were used in a study by Miller et al. (1996). In the cr oss-sectional study, high school students rated their perceived ability in ma thematics, as well as their endorsement of certain goals during their academic work, including learni ng for its own sake, for obtaining rewards (e.g., grades), and for pleasing the family. Wh en these variables were used in an all-
57 possible-subsets regression to predict differe nt types of engagement, self-efficacy was a significant predictor of self -reported persistence (e.g., â€œW hen I run into a difficult homework problem, I usually give up and go on to the next problemâ€) and self-regulation (e.g., â€œI organize my study time well for this classâ€). Self-regulation is one example of higher-o rder engagement. Whereas attendance, time on task, and persistence relate more to the quantity of effort, cognitive and selfregulatory strategies essentially indicate the quality of effort. The connection between these forms of engagement and self-efficacy is supported by a body of evidence just as extensive as that supporting the lower-orde r but prerequisite forms of engagement described thus far. (For reviews, see Pintrich, 1999, 2003; Pintrich & Schrauben, 1992; and Zimmerman, 2000.) In one study, for ex ample, approximatel y 200 seventh-grade students from both science and English classr ooms completed self-re port measures of course-specific self-efficacy, in terest in the material, and test anxiety (Pintrich & De Groot, 1990). The dependent variables were self-report measures of cognitive strategy use (e.g., use of rehearsal and elaboration to learn material) a nd self-regulation (e.g., monitoring oneâ€™s comprehension during reading) . With participants divided into high and low self-efficacy groups based on a median split, and with prior achievement (firstsemester grade) as a covariate, Pintrich a nd De Groot found that students high in coursespecific self-efficacy were more likely to report using cognitive and self-regulatory strategies during learning. The same va riables were measured in a study with 500 seventhand eighth-grad e students rating themselves in each of three subject areas: social studies, English, and mathematics (Wolters & Pintrich, 1998). Regression analyses showed that, although interest was the best predictor of cognitive strategy use, self-
58 efficacy nonetheless was a significant predicto r for all three subject areas, explaining between 2% (math) and 9% (English) of th e variance in the criterion. Wolters and Pintrich reported similar results for the pr ediction of self-regula tion. Moreover, unlike interest, self-efficacy was a significant pred ictor of studentsâ€™ actual grades in these subjects. Additional studies supporting the link between self-efficacy and higher-order engagement have been reported. For exam ple, Zimmerman and Martinez-Pons (1990) conducted structured interviews with students in grades 5, 8, and 11. The interviewer presented students with learning scenarios and asked them to describe what steps they would take. Responses were coded into a va riety of categories, including goal-setting and planning, seeking further informati on, rehearsing and memorizing, and selfevaluating. After the intervie ws students rated their verbal and ma thematical selfefficacy by indicating their degree of confid ence for each of a graduated series of domain-appropriate tasks. Across all three grades, multiple correlations of .41 (mathematics) and .42 (verbal) were reported, indicating that approximately 17% of the variance in self-reported use of active learni ng strategies was shared with variance in efficacy beliefs. A study offering stronger ev idence, in part because it controlled for aptitude, was reported by Bouffard-Bouchard, Parent, and Larive ( 1991). Junior high and senior high school students rated their efficacy for verbal tasks and then were observed as they tried to solv e four verbal concept-learning puzzles. In both age groups, and controlling for verbal aptitude, student s with higher self-effi cacy were better at monitoring their working time, more persistent, less likely to reject correct hypotheses prematurely, and better at solving conceptual problems. The results of this study were
59 well summarized by Bandura (1997): â€œThe se lf-assurance with which people approach and manage difficult tasks determines whet her they make good or poor use of their capabilities. Insidious se lf-doubts can easily overrule th e best of skillsâ€ (p. 35). Across both correlational and experimental research and across a variety of measures of engagement, ranging from the most minimal (staying in school) to the most cognitively sophisticated, self-e fficacy is strongly associated with greater investment in academic tasks. Effect of Utility Value on Senior Project Engagement Even if students feel competent and know what to do to succeed, they may choose not to engage deeply if the activity has litt le value to them (Wigfield & Eccles, 2000). The link emphasized in Figure 25 suggests that students will be more engaged in Senior Project when they perceive that the work is useful in preparing them for future personal and real-world challenges. The link between perceived relevance a nd engagement is central to both the expectancy value models (e.g., Wigfie ld & Eccles, 2000) and the authenticity framework (e.g., Newmann et al., 1992) de scribed earlier. However, calls for contextualizing academic learning and making its instrumental value clear are neither new (see, e.g., Bruner, 1966) nor restricted to these theories (see, e.g., Good & Brophy, 1987; Keller, 1987; Lepper, 1988; Pedersen, 2003; Resnick, 1987; Rinne, 1998; Small & Arnone, 2000). In fact, such calls are a majo r part of recent efforts to reform high schools (e.g., Dunn, 2001; Kantrowitz & Wingert, 2000; National Commission on the High School Senior Year, 2001b). These effort s are partly a response to a sentiment
60 Senior Project Engagement Senior Project Self-Efficacy Advisor Support Peer Support Parent Support Clarity of Expectations Autonomy Utility Value Sensitivity Novelty Previous Mastery Experience Academic Engagement Figure 2-5.Model diagram highlighting the e ffect of Utility Value on Senior Project Engagement. expressed clearly by a seni or who called the Nationa l Public Radio program Talk of the Nation during a show on senioritis: â€œI already knew what I wanted to become. I wanted to go into art illustration. . . . [Itâ€™s] pointless for me to sit here and learn what hydrogen bonding is because Iâ€”why am I gonna need to know that?â€ (Conan, 2005, p. 9). Many students share this view. For ex ample, Lowe (2003) spent 18 months observing classes and interviewi ng students at a restructuring high school in a rural area. One theme from the interviews was that students wanted lessons in school that had â€œrealworld significanceâ€ (p. 11) and that would help them prepare for their future careers. These findings are reinforced by results fr om the Colorado High School Senior Survey described earlier (Colorado School-to-Career Partnership, 1999). Of the nearly 9,000 seniors who completed the survey, 61% indicat ed that they were motivated by classroom instruction that included â€œs olving real-world problems.â€ This percentage was significantly higher than most of the other survey op tions for making classroom instruction more motivating (e.g., wa tching videos, using technology).
61 More oblique indications of the value th at high school students attach to â€œreal worldâ€ experiences have been found in studi es related to student employment. For instance, in interviews with 75 academically successful seniors from suburban Chicago high schools, Linton and Pollack (1978) learned that these st udents sometimes turned to part-time employment less because they n eeded the money and more because they wished to have experiences that felt more â€œrea lâ€ than school. In a book of essays written by students from another suburban high school in Chicago, the editors (Gilbert & Robins, 1998) interspersed qualitative analyses of the more than 100 essays submitted for the book. Gilbert and Robins noted the following th eme: on the job, students interact in an environment where they learn important social and occupational skills. Because school bears so little resemblance to the world they see in their jobs, students generally see school as having little releva nce to preparation for their postsecondary lives. These findings are consistent with a report from the U.S. Department of La bor (The Secretaryâ€™s Commission on Achieving Necessary Skills, 1991): Part of the difficulty is that employers and school personnel are passing each other like ships in the night. . . . As a cons equence of this miscommunication, secondary school students often see little connection between what they do in school and how they expect to make a living. They ther efore invest very li ttle effort in their education. . . . The sense that students cl early distinguish between what goes on in their classrooms and what goes on in the â€œr eal worldâ€ was palpable in focus groups convened as part of the SC ANS research. (pp. 20-21) The belief that high school is not very usef ul is not restricted to students who work or students who intend to enter the workfor ce instead of attending college. Rosenbaum (1998) surveyed over 2,000 seniors in 12 Ch icago high schools. In response to the statement, â€œSchool teaches me valuable skills,â€ nearly 30% of respondents rated themselves as neutral or in mild or str ong disagreement. There was little difference between workand college-bound students on this item and others measuring the
62 perceived relevance of school to oneâ€™s futu re life (e.g., â€œMy courses give me useful preparation Iâ€™ll need in lifeâ€) . Interviews with 50 seni ors in two of the high schools reinforced the impression that students struggle to see the connection between high school and their lives. Similar, if perhaps more alarming, results were found in a 1997 telephone poll of a nationally representative sample of 1,000 public high school students (Johnson & Farkas, 1997). Rating the statement, â€œI wonâ€™t really ne ed to know most of the things my school is teaching me when I get out into the real world,â€ 55% of respondents said that the stat ement was â€œveryâ€ or â€œsomewhatâ€ close to how they felt. The perception of schoolâ€™s usefulness appears to have declined over time and appears to decline as studen ts progress through school. Tren d data from the Monitoring the Future survey, which has sample d about 15,000 seniors annually since 1976, suggested a significant drop in the percentage of students who regard ed what they were learning in school as importan t for their later li vesâ€”from 52% in 1980 to 40% in 1999 (Boesel, 2001). In a cross-se ctional study reported by Harter (1996), students in grades 6, 7, and 8 rated the validity of a number of po ssible reasons that stud ents get â€œturned offâ€ to schoolwork. The higher the grade level, the more students felt that â€œschoolwork just isnâ€™t that important or relevant to studentsâ€™ lives.â€ These two trends parallel those noted earlier in this chapter: declining academic engagement over historical time and over the course of a studentâ€™s tenure in primary a nd secondary education. The following evidence suggests that this paralle l is not coincidental. Two small-scale studies, widely separate d in time, suggest a connection between utility value and engagement. In a classi c study Stinchcombe (1964) spent six months observing classes and interviewing students and faculty in a California high school. The
63 interview data indicated th at, especially for work-bound students, school effort was largely determined by the perceived relevance of the work to the world they soon would enter. An equal emphasis on the motivational potential of material connected to the â€œreal worldâ€ was noted by Cushman (2003) in her summary of extensive focus-group sessions with 40 high school students in New York, Rhode Island, and California. Diverse in socioeconomic status, ethnicity, and academic performance, the students in these groups frequently commented on the need to unders tand the usefulness of what they were learning. One student said, â€œJust saying you ne ed to pass math isnâ€™t enough. Show me how knowing pi is worth somethingâ€ (p. 10). Larger studies with quantita tive analyses support the preced ing qualitative research. In one study (Ainley, 1994), analysis of surveys completed by approximately 3,000 Australian seniors revealed a significant pos itive correlation (.49) between measures of the perceived relevance of schooling (e.g., â€œI ha ve acquired skills that will be useful to me when I leave schoolâ€) and deep appr oaches to learning (measured by items distinguishing superficial approaches from more effortful and cognitively demanding approaches to coursework). Correlations of similar magnitude between similar constructs were reported by Watkins and Hattie (1990) in a sample of 1,300 students attending grades 7, 9, and 11 at 18 different Australian high schools. As these studies controlled few variables, their simple correlational fi ndings are open to a va riety of alternative interpretations. More compelling is one of the findings from Rosenbaumâ€™s (1998) previously described survey of over 2,000 senior s in Chicago. In addition to rating their perceptions of the utility value of school, par ticipants rated their engagement in terms of time spent on homework and frequency of doing more than the minimum work needed to
64 pass their courses. After other factors (inc luding gender, race, and parental and peer support for school effort) were controlled in a multiple regression, future relevance had a significant unique effect on e ngagement, yielding a standard ized coefficient that was among the highest in a set of variables accounting for 34% of the variance in effort. As in the previous two studies, however, the inte rpretation that student s exert more effort because they perceive their work as useful mu st be balanced against other possibilities. For example, students may rationalize low effort post hoc by claiming that high school was not very useful. Two additional quantitative studies sugge st the potential motivating power of utility value. In one (Ass or, Kaplan, & Roth, 2002), appr oximately 900 Israeli students in grades 3-8 completed questionnaires about teacher behaviors in a particular class and their own behavioral, cognitive, and emotional engagement in the class. In a regression that included measures of the extent to which teachers allowed students choice and encouraged students to express themselves, th e teacher behavior of â€œfostering relevanceâ€ (e.g., â€œMy teacher talks about the connection between what we study in school and what happens in real lifeâ€) was the strongest pr edictor of behavioral, emotional, cognitive engagement in both age groups (grades 3-5 and 6-8). A very di fferent measure of engagement was used in a study by Wigfield a nd Eccles (1989). Th ey asked junior high and high school students about their intentions to continue taking mathematics coursesâ€” an example of continuing motivation (Maehr, 1976). For junior high students, the only significant predictor of these intentions was belief in the intrinsic value of mathematics. In contrast, for high school st udents, both intrinsic value and utility value were significant predictors.
65 All studies described thus far used off-line measures of engagement to test the link between perceived usefulness of schoolwork and academic engagement. Yair (2000) used an on-line measure, the Experience Sampling Method (Csikszentmihalyi & Larson, 1987), in which students complete a form describing their thoughts and behaviors at random times during school when they are signaled by a beeper. Yair collected 4,000 sampled experiences from 900 students in grades 6, 8, 10, and 12. Academic engagement was measured as a nominal variable indicating what students were th inking about at the time they were beeped, which could include the task at hand or some external preoccupation. Logistic regr ession was used to estimate the effects on this simple measure of engagement of predictor variab les that included perc eived relevance (e.g., â€œHow important was the activity in relation to your future goals?â€) , degree of challenge, instructional method (e.g., lect ure or discussion), and dem ographic variables. Higher perceptions of relevance pred icted greater engagement (i.e ., thinking about the task at hand) and lower susceptibility to external preoccupati ons. In fact, the observed decline in engagement across grade levels was attenuate d when perceived relevance was taken into account, suggesting that courses ar e seen as less relevant at higher grade levels. A study using the same on-line measurement but a more elaborated engagement index (a composite measure of self-repo rted interest, enjoyment, a nd concentration) found similar results in high school students (Shernoff et al., 2003). Polls of high school students, open-ende d ethnographic research, and an array of correlational studies suggest that students are more motivated when they perceive the work they are doing as useful and relevant to the â€œreal world.â€ The applicability of this conclusion specifically to Se nior Project was supporte d by a yearlong participant-
66 observation study of one English class that had just adopted Senior Project (Combs, 1995). Two excerpts from this disserta tion are especially relevant here: The Senior Project, as originally planne d, appeared to addres s the needs of the teachers. They wanted a program that would keep the seniors busy, invested in school, and producing high quality work . The Project, however, did not seem to address studentsâ€™ needs. Perhaps this was because no one ever asked them what those needs were. . . . These seniors de manded that the work they were asked to do in high school have a clear and direct relationship to the work they would be asked to do in college. Because this need was never overtly discussed, the ways in which the Project might have addressed it remained unclear. In fact, many students thought that this new program was taking time away from tasks, such as essay writing, which would be more useful in prep aring them for academic life in college. (p. 220) In her final interview, Mary [a stude nt] told me those who did not see the connection between the Project and college did not see its value. She said, â€œEveryoneâ€™s like, â€˜Iâ€™m not going to need this.â€™ They saw it in the research paper, but everything else that we did, theyâ€™re just like, â€˜We â€™re not going to have to do this in college; this is stupidâ€™.â€ (p. 201) Although this study focused on a single class in one school, the message emerging from it is so consistent with the othe r evidence I have reviewed that its Senior Project-specific conclusions provide firm ground for my hypothesis. Effect of Autonomy on Senior Pr oject Engagement and Self-Efficacy Learners who are free to choose the activity that they find most interesting and productive are more likely to be thou ghtfully engaged in a task and show persistence and se lf-regulation. (Pedersen, 2003, p. 54) Deprive [students] of self-determinati on and you have likely deprived them of motivation. (Kohn, 1993, p. 11) These statements cogently express the idea that autonomy, â€œa sense of being choiceful in oneâ€™s actionsâ€ (Connell & Rya n, 1987, p. 5), should be positively linked with academic engagement. Autonomy is conceptu ally distinct from both utility value and self-efficacy. Thus, it is logically possible for seniors to have great flexibility in choosing their topics and directing the course of their work (high Au tonomy) and yet feel that the
67 project is not ultimately very useful (low Utility Value). Indeed, this conceptual possibility is empirically supported by th e work of Combs (1995) described in the previous section. In addition, it is logically possible for a student to feel autonomous (â€œI got to choose a topic I really likeâ€) and yet not feel especia lly efficacious about the work because of limited prior success with writi ng, researching, and speaking. However, it seems likely in practice that, other things be ing equal, greater autonomy will contribute to higher self-efficacy. The two hypothese s are highlighted in Figure 2-6. Senior Project Engagement Senior Project Self-Efficacy Advisor Support Peer Support Parent Support Clarity of Expectations Autonomy Utility Value Sensitivity Novelty Previous Mastery Experience Academic Engagement Figure 2-6.Model diagram highlighting the e ffects of Autonomy on Senior Project SelfEfficacy and Senior Project Engagement. Autonomy is a central concept in two of the guiding theories for this researchâ€” Newmannâ€™s authentic-pedagogy model and Se lf-Determination Theoryâ€”and in a wide range of other social-science accounts of motivation (e.g., Br onfenbrenner, 1979; Csikszentmihalyi, 1990; Kohn, 1993; Saras on, 1990). In unpack ing the idea of authenticity, Newmann and his colleagues (e.g., Newmann et al., 1992) included the notion of ownership over the conception, executi on, and evaluation of oneâ€™s work. This freedom to make consequential, independent decisions is the de finition of autonomy I adopted for this study. Research rooted in SDT adopts a more far-reaching definition,
68 one that includes not only choicefulness but also a sense that oneâ€™s feelings and perspective are respected and that the work one is expected to do is relevant to personal goals (e.g., Deci, Vallerand, Pelletier, & Ryan, 1991; Skinner & Belmont, 1993). Because these extra components are indicators of other constructs in the present modelâ€” Advisor Support in the first case and Utility Value in the s econdâ€”they are not part of the current definition. Choice and control are eagerly sought by high school students. For example, in Cushmanâ€™s (2003) focus groups with 40 U. S. high school students, participants repeatedly mentioned how little input they we re allowed in their classes and schools. One student stated the point th is way: â€œI felt like school was keeping me from learning. I wanted to read books I chose and do my own ar t, but you didnâ€™t have timeâ€ (p. 100). As part of another focus group study (York-Barr & Paulsen, 1997), three different groups in a single suburban high school met to discuss ways to improve curriculum and instruction in the school. One group consisted of student leaders, another comprised students with â€œnon-dominant perspectives,â€ and a third incl uded students with special needs. Across all three groups, it was common for students to request more individualized work that allowed them choices regarding topics and wo rk methods. Gifted students may have an especially strong desire for academic autonomy. Thematic analysis of interviews with ten 15-18-year-old students, all of whom were identified as gifted but were significantly underachieving, revealed that th eir major reasons for not investing much effort were related to lack of opportunities for choice and control over th e content, process, and pace of their learning (Kanevsky & Keighley, 2003).
69 Larger observational and interview studies paint a similar picture of high school studentsâ€™ eagerness for autonomy. Lowe (2003) , distilling 18 months of observations and interviews with students at a restructuring high school, concluded that â€œstudents desired a greater sense of autonomy and control over thei r lives in schoolâ€ (p. 87). Taking a novel approach, Hoffman (2002) surmised that yearboo ks would reflect what is meaningful to students, and she therefore c onducted document analysis and interviews with students on the yearbook staffs at five suburban high schools. A common theme in the interviews was a desire for independenceâ€”for making c hoices and working without supervision. When asked to talk about their classes, st udents in all five schools described the yearbook class as one of the few places where they could determine the direction, amount, and content of their work on a daily basis. In contrast, academic classes were seldom mentioned positively in the interviews, and they were seldom featured at all in the yearbook itself. Hoffman (2005) replicated the study in five rural high schools and obtained nearly identical result s. Overall, students were critical of school routines for making them feel immature. One senior girl , asked what made students feel â€œgrown upâ€ at school, described this feeling well: â€œHow about the things that make us not feel grown up?â€ (p. 72). The previous comment suggests that the c oncern with autonomy may be especially strong among high school seniors. The research and experience of several authors, in conjunction with theories of adolescence a nd young adulthood, reinforce this suggestion. For example, Hughes and Orr (1989) reported the results of a survey of one schoolâ€™s senior class. Asked â€œWhat can we do to better teach Armwood High Schoolâ€™s students?â€ the 295 seniors who responded (a 45% respons e rate) commonly mentioned that the
70 faculty should give seniors more credit for the level of maturity they had reached and therefore grant them more decision-making opportunities in school. Similarly, Kastner and Wyatt (2002), in a book describing Kastne râ€™s experience counseling seniors and their parents, noted that seniors â€œt ell me how much they want to call their own shotsâ€”and how entitled they feel to do soâ€ ( p. 49). Indeed, theorists have suggested that the key factor underlying a young personâ€™s subjective sense of atta ining adult status, at least in mobile and industrialized countries like the United St ates, is accepting responsibility for oneself and making independent decisions (Arnett, 2000; Arnstein, 1980). Studies have shown that demographic milestones, such as comp leting education, getting married, and having children, are far less salient than autonomy when young people define adulthood (Arnett, 1997, 1998). According to stage-environment fit theo ry, most schoolsâ€™ failure to provide increasing opportunities for m eaningful autonomy is a fundamental cause of the welldocumented decline in academic motivation between elementary to middle school (Eccles & Midgley, 1989; Eccles et al., 1993; Midgley & Feldlaufer, 1987). Perceived autonomy in school is indeed a strong predictor of general engagement for students in their early teens. Eccles, Early, Frasie r, Belansky, and McCarthy (1997) studied 1,300 students in grades 7 and 8, two-thirds of th em African American. They used multiple regression to assess the influence of aut onomy support from the school (e.g., items asking students how often they did projects of th eir own choosing and made decisions about important classroom issues) on a measure of general school engagement that included enjoyment, effort, perceived importance of academics, and behavioral problems in school. After controlling for peer influen ce and parental support at home, autonomy
71 support remained a highly significant pred ictor of general school engagement. Conversely, in another study (Va llerand et al., 1997), lack of autonomy was a significant predictor of the ultimate disengagementâ€”dropp ing out of school. A large sample of students in grades 9-10 at Cana dian public schools completed a survey about their school experiences, and one year later, the resear chers checked school records. Ratings of perceived teacher support for autonomy (e.g., â€œI feel that my teachers pressure me to do what they wantâ€â€”reverse-coded) were signif icantly lower for the dropouts than for the persistent students. Finally, Cusick ( 1973) concluded from a six-month ethnographic study of a public high school that many stude nts were responding to their lack of decision-making power in school by paying onl y minimal attention, instead channeling their energy into areas where freedom was great er, such as friendships and extracurricular activities. Like the findings of most cross-sectiona l or non-experimental research, the results described above are open to many alternative interpretations. Perh aps the most obvious is that a positive, engaged disposition towards schoolâ€”for reasons unrelated to autonomyâ€”causes students to respond more pos itively to questions about many elements of their school experience, autonomy among them . Another plausible explanation is that more-engaged students elicit greater autonomy support from teachers. Such interpretations could apply equally to fi ndings from the 2005 High School Survey of Student Engagement (HSSSE), which incl uded over 80,000 students from U.S. high schools. Forty-nine percent of respondents indicated that th ey had at least some input into classroom decisions. Those who strongly agreed that they had such input were far more likely than those who strongly disagreed to say that they have â€œworked harder than
72 they expected to work in schoolâ€ (59% vs. 26 %) and that they â€œtake pride in their school workâ€ (86% vs. 34%) (HSSSE, 2005b). Additional studies with adol escent and young adult learners address some of these alternative interpretations and augment the co llective weight of evidence indicating that autonomy enhances engagement. Two such studies are experiments. In one (Zuckerman, Porac, Lathin, Smith, & Deci, 1978), yoked pair s of undergraduates solved puzzles, with one member of each choosing which puzzles to work and how much time to devote to each, while the other member of the pair wo rked with the puzzles and time allotments chosen by the other. Although participants in the â€œchoiceâ€ and â€œno choiceâ€ conditions had the same tasks in the same time peri od, those who had freedom of choice were significantly more intere sted in working on such puzzles later. In another study, Reeve, Jang, Carrell, Jeon, and Barch (2004) presen ted a workshop to train ten high school teachers to provide greater support for student autonomy; an additional ten teachers comprised a delayed-treatment control condi tion. Subsequent classroom observations showed that teachers in the experiment al group did use more autonomy-supportive behaviors. Furthermore, these behaviors we re an even stronger predictor of observer ratings of studentsâ€™ collective engageme nt (e.g., attention, ve rbal participation, persistence, and positive emo tion) after the intervention than were pre-intervention engagement ratings. As compelling as these re sults are, they must be viewed cautiously in this particular context because of the broad definition of â€œautonomy supportâ€ adopted by Reeve et al. As indicated in the first pa ragraph of this subsect ion, studies grounded in Self-Determination Theory, as th is one was, generally adopt a definition of autonomy that
73 includes not only choice but also a sense that oneâ€™s feelings and perspective are respected and that the work one is expected to do is linked to the realiza tion of personal goals. Longitudinal studies by Epstein and her colleagues also provide support for a causal interpretation of the association between autonom y and engagement. Epstein and McPartland (1976) conducted a longitudinal study with over 4,000 students in grades 4, 5, 6, 8, and 11. Students reported their level of day-to-day engagement in academic work (e.g., â€œIn class, I often count the minutes till it endsâ€â€”reverse coded). Across all grade levels, students with low grades who reported frequent opportunities to â€œshow what I can doâ€ in class rated their engagement as hi gher than students earn ing high grades but having few such opportunities. In a two-y ear longitudinal study using an expanded version of the same sample, Epstein (1984) compared older and younger students on the proportion of variance in Year-2 engagement me asures that could be explained by Year-1 scores. Given that initial scores for older students represent the longest accumulation of environmental influences, it is not surprising that Epstein f ound greater stability for the older students. However, even with Year-1 engagement measures controlled for students in grades 8 and 11, Year-2 student ratings of their decision-making power in school was a significant positive predictor of Year-2 engagement. Correlational evidence addressing the c onnection between autonomy and higherorder cognitive engagement was provided by Ramsden and colleagues. In one study (Ramsden & Entwistle, 1981), approximately 2,000 British undergraduates from 66 academic departments completed a course-per ceptions survey and an â€œapproaches to studyingâ€ inventory. Using disc riminant function analysis, wh ich narrows a large set of variables to determine an optimal subset that can discriminate between different groups
74 of interest, Ramsden and Entwistle found that departments with the gr eatest proportion of students adopting a deep, meani ng-oriented approach to studyi ng were best predicted by a function that included high le vels of autonomy (e.g., â€œWe seem to be given a lot of choice in the work we have to doâ€). In cont rast, departments with th e greatest percentage of students adopting a â€œreproducing orientat ionâ€ to studying (focused on memorizing) were best categorized by a f unction that included low levels of autonomy. This study and its basic findings were replicated with a sample of 1,500 Australian high school seniors (Ramsden, Martin, & Bowden, 1989). The plausibility of a causal link from autonomy to academic engagement is enhanced by the existence of credible mech anisms. For example, given the option to make choices related to the conception and execution of their work, people are likely to select according to their personal interests. The commonsense claim that people are likely to be more engaged when they are working on something related to enduring or newfound interests has long been a staple of philosophical work in education (e.g., Dewey, 1913; Whitehead, 1929), and it continues to be a central part of psychological theories of motivation (e.g., Mayer, 1998; Pari s & Turner, 1995; Schiefele, 1991). There is abundant evidence that student s who have greater personal in terest in the content of a task are more motivated to persist and to pr ocess task information with greater care (for reviews, see Hidi, 1990; Le pper, 1988; Renninger, Hidi , & Krapp, 1992; Schiefele, 1991). Creativity also is likely to be great er. Spooner (2002) conducted semi-structured interviews with 13 high school seniors who were nominated as especially creative by teachers and peers. Analysis of the intervie w transcripts suggested that creativity tended to flourish in environments that permitted stude nt control of the pace of their learning.
75 Moreover, the majority of the seniors identifi ed independent study (a requirement at their school) as something that exercised their creativity. Senior Project is based on exactly this premise of respecting choice and taking advantage of the motivational power of personal interest. Because the projects emerge from the in terests, talents, and strengths of the students, and because they elicit their cu riosity, they have the potential to be intrinsically motivating and to deflect stude ntsâ€™ initial and lingering resistance to the hard work demanded by the projec t. (Darling-Hammond et al., 1995, p. 89) Newspaper reports about Senior Project ofte n contain testimonials from students along these lines. For example, one senior from Souhegan High School in New Hampshire said of Senior Project, â€œIt gave me a chance to do something I always wanted to do. Thereâ€™s no reason to slack on this. Kids get to c hoose what they want to pursueâ€ (Vaznis, 2005, p. B7). Notwithstanding such testimonials, th ere has been some sugge stion that the selfdetermination in Senior Project can be more illusory than real. Winters (2000) conducted an intensive case study of one schoolâ€™s Seni or Project program. Based on in-depth interviews with and observations of students, Winters concluded, â€œI am convinced that students, when given freedom to direct their ow n studies, are not sure if they really have complete freedom to do soâ€ (p. 142). Faculty and administrators were not micromanaging seniorsâ€™ work, but there were subtle indications that certain topics were off limits. In another case study describe d previously, Combs (1995) found more blatant micromanagement and subsequent disengagement from project work. When the seniors became unruly or complacent, the teach er being observed by Combs responded by creating detailed tasks and assignments and a ttaching strict due dates. What these findings highlight is the importance of s ubjective student perceptions, rather than â€œobjectiveâ€ program features, as determinants of engagement.
76 In addition to influencing engagement vi a personal interest, autonomy support may affect engagement through its effect on se lf-efficacy. As indicated earlier in the conceptual separation of autonomy from the ot her variables in the m odel, it certainly is possible for a student to select a topic of great personal inte rest (high Autonomy) and yet still have little confidence in pursuing that topic through wri ting and research (low Senior Project Self-Efficacy). Nevertheless, given a fundamental desire to feel and express competence, it seems reasonable to suppose th at students would make choices likely to bring them more success. Thus, self-efficacy for project work would be greater for students who feel that they have been able to make their own decisions throughout the project than for students who feel less self-determined. A small body of evidence supports this conjecture. Williams and Deci (1996) studied second-year medical students in a patient-interviewing course. Over th e five months of the class, perceived competence in interviewing skills increased only for students who felt that the skills were taught by a teacher who supported their au tonomy. Because Williams and Deciâ€™s definition of autonomy support was rooted in SDT, it included not only allowing choice but also understanding the stude ntâ€™s perspective (e.g., â€œMy in structor tried to understand how I saw things before suggesting a new wa y to do somethingâ€). Again, the findings must be viewed with due caution in the cont ext of the current defi nition of Autonomy. Caveats also apply to earlie r research by Deci and collea gues (Deci, Schwartz, Sheinman, & Ryan, 1981). Based on elementary teachers â€™ responses to vignettes, Deci et al. selected three teachers with high autonomy -support scores (indicating a preference for helping children to think for themselves to resolve problems) and three with low scores (indicating a controlling orientat ion). Students in these cl assrooms completed survey
77 measures of perceived cognitive competence in the first week of school and then again two months later. The items on the perceivedcompetence scale, more general than those on a self-efficacy scale, asked students about their confidence in such skills as understanding what they read, remembering ma terial in school, and solving problems. The children in autonomy-supportive classr ooms experienced larger increases in perceived academic competence during the firs t two months of the school year than did students in controlling classrooms. It is a refrain among high school students, es pecially seniors, that they do not have enough freedom in school to study the topics th ey like and to use thei r individual gifts to pursue work that is meaningful to them. By exploiting the motivational power of personal interests and greater self-efficacyâ€”w hich has been demonstrated across the wide variety of studies cited aboveâ€”support fo r Autonomy has the po tential to increase student engagement in Senior Project. Effect of Clarity of Expectations on Se nior Project Engagement and Self-Efficacy Researchers who study classroom assessment seem to agree unanimously that students are likely to be more motivated when expectations about th e nature, quality, and time frame of their work are clear (e.g., Ar ter & McTighe, 2001; Black & Wiliam, 1998; Natriello, 1987; Stiggins, 2001; Wiggins, 1989). This principle extends beyond education to the workplace, where clear de scriptions of expected performance and feedback on the adequacy of performance have long been considered key to employee productivity (e.g., Gilbert, 1978). An emphasi s on explicit expectati ons is not at odds with the importance of autonomy as an in fluence on engagement. Autonomy does not imply freedom without limits; instead, it m eans choice within parameters that give students clear guidelines for st ructuring their work. Thus, the hypotheses represented in
78 Figure 2-7 are that clear exp ectations contribute to Seni or Project Engagement both directly and through their effect on a studentâ€™s Senior Project Self-Efficacy. Senior Project Engagement Senior Project Self-Efficacy Advisor Support Peer Support Parent Support Clarity of Expectations Autonomy Utility Value Sensitivity Novelty Previous Mastery Experience Academic Engagement Figure 2-7.Model diagram highlighting the effe cts of Clarity of E xpectations on Senior Project Self-Efficacy and Se nior Project Engagement. By what mechanisms might clear expect ations enhance academic engagement and self-efficacy? First, clear parameters can help students avoid the misdirected effort and increased evaluation anxiety that accompany task uncertainty; students do not have to guess about what is important (Albrech t & Adelman, 1987; Arter & McTighe, 2001; Crooks, 1988; Csikszentmihalyi, 1997; Natr iello, 1987). Second, explicit guidelines from a teacher help students make the educat ionally desirable transition from external feedback to self-monitoring. Th is internalization, and the greater initiativ e it supports, are not possible when standards and expect ations are unclear (B utler, 1988; Sadler, 1989). Other mechanisms are evident when we consider specific means by which educators can promote clear expectations. Pr oviding feedback is one of those means. Formative assessment, appraising student work for the purpose of improving it rather than summarily judging it, is the increasingly popular way of referring to this kind of
79 feedback (e.g., Black & Wiliam, 1998; Tuns tall & Gipps, 1996). Done well, formative assessment conveys clear information about progress to date and provides suggestions for improving the work. Knowing that one has made progress and having information about how to advance even farther are ingredients for creating a greater fe eling of control and self-efficacy (Bandura, 1986; Locke & Latham, 1990; Rinne, 1998; Schunk, 1989, 1991; Schunk & Pajares, 2002). Another means of promoting clear expectations is dividing long-term goals into a series of smaller and more proximal subgoals (Bandura & Cervone, 1983; Bandura & Schunk, 1981; Brophy, 1987; Crooks, 1988; Pajares, 2002). Pajares stated the point well: Working toward long-term (distal) goals is a necessary ingredient of school life, but it can be tough on a studentâ€™s motivation. Short-term, proximal goals are more easily digestible for students, and they have the added benefit of raising selfefficacy. Not only do proximal goals make a task appear more manageable, but the more frequent feedback that is provided can convey a sense of mastery. (p. 121) In most cases, Senior Project is exactly th e kind of novel task in which feedback and subgoals can provide the clear task parameters needed to overcome initial uncertainty (Kuhlthau, 1985; Small & Arnone, 2000). For each of the reasons described above, clear expectations may promote task engagement. Evidence that they in fact do so has taken a variety of forms. Natriello (1984), for example, conducted 65 intervie ws with students in two suburban high schools. A common complaint of student s who had disengaged from school (as measured by absenteeism and self-reported cheating and low effort) was that they sometimes received unsatisfactory grades be cause they had misunderstood the criteria by which they were to be evaluated. Un fortunately, one of the causes of such misunderstanding, infrequent feedback, appear s to be common in high schools. About half of the nearly 81,000 high school students responding to the 2005 High School
80 Survey of Student Engagement said that th ey â€œneverâ€ or â€œonly sometimesâ€ received prompt feedback from teachers on their wor k. Furthermore, students who said they â€œfrequentlyâ€ received prompt feedback were more likely to report that th eir schoolwork made them â€œcurious to lear n other thingsâ€ (HSSSE, 2005a). Classroom-based studies reinforce the asso ciation between clear expectations and academic engagement. One small study compared four teachers of Advanced Placement (A.P.) United States History, two considered effective and two ineffective based on how their studentsâ€™ A.P. exam scores compar ed with the national average (Henderson, Winitzky, & Kauchak, 1996). Data on collectiv e engagement were gathered via eight unscheduled observations in each classroom, a nd data on feedback were collected from the observations and from student surveys. In the classrooms of the effective teachers, engagement rates were higher, as was the fre quency of assessment and feedback. A more formal quantitative study in high school cl assrooms was conducted by Thomas et al. (1993). Students from 12 biology classes in seven schools complete d preand post-unit surveys asking about the degree of initiativ e they took in their studying. Information about the amount of guidance and feedback provided by the teacher was collected via teacher interviews, classroom observations, and document analysis by the authors. Using hierarchical linear modeling, Thomas et al. found that more guidance and feedback (e.g., practice exercises and written comments on te sts) predicted greater change in selfreported initiative in studying. Finally, in a study described prev iously that included approximately 1,500 Australian high school senior s, Ramsden et al. (1989) administered a survey that included items re lated to the clarity of expectations in the school (e.g., â€œThe teachers here make it clear right from the star t what they require from usâ€). Additional
81 items addressed the depth of studentsâ€™ engageme nt in learningâ€”specifi cally, the extent of their genuine interest in learning and the extent of their higher-order cognitive processing of material (e.g., using elaborat ive strategies instead of just memorizing). The analysis revealed a modest but significant positive co rrelation between clarity of expectations and interest in learning ( r = .24) and also found that the items measuring clarity loaded factoranalytically with those measuring level of interest and higher-order processing. The association of clear exp ectations with engagement a pplies equally to adults in the workplace. For example, Wedman a nd Diggs (2001) conducted interviews with faculty in a teacher-education program to find out why the schoolâ€™s costly investment in technology had not produced much engagement in technology integration. A major reason given by the faculty was that expectati ons for technology use we re unclear, in part because faculty seldom received regular and helpful feedback about their efforts to use technology in the classroom. At a different university, th is study was replicated and extended to include not only teacher-educa tion faculty but also their students and cooperating teachers from the district. All thr ee groups identified unclear expectations as a primary reason for low engagement in tec hnology integration (Sch affer & Richardson, 2004). A much larger study (Dean, Dea n, & Rebalsky, 1996) posed the following question to over 800 people in different wor kplaces: Improvement in which one of the following areas would best enable you to do your job better? The forced-choice question was followed by six options, among them be tter resources, better incentives, more systematic training, and clearer expectations and feedback. The latter option was the most popular of the six choices in all different employee gr oups, ranging from a low of 29% to a high of 40%.
82 Although the cumulative weight of the preceding studies makes a strong case for the value of clear guide lines in increasing engagement, i ndividually the studies are open to the alternative interpretations to which corr elational and descriptive research are easily vulnerable. Experimental research makes the case stronger. Most of these studies have used elementary-school samples; caution is therefore warranted in generalizing to the older students who are the focus of the current study. However, one experiment (Morgan, 1985) used a sample of college st udents to test whether intrinsic motivation would be increased by one of the aforementione d sources of clear guid elines: larger goals broken down into subgoals. Approximately 200 students in an educational psychology class were divided into four groups: one r eceived training at the beginning of the semester in setting subgoals, one was taught to formulate long-range course goals, one received training in monitoring their time e xpenditure, and a control group had no special training. At the end of the course, the students in the s ubgoal-setting group rated their interest and engagement in the course as higher than did the other three groups. The kind of self-feedback exemplified in the subgoal condition of the preceding study is matched in importance by feedback fr om other people. Butler and Nisan (1986) asked 260 sixth-grade students to work on puzzl es (e.g., using the letters of a long word to form other, shorter words) under one of three conditions: feedback consisting of one sentence reminding students about the criteria for success (e.g., â€œThe words you wrote were correct, but you did not write many word sâ€), strictly numerical grades, or no feedback at all. In a sec ond session students worked on alternate versions of the tasks. Compared to students in the other two groups , students in the feedback group expressed greater interest in the tasks, a greater willingness to do them again, and a greater
83 willingness to attribute their success to effort and interest rather than to external factors (e.g., an easy task). Butler (1988) essentially replicated this study in a different sample of 132 students in grades 5-6. However, the sa mple was purposefully chosen to include students with high and low school achievem ent (based on a composite of report-card grades in mathematics and language arts). At both achievement leve ls, students receiving individual feedback expressed greater interest in the tasks and requested more additional tasks during a free period. Finally, Elaw ar and Corno (1985) trained one group of primary mathematics teachers to provide detailed written feedback on homework. A control group followed the usual practice of marking homework without comments. To control for the possibility that the training changed underlying teaching strategies, a third group was trained but then provide d half of their cl asses with full feedback and the other half with only grades. Those sixth-graders who received full feedback showed greater gains in their enjoyment and valuing of math ematics. Impressively, feedback accounted for 57% of the variance in post-treatment enjoyment and valuing, compared to 16% accounted for by pre-test attitudes. The experimental, correlational, and descrip tive studies cited thus far establish that engagement is more likely in environments that promote clear expectations and a clear understanding of how one is doing relative to those expectations. That higher selfefficacy is a primary mediator for this eff ect has been claimed often (e.g., Bandura, 1986, 1997; Brookhart, 1997), and there is evidence to support it. Kuhlthau (1985) observed 27 high school seniors while they were working on research pape rs for their A.P. English classes. She also conducted in -depth interviews with six of these students. Kuhlthau found that most students felt very uncertain a bout what was expected of them. Recalling
84 the earlier point that autonomy and clear expe ctations are not antagonistic, what the A.P. teachers viewed as freedom and trust, the students experienced as a lack of clear guidance. A much larger study, encompassi ng four U.S. high schools and approximately 1,700 participants, was conducted by Duckwo rth, Fielding, and Shaugnessey (1986). Survey measures of course-specific self-effi cacy (e.g., â€œIf I study hard for this class, the effort is rewardedâ€) correlated strongly w ith measures of course clarity (e.g., â€œI know what I am expected to be learning in this classâ€; r = .61) and measures of teacher feedback (e.g., â€œWhen I miss something on a te st in this class, the teacher gives me specific feedback about what I need to study againâ€; r = .53). As always, such zero-order correlations should be interpreted only as indi cations of a strong asso ciation between selfefficacy and clarity, not as proof that increases in clarity cause increases in self-efficacy. Another, more sophisticated correlationa l study tested a model of the effects of teacher behaviors on studentsâ€™ perceived competence, autonomy, relatedness and, ultimately, their academic engagement (T ucker et al., 2002). Approximately 100 African-American students in grades 1-12, all part of an after-school tutoring program, participated in the survey study. In the pa th analysis, the direct link from â€œteacher structureâ€ (measured by items related to clea r expectations and f eedback) to perceived competence was among the strongest paths in the model. In fact , no other variableâ€” including teacher warmth, teacher support fo r autonomy, grade level, or sexâ€”had as significant an effect on perceived competence. It is important to note, however, that the perceived-competence variable was measured not strictly as self-effi cacy. Rather, like Skinnerâ€™s concept of perceived control, it also included items related to locus of control and attributional style.
85 Experimental evidence for a causal link between clear expectations and selfefficacy also exists. Jamison (1993) starte d her dissertation with the observation, based on her experience as a teacher, that cert ain teacher behaviors reliably preceded breakdowns in student engage ment. Providing unclear inst ructions and inconsistent feedback were among these behaviors. Ja mison conducted an experiment in which college students learned how to tie complex knots under one of three types of tutelage: a facilitator who had received instruction only in knot-tying, a facilitator who was trained in knot-tying and some general principles of learning, or a facilita tor who had received training in knot-tying, decomposing tasks into smaller parts, and providing formative feedback. Participants in the latter group reported greater engagement; were rated as more attentive, self-initiati ng, and interested by an obser ver; and expressed greater feelings of self-efficacy (e.g., â€œFeedback from my facilitator helped me feel competentâ€). Two additional experiments were conducted with elementary students. Schunk and Swartz (1993) asked fifth-graders to rate their self-efficacy for five paragraph-writing tasks (e.g., deciding the main idea for a new paragraph) before an intervention. All participants then received writing-strategy instruction, but students in one condition received specific feedback about their use of the st rategies in a subsequent session. With pre-intervention efficacy ratings controlled, Schunk and Swartz found that students in the feedback condition judged thei r post-intervention self-efficacy for the writing tasks as higher than did students in the other conditions. In addition, mo re than two-thirds of the variance in post-intervention writing skill wa s accounted for by these later self-efficacy ratings. In a younger sample of 40 stude nts (ages 7-10 years), Bandura and Schunk (1981) studied the effects on motivation of breaking complex tasks into subgoals.
86 Participants engaged in self-directed learni ng of a mathematical process under one of three conditions: proximal subgoals were made e xplicit, distal goals were made explicit, or no goals were provided. Students in the proximal-subgoals condition demonstrated the greatest increase in self-efficacy and intrinsic interest in arithmetic, as well as the best performance at the end. In summary, studies using a variety of methods and samples suggest that providing people with clear expectations and information about how th ey are performing relative to those expectations offers motiva tional leverage. Part of this effect appears to operate via improved self-efficacy. Especially for larg e, extended assignments such as Senior Project, which can seem overwhelming to stude nts, explicit expect ations and frequent feedback can make the target appear much clearer (Blume nfeld et al., 1991; McLain, 2002; Pajares, 2002). Effect of Sensitivity on Senior Project Engagement The evidence described in Chapter 1 indict s the high school senior year as a period of minimal academic growth a nd a â€œfarewell tour of adol escence and schoolâ€ (National Commission on the High School Senior Year, 2001a, pp. 16-17). What this indictment obscures, however, is the high level of stre ss that many seniors e xperience. Recognizing this problem, one research team (Jason & Burows, 1983) actually developed and tested a â€œtransition trainingâ€ program for high sc hool seniors that included relaxation and cognitive restructuring activities. (See also Parker, 1992.) The contrast between this stress and what seniors expected of their la st year in high school has been a regular subject of commentary by people who study th e senior year. For example, based on focus groups and surveys in spring 1992, Zuke r (1997) concluded that most seniors felt they had little forewarning of how busy they would be with schoolwork, college
87 applications, and other responsibilities. Sizer (2002, 2003) found the same contrast between seniorsâ€™ expectations of relative ease and the more strenuous reality in her interviews with more than 150 seniors in 26 diverse U.S. high schools. A heavily involved senior in West Virginia, writing a guest editorial for her hometown newspaper, offered an especially telling testimonial: But now that I am here, I wonder why nobody ever warned me about the stress I am experiencing. I never expected to get here and cry my last year of high school away. This has been the most stressful year of my life. (Chapman, 2004, p. 6C) Such evidence suggests that students are more likely to be engaged in Senior Project if deadlines and additional structural features of the project are sensitive to other demands on seniorsâ€™ time and energy. Positiv e empirical correlations between Sensitivity and the other variables described thus far certainly seem likely. However, at least theoretically, Sensitivity is distinct from the other constructs. For example, a student may perceive that she has many choices regardi ng Senior Project (high Autonomy) and yet feel that the deadlines are unreasonable (low Sensitivity). Similarly, flexible deadlines (high Sensitivity) may coexist with vague exp ectations (low Clarity of Expectations). Regardless, with other things be ing equal, it is reasonable to conjecture that students who feel that the Project works with them rather than against them are likely to be more engaged. This hypothesis is represented by the highlighted path in Figure 2-8.
88 Senior Project Engagement Senior Project Self-Efficacy Advisor Support Peer Support Parent Support Clarity of Expectations Autonomy Utility Value Sensitivity Novelty Previous Mastery Experience Academic Engagement Figure 2-8.Model diagram highlighting the effect of Sensitivity on Senior Project Engagement. Among the major sources of additional stre ss for high school seniors is part-time employment. Part-time employment among American seniors is common. Data collected between 1985 and 1989 from a na tionally representative sample of 70,000 seniors indicated that 75% work ed part-time during the school y ear. Nearly half of these students worked 20 or more hours per week (Bachman & Schulenberg, 1993). Very similar results were obtained in a dataset compiled in the la te 1990s by the U.S. Bureau of Labor Statistics based on interviews w ith a national sample of 9,000 students (U.S. Department of Labor, 2005). Viewed alongs ide corresponding stat istics for other countries obtained from TIMSS and other studi es, these figures suggest that the United States is one of the only nations in whic h it is common for 17-year-olds to be both students and workers (Bracey, 2004; Larson & Verm a, 1999; National Research Council & Institute of Medicine, 1998). While it is true that seniors sometimes turn to jobs in order to obtain more â€œreal worldâ€ experience (Gilbert & Robins, 1998; Linton & Pollack, 1978), they often find a significant amount of stress both in the job it self and in balancing the job with school-
89 related work. Amen and Reglin (1992) admi nistered an open-ended survey to 95 seniors at an urban high school in North Carolina. Asked to list the t op five sources of stress in their lives, 60 students listed employment, noting (among other things) that working made it difficult to complete homework. G ilbert and Robins (1998) uncovered a similar theme in compiling and analyzing more than 100 essays submitted by students asked to write about what concerned them most in high school: Students who find out at 2:00 pm that they mu st read an entire chapter before class the next day, in addition to working to clos ing that night, will not be able to finish both tasks. Most often, it will be the sc hoolwork that will suffer. Advanced notice on assignments allows students the be st opportunity for getting both tasks completed. (p. 101) These same sources highlight an additional source of stress for high school seniors: applying to college. Stress can accrue at mu ltiple points of this process. One happens early, when students are writing essays and completing forms for admission and financial aid. In his ethnography of a highly-rega rded California high school, Humes (2003) described the panic that often attended the writing of college essays. This panic begins earlier for many students because of the increasing popularity of early-admissions programs (Pope, 2005). Aggravating the proble m further is the growing competition for places in many colleges and universities, which encourages students to treat the application process as a cont est to â€œsellâ€ themselves (K astner & Wyatt, 2002; Sizer, 2002). Still another stress-point is conflicts with parents ov er application deadlines and choices of places to apply (Mayher, 1998; Thompson, 1996). The stress does not necessarily end once the forms are complete d and the essays written. The nonchalance that is symptomatic of senioritis once college applications have been submitted may mask the anxiety that many students feel while wai ting to learn their fates (Sizer, 2002; Zuker, 1997). Indeed, Sizer (2003) suggested that senior itis is really more of a â€œsick leaveâ€ than
90 a restâ€”a point that was reinforced by a senior interviewed for a New York Times story (Lombardi, 2003): Emilie Curran, a senior at Horace Greel ey High School in Chappaqua, felt so impassioned about her right to slump that she wrote a column for her school newspaper in which she argued that parent s and teachers should cut seniors some slack. After the stressful college appli cation process, seniors needed time â€œto recuperate fully,â€ she noted. (p. 1) A small body of evidence supports the hypothesis that, when educators pay attention to these sources of stress in seniorsâ€™ lives, they are more likely to promote academic engagement. For example, Cushma n (2003) learned in her focus groups with 40 students in three major U.S. cities that student motivation was higher when teachers had asked certain questions before planning an activity. Among these questions were the following: Does the activity c onflict with studentsâ€™ physical needs or other priorities in and out of school? Have students had enough preparation time to do the task well? Seniors in the focus groups specifically me ntioned college-application deadlines as something that teachers should consider when deciding due dates for work. In another study, described previously to support the incl usion of Clarity of Expectations in the model, Natriello (1984) conducte d 65 in-depth interviews with students identified as disengaged from school. Being assigned too much work at one time was a commonly mentioned antecedent of disengagement. A group of approximately 8,500 seniors in the Montgomery County (MD) Public School Dist rict, surveyed in 2004, was asked about issues or experiences that had interfered with their high school education (Wimberly & Bernstein, 2005). Thirty-three percent mentioned activities outside of school, 31% referred to a job, and 28% mentioned family oblig ations. That students can identify such factors as interfering with th eir full engagement in academic work suggests that educators
91 may be able to increase engagement by â€œworking aroundâ€ such factors (without sacrificing rigorous expectations). Findings from two formal research studies reinforce the link between engagement and sensitivity to other demands on studentsâ€™ time and energy. Yair (2000) used the Experience Sampling Method to collect measures of in-class engagement for nearly 900 students in grades 6-12. As indicated earlier, engagement in this study was defined dichotomously as a match or mismatch betw een the content of the lesson and what the student was thinking about when beeped. The linear reduction in engagement with grade level virtually disappeared when time spent in external pursuitsâ€”i ncluding paid work, chores, and socializing with friendsâ€”was added to the prediction equation. Yair described these results as follows: â€œExternal nonschool factors are in a constant tug-ofwar with instructional charac teristics over studentsâ€™ attent ion in academic classesâ€ (p. 260). Another study described previously (M acIver et al., 1991) tested alternative models to explain changes in effort over the course of a semester for junior high and high school students. The independent variables were changes in intrinsic interest, in extrinsic pressures, in perceived utility value, and in self-concept of ability. The model explained more than twice as much variance in the juni or high sample (70%) as it did in the senior high sample (34%). The rese archers concluded as follows: Apparently the four psychological factor s assessed in the study [the four independent variables listed above] give a more complete picture of the determinants of effort on school tasks for the younger adolescents. Further research needs to examine additional fact ors that may influe nce older studentsâ€™ effort on school tasks. We suspect that competing activities such as athletic training, jobs, and peer rela tionships affect time spent on schoolwork for senior high school students more than for juni or high school students. (p. 208) In conclusion, high school seniors face an array of commitments and potential sources of stress that seem unlikely to change significantly in the near future. Students
92 who perceive that their Senior Project program is sensitiv e to these other demands are more likely to have both the time and the fa vorable disposition towards the project that would promote greater engagement. The ch allenge for program directors is being sensitive and flexible without falling into th e bargain or treaty of lowered expectations that so many educational commentators have noted (Boyer, 1983; Goodlad, 1984; McNeil, 1984; Sedlak et al., 1986; Sizer, 1984). Effect of Novelty on Senior Project Engagement [W]e are bored, not lazy. The reason why we are not doing our homework or studying for tests is not because we are re belling or because we donâ€™t think we have to. And really, itâ€™s not because we just want to go out and have a good time with our friends. . . . It all comes down to bor edom and the desire for something new. . . . The structure is the same, the peopl e are the same, and while it has all been worthwhile, we are ready for something new. Senior year shoul d be different than any other year for all students. (Cyr, 2001) Some students may disagree with parts of what this Maine high school senior wrote in a newspaper editorial. For example, some might argue that thei r lackadaisical work habits are partly an expression of rebellion. Howe ver, few are likely to contest Cyrâ€™s call for novelty in the senior year. For exampl e, based on three years of surveying and interviewing seniors particip ating in a dual-enrollment pr ogram, Rocklein (1994) noted that seniors often complain about the samene ss of their routines. Sizer (2002) reached a similar conclusion based on her interviews with over 150 seniors in 26 U.S. high schools. A senior caller to the Talk of the Nation radio program (Conan, 2005) echoed this feeling when he said, in favor of an internship pr ogram, â€œWell, it kind of gets you out of the school. I think itâ€™s quiteâ€”itâ€™ s so monotonous just going to the same school every day for, like, four yearsâ€ (p. 4). High school reformers generally agree with the assessment that â€œbusiness as usualâ€ is not the way to engage seniors (Lippard, 2000; Miller , 1972; National Commission on
93 the High School Senior Year, 2001a). For ex ample, based on his experience as the principal of a laboratory school, Miller noted over three decades ago that [s]ince most seniors have little appetite to be fed the same basic academic program and class pattern they experienced in the previous years, and since teachers are most unhappy to face disgruntled students da y after day, it should not require too much prodding to convince the faculty to break the senior slump by altering the traditional routine of classes. (p. 75) These words recall the previously de scribed recommendation of the National Commission on the High School Senior Year to â€œmove away from a system in which the senior year is just more of the sameâ€ ( p. 22). These observations from educators and students suggest that a Senior Project program that feels more novel should, other things being equal, promote greater engagement than a program that feels like â€œmore of the same.â€ This hypothesis is represented in Figure 2-9. A link between motivation and the experien ce of novelty is emphasized in several bodies of practice-based and theoretical lite rature. For example, instructional-design researchers have suggested that change and variety are essential ingredients of lesson plans that maintain student interest (Arnone, 2003; Keller, 1987; Kopp, 1982; Rinne, 1998). Psychologists adopting an organismic approach to human motivation (see Chapter 1) similarly have stressed the importance of nonroutine tasks for fully engaging people
94 Senior Project Engagement Senior Project Self-Efficacy Advisor Support Peer Support Parent Support Clarity of Expectations Autonomy Utility Value Sensitivity Novelty Previous Mastery Experience Academic Engagement Figure 2-9.Model diagram highlighting the effect of Novelty on Senior Project Engagement. (e.g., Csikszentmihalyi, 1990; Deci, 1975). W ithin educational psychology, novelty has been noted as a potential c ontributor to motivation a nd engagement (e.g., Brophy, 1987a, b; Lepper, 1988; Stipek, 2002). S tipek expressed the point clearly: Think about how exciting it is to do rou tine tasksâ€”for example, mowing the lawn, setting the table, or bala ncing your checkbook. The same kinds of tasks day after day at school are just as uni nteresting for children as routine tasks are for adults. Sometimes studentsâ€™ academic motivation ca n be increased by simply making sure that there is variety in the kinds of tasks they ar e asked to do. (p. 126) Perhaps the strong prima facie appeal of the claim that novelty enhances engagement partly explains why research on its motivational power has been limited. In one study already described, Kanevsky a nd Keighley (2003) conducted in-depth interviews with ten high school students who were identified as gifted and were underachieving. Thematic analysis of the interview transcripts led the authors to conclude that these students found school boring in part because they sought novel, openended experiences rather than the familiar, decontextualized tasks of school. In another study (Pederson, 2003), 66 sixth-graders were ex posed to a novel problem-based-learning (PBL) unit in their science classes. Student s completed surveys on their attitudes towards
95 science problem-solving both before and after the unit, and they al so participated in interviews after the unit. On the surveys students expressed great er interest, stronger preference for challenge, and greater dedica tion to â€œfiguring things outâ€ in the PBL environment. Asked during the interviews to explain their greater engagement, students mentioned the relative freedom and absence of te acher control, as well as the change in routine. The frequency of the latter respons e led Pedersen to ask whether most of the positive effects of PBL were due to novelty alone. Thus, â€œwould students who spend most of their educational careers in student -directed learning environments express more intrinsically motivated behavior ?â€ (p. 75). Other studies hint at the positive motivationa l consequences of novelty. In research described earlier, Sandberg (1981) collected longitudinal survey da ta on students in a Pennsylvania high school that recently had redesigned its junior-s enior curriculum to include more opportunities for independent a nd interdisciplinary st udy. Contrary to the trend in two comparison schoolsâ€”as in school s across the country (see Chapter 1)â€”most of the seniors in the rest ructured school expressed more positive attitudes about their work than they had the previous year. This trend did not hold, how ever, for students in the restructured school who already had part icipated in the rede signed curriculum as juniors. Sandberg suggested that, for thes e students, much of the engagement-promoting novelty of the curriculum may have worn off. Further evidence for the va lue of novelty comes from Hoffmanâ€™s (2002) study of students in the yearbook classes of five s uburban high schools. Noting that academic work seldom was mentioned in the focus groups as being especially memorable, Hoffman asked students for reasons. Some of their responses are worth quoting:
96 Maybe the reason no oneâ€™s talking about acad emics is because itâ€™s, like, something everybody has to do. And, like everybody knows about it so they donâ€™t care about it? So you tell the things that make you different, outside of school. (p. 34) Our academics are just an ordinary, like , everyday thing. Theyâ€™re what we do every day, and so weâ€™re pretty sure that nothing really spectacular will happen, that weâ€™ll really remember. They are th e same since first grade. (p. 34) For high school seniors, a signi ficant part of novelty is a feeling of distinctiveness, of special rights and responsibil ities granted to them by virtue of their advanced status. Senior honoring ceremonies, class trips, prom , and premium parking spaces appear to be insufficient (Kessler, 1999/2000; Zuker, 1997) . As indicated in Chapter 1, and as described so well by Dâ€™Orio (2002) below, the absence of such novelty is a primary trigger for the ennui of senioritis: Letâ€™s assume youâ€™ve had the same job for 12 years. In that time, youâ€™ve changed in numerous ways, the typical maturation that follows college, marriage, buying a house, and having children. But today, when you come into work, you realize that the job is still relatively unchanged from when you started, a nd it doesnâ€™t use the talents youâ€™ve developed in the intervening years. Worse yet, youâ€™re treated as if youâ€™re still 22, with bosses overseeing your every move and telling you whatâ€™s good for you. Chances are youâ€™d be bored. Now substitute school for job in the above example . . . [and] you have a good idea of what senior year feels like. (p. 27) Data supporting the desire for distinctiven ess among seniors comes from a survey I administered to 45 seniors in early spring at an independent, college-preparatory school in New Hampshire (Duff, 2004). Forty-seven pe rcent disagreed with the statement, â€œI am satisfied with the number of courses that are just for seniors.â€ Furthermore, the statement, â€œI would like to see more school-relat ed activities (e.g., field trips) that are just for seniors,â€ elicited strong agr eement from 75% of the class. As its name implies, Senior Project is just for seniors. This element of novelty and others have been mentioned as points of motivational leverage for seniors (DarlingHammond et al., 1995; Sizer, 2002). For ex ample, Darling-Hammond et al. wrote,
97 â€œStudentsâ€™ research often takes them beyond the confines of the school. They conduct research the way social science research ers do, and the authen ticity and nonroutine quality of the process impresses and excites themâ€ (p. 93). These remarks suggest that the nonroutine quality of Senior Project may be subsumed by some of the other variables already in the model. For example, Senior Project may be the first time that students have done work that they consider highly useful (Utility Value), and it may be the first time that they have enjoyed such freedom in pursuing a topic of their choice (Autonomy). However, Novelty also may be distinguishable from these other variables. At a school where project-based learning is comm on throughout high school (e.g., Lippard, 2000; Meier, 1995), not just in the senior year, Se nior Project might be characterized by high Utility Value and Autonomy but relatively low Novelty, which in turn might reduce the motivational potential of the project. Effect of Advisor Support on Senior Project Engagement and Self-Efficacy As dual-career, single-parent, and wh olly dysfunctional households have all increased, fewer adults know young people we ll. There is insufficient connective tissue among watchful, concerned adults and youth. So young people often fall through the cracks. They become invisi ble to grownups, influenced mainly by peers. In these circumstances, the role of teacher as caring adult assumes greater consequences. . . . Shocked by an adult who cares, a student pays that adult back by engaging in a learning ac tivity that otherwis e he or she might reject. (Powell, 1996, p. 56) Students working on Senior Project have an advisor. In line with Powellâ€™s commentary, it seems reasonable to conjecture that the amount and quality of support they receive from that adult should affect their engagement in the project. This hypothesis is firmly rooted in the conceptual framework for this disse rtation, as it draws upon the concept of school membership in Newmannâ€™s model (e.g., Newmann et al., 1992) and the concept of relatedness in Se lf-Determination Theory (e.g., Ryan & Deci,
98 2000). Relationships between teachers and st udents are a common element in many other accounts of student motivation (e.g., Dar ling-Hammond, Chajet, & Robertson, 1996; Erickson & Shultz, 1992; McQu illan, 1997; National Resear ch Council & Institute of Medicine, 2004). In fact, these sources and others believe that close student-teacher relationships are a major reason that stude nts in small schools generally have higher achievement, better attendance, and lower rate s of misbehavior and dropout than students in larger schools (Finn, Pannozzo, & Achi lles, 2003; Lee & Smith, 1995; Murphy, Beck, Crawford, Hodges, & McGaughy, 2001; Newmann, 1989; Toch, 2003). Several mechanisms might underlie the effect of Advisor Support on project engagement. As Powell (1996) suggested, students may feel a tacit quid pro quo to work hard when the adult has been highly supportiv e. Also, as sugges ted in Banduraâ€™s (1986, 1997) self-efficacy theory, when students know that someone supports them, they may feel more efficacious in their work, which in turn would promote engagement. Both this efficacy-mediated effect on engagement and a direct effect on engagement (more accurately, an effect mediated by unmodeled factors) are represented in Figure 2-10. Senior Project Engagement Senior Project Self-Efficacy Advisor Support Peer Support Parent Support Clarity of Expectations Autonomy Utility Value Sensitivity Novelty Previous Mastery Experience Academic Engagement Figure 2-10.Model diagram highlighting the effe cts of Advisor Support on Senior Project Self-Efficacy and Senior Project Engagement.
99 What does it mean for an advisor to be supportive? The â€œteacher supportâ€ construct has been operationali zed in different ways by diffe rent researchers, but some common denominators are evident (Ryan & Pa trick, 2001). They are instrumental help, encouragement, and emotional support. In strumental help includes communicating with students directly and regularly about their progress and sitti ng down with them to clarify material when necessary (Davidson & Phelan, 1999). Encouragement is conveying confidence in the studentâ€™s ability to su cceed (Hudley, Daoud, Polanco, Wright-Castro, & Hershberg, 2003). Alongside this validation should be warmth, the emotional support that students feel when thei r teachers are honest, fair, a nd interested in them as individuals (Firestone & Rosenblum, 1988; Hudley et al., 2003; Wentzel, 1997). Evidence linking these forms of teacher s upport to student academic engagement is abundant. Skinner and colleagues have provide d much of that evidence for students at the elementary level (e.g., Furrer & Skinner, 2003; Skinne r & Belmont, 1993; Skinner, Wellborn, & Connell, 1990). While intuition might suggest that teacher support would be less important for older students, with their increasing need for autonomy, the evidence suggests otherwise. Klem and C onnell (2004), for example, compared data on engagement and teacher support (e.g., â€œMy teacher cares about how I do in schoolâ€ and â€œMy teacher thinks what I say is importan tâ€) from approximately 2,000 elementary and 2,000 middle school students. High levels of teacher support were associated more strongly with self-reported engagement for th e middle school students. Further evidence for the link between teacher support and academ ic engagement in middle school students has emerged from studies using large, nationally representative crosssectional data sets (e.g., Voelkl, 1995, with an emphasis on emo tional support), as well as longitudinal
100 research using smaller, localized data (e.g., Murdock, Anderman, & Hodge, 2000, focusing on teacher encouragement; and We ntzel, 1997, emphasizing emotional support and instrumental aid from teachers). The importance of teacher support fo r young adolescents is not at all counterintuitive, according to Eccles, Midgl ey, and their colleagues (Eccles & Midgley, 1990; Eccles, Midgley, & Adler, 1984; Midgley , Feldlaufer, & Eccles, 1989). In their bid for autonomy, young adolescents primarily seek independence from their parents; therefore, nonparental adults, teachers prim ary among them, become especially important role models and sources of support during this time. Unfortunately, at precisely the time when such support becomes so vital, relati onships with teachers tend to become less personal. Several longitudinal studies of students transitioning from elementary to middle school have indicated that students perc eive their relationships with teachers as becoming less nurturing (Eccles et al., 1993; Fe ldlaufer, Midgley, & Eccles, 1988; Furrer & Skinner, 2003). These authors have suggested that the parallel be tween this trend and the documented decline in engagement afte r elementary school is not coincidental. Compared with middle school students, hi gh school students appear to be no less interested in, and no less affected by, st rong emotional and instrumental support and encouragement from their teachers. At the mo st basic level of engagement, studies have suggested that students who dr op out of high school may do so partly because they have poor relationships with their teachers (B earden, Spencer, & Moracco, 1989; Williams, 1987). Bearden et al. interviewed 400 hi gh school dropouts, and Williams compared responses of 50 graduates and 50 dropouts to a standardized interview. In both studies, students who left school before graduating re ported that their teachers generally did not
101 care about them and did not provide the s upport they needed. By themselves, such findings do not establish that low teacher support was a dropo ut-precipitati ng factor for these students. Perhaps both dropout and lo w teacher support were caused by a third factor, such as disruptive beha vior in the classroom and cons equent low grades. In the context of the evidence presented below, however, it would be difficult to dismiss low teacher support as merely an epiphenomenon, rather than a cause, of low engagement. Local and national polls and surveys provide some of this evidence. For example, in the Colorado High School Senior Survey administered to nearly 9,000 seniors (Colorado School-to-Career Partnership, 1999), students we re asked to indicate on a checklist which aspects of school motivated th em to learn. Seventy-nine percent checked off â€œwhen I like the teacher,â€ and 62% marked â€œwhen the teacher takes a personal interest in me.â€ Another local study (Roth & Damico, 1994) used interviews with approximately 200 sophomores and seniors from six Florida hi gh schools. Qualitative analysis of the transcripts suggested that stude nts were most likely to have trouble engaging with course material when they were unsure if the teacher cared about them personally. A national phone survey of 1,000 public hi gh school students conducted by Public Agenda (Johnson & Farkas, 1997) asked stude nts how many of their teachers (most, some, or very few) did each of the following: treated students with respect, personally cared about their students as people, and provi ded students with a lot of individual help. The percentages saying that most of thei r teachers did so were 41%, 30%, and 31%, respectively. Asked whether these teacher be haviors would help them learn a lot more, a little more, or no more, students responded as follows: 69% said they would learn a lot more with a respectful teacher, 64% said they would learn a lot more with teachers who
102 cared about them as people, and 69% said th ey would learn a lot mo re with teachers who provided individual help. Other large studies confirm that students place a premium on teacher support. In an 18-month ethnographic study of a restruct uring high school, Lowe (2003) drew conclusions based on interviews with student s and observations of their daily lives in school. Among the strongest themes was one related to teacher s upport: â€œStudent after student indicated a greater willingness to ta ckle hard work, and spoke of having far greater enjoyment in schooling in general, when the teacher at the front of the room showed interest and concern in each studentâ€™s successâ€ (p. 12 ). In a research project spanning 16 diverse secondary schools in the late 1980s and ea rly 1990s, McLaughlin and colleagues at the Center for Research on the Context of Teaching at Stanford collected interview and survey data from students (McLaughlin, 1994). One of their conclusions was related to teacher support: Students told us â€œthe way teachers treat you as a studentâ€”or as a person, actually,â€ counted more than any other factor in the school setting in determining their attachment to the school, their commitme nt to the schoolâ€™s goals and, by extension, the academic future they imagined for themselves. (p. 9) Finally, McLaughlin (1994) offered an inte rpretation in terms of student-teacher relationships of findings first reported by L ee and Smith (1994) a nd later published more formally in Lee and Smith (1995). Anal yzing longitudinal data from over 11,000 students enrolled in more than 800 high school s, Lee and Smith found that students in schools designated as â€œrestructuredâ€â€”meaning that they had in place three or more of 12 practices that the researcher s defined as significant de partures from conventional practiceâ€”showed greater gain s in engagement from grade 8 to grade 10 than other students. These results held even after taking into account school demographics and
103 degree of academic emphasis (measured by the average number of math and science courses taken by students at ea ch school in grades 9 and 10). Looking at the diverse practices defining â€œrestructure dâ€ schools, such as havi ng students keep the same homeroom throughout high sc hool and giving students oppo rtunities for independent study with faculty, McLaughlin (1994) noted a common denominator: all may contribute to a personalized school setting in which d eep student-teacher relationships can develop. The findings of these large studies are re inforced by small case studies. In the study by Kanevsky and Keighley (2003) that has been described twice, the most powerful theme in the interviews with ten gifte d, underachieving high sc hool studentsâ€”stronger than the themes of autonomy and novelty already mentionedâ€”was the value of caring teachers as catalysts for engagement. A similar interview study by Emerick (1992) identified strong relationships with teachers as one of six factors that moved ten gifted high school students from chronic underachieve ment to academic success. In another study outlined previously (Spoone r, 2002), data were collected via interviews with 13 high school seniors nominated as especially creative by their peers and teachers. Qualitative analysis of the interview transcript s revealed a theme: the teachers to whom these students responded best were those w ho cared about studentsâ€™ learning and their general well-being. Such themes are not limite d to gifted students. In a focus group of nine seniors convened by Education Week to discuss what did a nd did not work in their schools, much talk centered on the impor tance of teacher caring (Keller, 2003). One boy who transferred to the school in hi s senior year talks about a teacher who â€œwhen I first got here, helped me a lot. Everybody else was like, donâ€™t care, â€˜cause I came late. She was the only person who rea lly helped me to get my work done . . . by telling me to stay afte r class and get the work done an d all that. It was nice.â€ (p. 5)
104 Finally, in a case study of one student, a recent high school valedictorian wrote a commentary for Educational Leadership titled, â€œWhat Our Teachers Should Know and Be Able to Doâ€ (Belton, 1996). In line with all the evidence pres ented thus far, Belton asserted that, â€œIf students like a teacherâ€”regar dless of the grades they receiveâ€”they will want to perform to please the teacherâ€ (p. 67). As indicated earlier, such quid pro quo is only one possible mechanism for the effect of teacher support on academic engageme nt. Another is self-efficacy. One piece of evidence for this link comes from an et hnographic study of an English classroom in a low-income rural high school (Dillon, 1989). Based on field notes and interviews with the students, Dillon concluded that teach ers contribute to adolescentsâ€™ sense of competence when they convince students that they want the students to learn and that they care about the students as people. Several experimental studies with college undergraduates also document an effect of teacher encouragement, help, and caring on student self-efficacy. In one study (Tuckman & Sexton, 1991), students in an educational psychology class had an opportunity, each week for ten weeks, to earn bonus points by writing test items related to the course content. Each week half of the students received personalized and encouraging (though not partic ularly informative) feedback, such as â€œYour items show good imagination and clever ness.â€ The other half received neutral feedback using words such as â€œacceptable.â€ Ev en after controlling for initial course selfefficacy and student-rated importance of doi ng well in the course, Tuckman and Sexton found that students who received encouragi ng feedback had significantly higher course self-efficacy at the end of the ten weeks. In another study, 115 male undergraduates rated their interest in and self-efficacy for a physical task and then partic ipated in the task
105 under conditions of positive, negative, or no feedback (Vallerand & Reid, 1984). The feedback was bogus and was designed to be en couraging (or discourag ing) rather than highly informative about where one stood or how to improve (e.g., â€œIt looks like you have a natural ability, and it shows in your pe rformanceâ€). After cont rolling for pre-test levels of interest and self-efficacy, Vallerand and Reid found that postactivity levels of interest and perceived competence were hi ghest in the positivefeedback condition. Moreover, path analysis suggested that pe rceived competence mediated the effect of positive feedback on interest. Finally, Goodenow (1993) surveyed 350 stude nts in grades 6-8, asking them about their course-specific expectancies (e.g., â€œI e xpect to do very well in mathâ€) and valuing of the subject (e.g., â€œUnderst anding math is important to meâ€). These dependent variables were then regresse d on a number of independent measures, among them teacher support (e.g., â€œMy science teacher is interested in what I have to sayâ€), general sense of classroom belonging, and positive interactions with classmates. Teacher support was a significant predictor of expectancy of succe ss; in addition, unlike the other predictors, teacher support was an equally robust predicto r for both the sixthand eighth-graders in this middle-school cohort. Darling-Hammond et al. (1995) described one of the essential ingredients of Senior Project as â€œrelentless teacher encouragementâ€ (p. 92). The experimental, correlational, and descriptive studies reported above, each operationalizing teacher support and engagement slightly differently, neverthe less converge on the same conclusion. Having someone squarely in their corner makes students feel more efficacious, and it also encourages a form of â€œpaybackâ€ in which st udents feel a sense of obligation to invest
106 effort for that person. Other things being e qual, therefore, engagement should be greater for seniors who perceive their proj ect advisor as highly supportive. Effect of Peer Support on Senior Project Engagement Peers have great potential to support or undermine the engagement of classmates (Newmann, 1998). Indeed, one of the factors be lieved to exacerbate senioritis is its contagious nature (DeFao, 2005; Herring, 2001; Sizer, 2002). An apt description of the conditionâ€™s infectiousness was offered by a hi gh school senior who was an invited guest on an April 2005 airing of the Na tional Public Radio program, Talk of the Nation . Responding to a fellow seniorâ€™s suggestion th at senioritis is cont agious, the student replied: That I would say is one of the primary pr oblems with senioritis. You can go into the second semester of your senior year thinking, â€œIâ€™m going to keep up myâ€” maintain my average. Iâ€™m going to do fine.â€ And then you start seeing your friends, who have always been at the same level as you, begin slacking off and maybe not showing up to class a few times a week, and it does have a tremendous impact on your state of mind. (Conan, 2005, p. 6) The potential influence of peers on a seni orâ€™s state of mind is unsurprising given how much time and energy seniors in vest in their interactions w ith friends. To wit, in the 2005 High School Survey of Student E ngagement, which included nearly 81,000 students, 57% of seniors, compared with 43% of ninth-graders, re ported spending eight or more hours per week socializing (HSSSE , 2005b). A much smaller interview-based study with 75 seniors from five Chicago hi gh schools reinforces the emphasis on friends in the senior year (Linton & Pollack, 1978) . Although engaged enough academically to be in the top 15% of their respective se nior classes, these students nevertheless overwhelmingly identified the most rewarding aspect of school as spending time with friends.
107 Even for the most committed students, it is difficult to sustain interest and effort in school when friends and peers are enjoying th e â€œfarewell tour of adolescenceâ€ described by the National Commission on the High School Senior Year (2001a) based on its focus groups and other research. However, if a stude ntâ€™s friends enjoy Seni or Project and work diligently on it, then it seems likely that th e student would be relatively more engaged than if his or her friends reje cted the project. This hypothe sis is represented in Figure 211. Senior Project Engagement Senior Project Self-Efficacy Advisor Support Peer Support Parent Support Clarity of Expectations Autonomy Utility Value Sensitivity Novelty Previous Mastery Experience Academic Engagement Figure 2-11.Model diagram highlighting the e ffect of Peer Support on Senior Project Engagement. The decision to include a peer variable in this engagement model is based not only on the preceding considerations, but also on the conceptual framework for this study. Peer support enters most cl early in the guise of school membership described by Newmann (Newmann et al., 1992) and in the concept of relatedne ss detailed in SelfDetermination Theory (e.g., Ryan & Deci, 2000). It also appears in more detailed treatments of expectancy value theories (e.g., Wigfield & Tonks, 2002), as well as in other, independent accounts of academic motivation (e.g., Birman & Natriello, 1978; Wentzel, 1994).
108 Traditionally the influence of peers on academic engagement has been seen as mostly negative. In his classic study of the social ecology of ten Chicago-area high schools in the late 1950s, Colema n (1961) described an â€œadoles cent societyâ€ in which one of the keys to becoming and remaining popul ar with peers was to devalue academic achievement. In the two succeeding decades , Bronfenbrenner (1979) and Bishop (1989) reinforced this picture of the peer group as a pejorative influence on school motivation. Modern sociological and educational researcher s have tended to view these portrayals as too monolithic and simplistic, insisting instead that â€œthere is as much variation in values and aspirations within the youthful gene ration as between young people and adultsâ€ (Brown, 1990, p. 175; see also Brown, 1993; Entwistle, 1990). Thus, while there are indeed strong associations between the traits and behaviors of adol escents and those of their friends, sometimes those traits and behaviors are ones that actually promote academic engagement (Crosnoe, Erickson, & Dornbusch, 2002; Davies & Kandel, 1981; Epstein, 1983; Hartup, 1993). By themselves, the strong associations between the attitudes and behaviors of students and those of their friends do not provide clear evidence of peer influence . The similarity instead may indicate that students c hoose to associate with people already like themselves on dimensions such as school engagement (Brown, 1990). Perhaps more likely, the similarity between self and peers re flects the operation of both factors: students initially select friends who are like them, a nd the similarity increases over time due to the peer-influence mechanisms of modeling and norm-setting (Ryan, 2000; Warr, 1993). However, regardless of the exact reasons fo r the likeness between students and their close peers, simply knowing that degree of peer buy-in to Senior Project (or any other
109 academic endeavor) is a good predictor of a studentâ€™s own engagement would suggest valuable program improvements. Evidence for the potential contribution of peer support to academic engagement comes from a variety of large survey studies. In one of the earliest, Natriello and McDill (1986) used data collected in 1964-65 fr om approximately 12,000 students in 20 U.S. public high schools. Student e ffort was measured by self-r eported time spent daily on coursework outside of class, and peer support was measured (imprecisely) by a dichotomous item asking if the respondent thought it was important to earn good grades in order to be popular with oneâ€™s peers. Mu ltiple regression reveal ed that peer support had a stronger direct effect on homework effort than all variables in the model (which included educational characterist ics of parents and the studen tâ€™s academic track) except gender. In a survey study that included students from nine high schools in California and Wisconsin, Steinberg, Dornbusch, and Brown (1992) used items that directly asked students to rate how much their parents and fr iends encouraged them to succeed in school and to seek further education. Parental en couragement was the strongest predictor of long-term educational goals, but the encourag ement of peers had a stronger relation to studentsâ€™ daily school engage ment, including how much tim e they spent on homework, how they behaved in the classroom, and whet her they enjoyed coming to school. This study does not suggest that peers have no influence on the long-term educational plans of students. Indeed, Hallinan and Williams (1990) found in a sample of 20,000 dyadsâ€” each consisting of a senior and his or her best friendâ€”that the college plans of close friends are strong predictors of a studentâ€™s own intention to enroll in college. This
110 conclusion was based on a logis tic regression (the dependent variable being a simple yesno question about intention to enroll in college) using data from the 1980 High School and Beyond dataset. Studies using smaller samples have conve rged on similar conclusions about the potential of peer support to influence academ ic engagement. For example, in a study described previously, Rosenbaum (1998) surv eyed over 2,000 seniors in 12 Chicago high schools. School effort was measured by ite ms asking about time spent on homework and whether the student consistently performed up to his or her po tential in classes. Students also answered questions about the school-related at titudes and behaviors of their close peers, from which Rosenbaum computed indi ces of â€œrebellious p eersâ€ and â€œpro-school peers.â€ Both indices were significant predictors of effort, in the directions expected, even when the studentâ€™s perception of schoolâ€™s ut ility value was included in the regression. This finding, along with the othe rs reported thus far, are easily interpretable in terms of peerselection rather than peer influence. More compelling evidence for peer influe nce on academic engagement comes from a longitudinal study by Mounts a nd Steinberg (1995). For a sa mple of 500 students in grades 9-11, Mounts and Steinberg regressed studentsâ€™ current GPAs on their GPAs one year earlier and the GPAs of their closest friends one year earlier. They found that students with close friends who were successf ul in school improved their own GPAs over the one-year span more than students with ini tially comparable levels of achievement but less academically-oriented friends. The inference that the improved GPA was the result of greater engagement (not measured) seems reasonable given the preceding and following studies. Murdock et al. (2000) coll ected data on peer e ducational aspirations
111 (e.g., â€œMost of my good friends wonâ€™t drop outâ€) at grade 7 and used these data as predictors of grade-9 engagement (self-ratings of attendance, participation in class, and consistent completion of homework). Anal ysis of the data for the approximately 200 students in the study showed that, even afte r controlling for grade-7 academic selfconcept and standardized test achievement, peer aspirations in grade 7 were significant predictors of a studentâ€™s own acad emic engagement in grade 9. Finally, in a rare experiment al study on peer influence, Berndt, Laychak, and Park (1990) randomly assigned 118 students in grad e 8, each paired with a close friend, to one of two conditions. In the experimental group, pairs discussed dilemmas related to school motivation. For example, students discussed how to respond to a situation pitting an attractive music concert against the need to st ay home and study for an important test. Dyads in the control group discussed topics un related to school. Both before and after these discussions, all students made indepe ndent decisions about how they would respond to the dilemmas discussed in the experiment al group. Comparing preand post-responses to the scenarios, Berndt et al. found that only students in the experimental condition became more similar in their responses and that the shift in responses was not â€”contra Coleman (1961) and othersâ€”consistently in the direction of less achievement-oriented behavior. In addition, conten t analysis of the discussions suggested that it was not conformity pressure at play in the shift, but rather the exchange of information in dyads. In other words, peers influence each other in a variety of ways, not just through modeling and norm-setting but also through sharing info rmation and different perspectives with one another.
112 Using a variety of methods and measures of engagement and peer support, the preceding studies establish the potential of st udentsâ€™ close peers to undermine or promote their academic engagement. Determining whether this potential is reali zed in the case of Senior Project is one of the goals of this study. One in-depth, longitudinal study of a change process at a secondary boysâ€™ school in Australia suggests that harnessing the â€œpower of peersâ€ is an effective way to increase buy-in (Dowson & Cunneen, 1998). To promote a recent school restructuring effort , administrators identified especially influential members of the senior cohort (t he â€œmovers and shakersâ€ [p. 8]) and made special efforts to interest them in the program. This process even applies to students w ho, in the past, have used their social credibility amongst peers to undermine the schoolâ€™s academic and pastoral aims. There have been some spectacular successes at tributable, in part, to the nurturing of such studentsâ€™ egos by saying: â€œyouâ€™re a l eader,â€ rather than berating them about past misdemeanours. (p. 8) The school also invited past students to tell of their experiences in the new program. The lessons in this study may turn out to be rele vant to people who coor dinate Senior Project programs. Effect of Parent Support on Senior Project Engagement and Self-Efficacy Based on its focus groups with student s and parents around the country, the National Commission on the High School Senior Year (2001b; see also Miller, 2001) concluded that many parents inadvertently en courage their children to approach senior year less seriously. Among the ways that pare nts may have this effect are by allowing their children to work long hours and by show ing less overt interest in their schoolwork (often motivated by a respect for the seniorâ€™s growing independence). Students certainly may be actively engaged in school even wh en their parents are not highly involved; however, other things being equal, student s whose parents show greater support for
113 Senior Project are likely to be more engaged in the project. Represented in Figure 2-12, this hypothesis is grounded in self-efficacy theory and expectancy value models. According to these theories, parents can pr omote a feeling of competence in their children and provide other forms of support for greater engagement (e.g., rewards for good performance). Senior Project Engagement Senior Project Self-Efficacy Advisor Support Peer Support Parent Support Clarity of Expectations Autonomy Utility Value Sensitivity Novelty Previous Mastery Experience Academic Engagement Figure 2-12.Model diagram highlighting the e ffects of Parent Suppor t on Senior Project Self-Efficacy and Senior Project Engagement. Before reviewing evidence for the effect of parent support on academic engagement and self-efficacy, it is important to establish a working definition for Parent Support. There is substantial variability in the conceptualizati on and measurement of parent involvement (Hickman, Greenwood, & Miller, 1995). A widely used typology (Epstein, 1987, 1992) distinguish es home-based and school-bas ed parent involvement. School-based support, consisting primarily of attendance at school events (e.g., athletic events and parent-teacher conferences), is no t the focus of the study reported here. More relevant here is home-based involvement, wh ich Grolnick and Slowiaczek (1994) divided into two basic kinds: personal and intellectua l. Personal support includes encouraging and reinforcing school success, monitoring the childâ€™s progress in school, and talking
114 with the student about long-term educatio nal plans (see, e.g., Fehrmann, Keith, & Reimers, 1987; Muller, 1998). Intellectual s upport consists of activities such as providing cognitive stimulation (e.g., trips to museums) and helping students with homework. The Parent Support va riable in the current study is rooted in the intellectual and personal forms of parent involvement. So defined, Parent Support has been a focus of educational reform for decades, appearing prominently in the A Nation at Risk report (National Commission on Excellence in Education, 1983), more recently in the Goals 2000: Educate America Act (Carey & Farris, 1996), and peri odically in the general edu cation literature (e.g., Seeley, 1989; Walberg, 1984). In a 2000 Phi Delta Kappa mail survey of 2,000 K-12 public school teachers (25% of them high school teachers), the most common answer to the question of how to improve public schools was to involve parents more in the education of their children. This response also was the most frequent in the 1989 and 1996 versions of the survey (Langdon & Vesper, 2000). Because the authors did not disaggregate the results by grade level, it is impossible to know whether the data from the high school teachers alone followed this trend. However, what is clear from a variety of studies is that parents are significantly less involved pe rsonally and intellectually in the education of high school students than they are for st udents at earlier grade levels (Marks, 2000; McLain, 2002; National Research Council & Institute of Medicine, 2004; Simon, 2001; Steinberg, Lamborn, Dornbusch, & Darli ng, 1992; Walker & Hoover-Dempsey, 2001). The lower levels of parent involvement in high school are unsurprising. Older students generally demand more autonomy from their pare nts; students therefore may discourage overt parent support, and parents may feel equall y that their involvement is
115 less appropriate as their children progress through school (Carnegi e Corporation, 1989; Eccles & Harold, 1993; Muller, 1998). When students reach high school and the work becomes more cognitively sophisticated or esot eric, parents may feel less efficacious in helping their children (Eccles & Harol d, 1993, 1996; Muller, 1998). Nevertheless, parents, teachers, and adolescents may be assuming too much and thereby neglecting a valuable opportunity. Even for adolescents, parents tend to be more involved if they perceive that teachers and students want and expect th eir involvement (Deslandes & Bertrand, 2005; Hoover-Dempsey et al., 2001). For example, Deslandes and Bertrand surv eyed the parents of nearly 800 students in grades 7-9. They measured parent involve ment using items about the extent to which the parent encouraged the childâ€™s academic e fforts and helped the student study; they also measured parent perceptions of studentsâ€™ invitations for academic involvement (e.g., â€œMy adolescent has asked me to listen to him/her read something he/she wroteâ€). There was a decline in student invitations across the grade levels. Ho wever, at all grade levels, requests for help were the best predictor of parent support, accounting for more than 25% of the variance beyond that expl ained by the control variables (e.g., parent gender, family size). In a similar study comparing student s at grades 5, 8, and 11, Walker and HooverDempsey (2001) found that parent help w ith homework was unrelated to student invitations at grade 5 but highly relate d to student invitations at grade 11. By itself, showing that parents are more likely to be involve d in adolescentsâ€™ schooling when they perceive that such in tervention is welcome does not show that parent support actually hei ghtens student engagement. A variety of studies, however, suggest that it does. While parent involvem ent is not a strong predictor of high school
116 studentsâ€™ scores on standardized tests of achievement (Keith, Reimers, Fehrmann, Pottebaum, & Aubrey, 1986; Muller, 1998), it is a significant predictor of school gradesâ€”which are more sensitive than test sc ores to student motiva tion and effort (Keith et al., 1993). Clark (1983) studied ten African-American hi gh school students and their families in depth, asking what distinguished the families of academically successful students. The successful students had pare nts who adopted a styl e that Clark called â€œsponsored independence.â€ This style incl uded consistently monitoring schoolwork, encouraging academic pursuits, and requiring th at their children earn the privileges of independence. Taking a quantitative appro ach, Hickman et al. (1995) collected data using structured interviews with 47 mothers of high school students in one community. Achievement was measured by GPA, and di fferent forms of parent support were measured using questions about the number of hours spent in particul ar activities. Homebased parent involvement (measured by ques tions about help with schoolwork, course selection, and post-secondary plans) was th e only significant predictor of GPA in a regression analysis. Larger quantitative studies also indicate that parent support still matters for high school students. In a study with over 500 Ca nadian ninth-graders, Deslandes, Royer, Turcotte, and Bertrand (1997) re gressed year-end GPA on meas ures of parent support. Even when general parenting-style variables (e.g., degree of cont rol and warmth) were part of the model, â€œaffectiv e support for schoolâ€ was a significant predictor of grades, accounting for approximately 10% of the va riance. Exactly how Deslandes et al. operationalized â€œaffective support for schoolâ€ is unclear because they did not provide a complete listing of items used to measure i t; however, the sample item, â€œA parent gives
117 me encouragement about school,â€ conveys some idea of the constructâ€™s content. Finally, a study using data from more than 11,000 se niors who participated in the National Educational Longitudinal Study (NELS:88) found that greater parental involvement was associated with higher grades even after co ntrolling for race, family structure, gender, prior achievement, and socioeconomic stat us (Simon, 2001; see also Trusty, Watts, & Erdman, 1997). The study by Simon (2001) also found th at two engagement measures, attendance and degree of preparation for classes, were pos itively associated with parent involvement after controlling for the demogr aphic factors. These findings suggest that parent support may contribute to studentsâ€™ academic perfor mance partly through its effect on studentsâ€™ engagement. Additional evidence reinforces th is interpretation. For instance, Fehrmann et al. (1987) used data from over 28,000 high school seniors w ho participated in the 1980 High School and Beyond Longitudinal Study. Parent involvement was measured by student reports of the extent to which th eir parents keep track of their academic performance and talk with them about post-s econdary options. In a path analysis that included the background variables of intellect ual ability, ethnicity, gender, and SES, the path from parent involvement to time spen t on homework (a measure of engagement) was highly significant and meani ngful, second only to the eff ect of ability on homework investment. A similar link between parent support and time spent on homework was found by Trusty (1996) in an analysis of data from 9,000 seniors who participated in the 1992 cross-sectional sample of NELS:88. Parent involvementâ€”measured by student reports of how often their parents discussed school courses, grades, topics studied in class, and issues outside of school with the studentâ€”correlated significantly (though
118 modestly: r = .23) with time spent on home work. A similar correlation ( r = .25) was found between the parent support measure a nd a measure of positive attitudes about school (e.g., a belief that a ttending classes is important and that continuing oneâ€™s education is valuable)â€”attitudes that would support student engagement. Additional cross-sectional studies conve rge on similar conclusions. One that already has been described (Marks, 2000) included a measure of parent involvement (extent of parent-child discussion about current academic matters and postsecondary concerns) and a measure of academic engagement (student reports of effort, attentiveness, boredom, and homework comple tion). Marks used hierarchical linear modeling to analyze the data from approximately 4,000 student s in grades 5, 8, and 11 at 24 different schools. Parent involvement wa s a positive predictor of engagement at all three grade levels, with the largest coefficient ( = .19) found for the high school students. In another stud y, Gonzalez, Doan Holbein, and Quilter (2002) surveyed 200 students in grades 9 and 10, asking how involved their pare nts were with homework, course selection, and monitoring of academic progress. These items, which addressed home-based parent involvement, were comb ined with ones related to school-based involvement (i.e., attending extracurricular ac tivities); given this hybr idized measure, the findings should be viewed cautiously in the cu rrent context. The dependent variable was a measure of academic â€œmastery orientationâ€ (seeking challenge, being persistent, and experiencing satisfaction in school). In a mu ltiple regression including ethnicity, gender, and scores on a general parenting-style inve ntory, parent involvement was a significant positive predictor of mastery orientation.
119 The preceding findings from cross-secti onal studies certainly are open to noncausal interpretations. They al so are open to a reverse causal interpretation: students who are engaged for reasons other than parental involvement may effectively â€œdraw inâ€ their parents (Trusty, 1996). Neverthe less, longitudinal studies dem onstrate show that, indeed, parent involvement and encouragement are active ingredients cont ributing to student engagement. In a study by Steinberg, Lamborn, Dornbusch, and Darling (1992), approximately 6,000 high school students partic ipated in two rounds of data collection one year apart. In addition to measuring general parenting style (e.g., authoritative vs. permissive), Steinberg et al. measured â€œp arent involvementâ€ (e.g., help on homework, monitoring of academic progress) and â€œparen t encouragementâ€ (e.g., degree of verbal support for hard work). The dependent vari able was â€œschool engagement,â€ measured by student responses to items about effort a nd attention in class, extent of school misconduct, and degree of bonding to teachers. Even after controlling for year-1 scores on the engagement measure, Steinberg et al. found that year-1 aut horitative parenting was a strong positive predictor of year-2 engagement. Most important for the present argument, the results of a mediation analysis showed that this re lationship depended on the intervening influence of parent involvement and encouragement. That parent support for education can pr omote academic engagement is now clear, but it is not yet clear exactly how such support works. Hoover-Dempsey and Sandler (1995) offered a theoretical model to explai n how parent involvement and encouragement can enhance childrenâ€™s school outcomes. M echanisms of parental influence include modeling (i.e., behaving in wa ys that demonstrate that school-related activities are worthy of adult time and interest), providi ng reinforcement, and offering instrumental
120 help and verbal persuasion that contribute to a studentâ€™s sense of academic self-efficacy. This notion of parents as self-efficacy â€œsocia lizersâ€ for their children also appears in other, more general theories of academic motivation and engagement (e.g., Wigfield & Eccles, 2000; Zimmerman, 2000). The hypothe sized influence of parent support on childrenâ€™s self-efficacy may be stronger for younger students, who generally spend more time in the home and are more dependent on their parents (e.g., Jonson-Reid, Davis, Saunders, Williams, & Williams, 2005). Nevertheless, there is evidence that pare nts may continue to have an effect on adolescentsâ€™ self-efficacy. For example, Tu rner and Lapan (2002) asked 139 students in grades 7 and 8 to complete a survey in wh ich they rated 90 different occupations on the following dimensions: personal interest, self-e fficacy (i.e., confidence in being able to do that kind of job), and perceptions of whether their parents would support their participation in that kind of occupation. The 90 occupations were grouped into six categories according to â€œHolla nd typeâ€ (e.g., Conventional, Ar tistic, and Investigative; Holland, 1992). Not only was self-efficacy a sign ificant predictor of career interest, but parent support accounted for between 29% a nd 43% of the variance in career selfefficacy for each of the career types. A nother study (Ferry, Fouad, & Smith, 2000) used an academic rather than vocational criter ion with a sample of about 800 college undergraduates. In a path analysis, parent encouragement (e.g., â€œM y parents encouraged me to put a lot of effort into math in grad es 9-12â€) was a strong di rect predictor of high school math and science GPA. These grades , in turn, mediated a significant indirect effect of parent encouragement on math/science self-efficacy. Additional authors have reported moderate to strong positive corre lations between parent involvement in
121 education-facilitating beha viors (e.g., offering school-re lated encouragement and assistance with homework) and academic self-e fficacy of both early and late adolescents (Hoover-Dempsey & Sandler, 2005 ; Wettersten et al., 2005). Through their potential impact on student self-efficacy and other engagementpromoting beliefs and behaviors, parents can have significant impact on a Senior Project program. The importance of maximizing cooperation and minimizing antagonism between schools and parents has long been emphasized (e.g., Ep stein, 1986), and it appears to be especially important in the cas e of a major graduation requirement such as Senior Project (McLain, 2002). Indeed, stor ies about parents protesting and undermining Senior Project have been repor ted in newspapers. For example, a vocal group of parents in Washington State complained loudly when Se nior Projects were fi rst required in 1993. Some wore black armbands to graduation af ter three seniors were barred from the ceremony for plagiarizing parts of their Seni or Project papers; their argument was that cheating on any other assignment would have resulted only in an F, not a ban from graduation (Tomsho, 2005). In fall 2004 parent s in East Greenwich, Connecticut, packed school-board meetings to protes t the districtâ€™s plan to requ ire Senior Projectsâ€”a move to which the district leaders responded by ma king the project optiona l (Archer, 2005; see also Mitchell, 2005). Chapter Summary and Prospectus The purpose of this chapter was to explain and defend each path in the proposed model of Senior Project engagement. A re view of the literature on student motivation and engagement suggested that two major kinds of factors have great explanatory potential: social support variab les and task-feature variable s. The primary sources of social support for academic engagement are parents, peers, and teachers. Important
122 features of the task are studentsâ€™ perceptions of its usefulness, nove lty, clarity, respect for their autonomy, and sensitivity to other dema nds on their time and energy. Student selfefficacy for the project is at the heart of th e model, contributing directly to engagement and serving as a mediator for the effects of certain variables on engagement. The other variables in the model, Academic Engageme nt and Previous Mast ery Experience, are included primarily as â€œcontrolâ€ variables li kely to be important but less immediately manipulable by Senior Project coordinators . The goals of the following study are to determine whether the basic structure of the model captures the patter ns of variation in the data and to learn which variables are the most significant c ontributors to student engagement in Senior Project.
123 CHAPTER 3 RESEARCH DESIGN AND METHODS To test the model described in Chapter 2, I developed a survey to be completed by high school seniors participating in Senior Project. The first purpose of this chapter is to describe the development and administration of the survey. I then outline important characteristics of the partic ipating schools and students. The remainder of the chapter offers a detailed account of the analysis, incl uding the refinement of the scales used to measure the variables, the checking of statis tical assumptions, the execution of the path analysis and refinement of the model, and the cross-validation. Finally, I outline some limitations of the methodology. Instrument Content and Format of Survey The variables in the current model, su ch as Autonomy and Senior Project Engagement, are not directly observable or measurable. Whereas one can measure simple variables such as income and political party affiliation directly, or at least without much ambiguity, most psychological concepts present a greater measurement challenge. Each variable in the model was operationalized by a set of survey questions meant to gauge the degree to which a student experien ces or manifests the underlying variable. For instance, the nine items related to Seni or Project Engagement address some of the likely behavioral, cognitive, and emotional expressions of underlying engagement (e.g., â€œHow often do you feel proud of the work you do on the project?â€).
124 Decisions about item content were based on both the preceding review of literature and the distinctive features of Senior Project. For exampl e, Advisor Support (e.g., â€œI get a lot of encouragement from my faculty adviso r for Senior Projectâ€) is a version of the general teacher-support construc t (described in Chapter 2) tailored to the specifics of Senior Project. Items for the two engage ment variables (Academic Engagement and Senior Project Engagement) were written in parallel. Maintaining a consistent operationalization of engagement would have two desirable consequences. First, it would create the strongest possible control va riable, rendering all the more impressive any observed effects of other variables on Se nior Project Engagement. Second, keeping the item sets parallel would enable compar isons that might prove useful later. For example, the question of whether students ar e more engaged in Senior Project than in regular courseworkâ€”basically tantamount to wh ether Senior Project actually alleviates some of the symptoms of senioritisâ€”require s that both areas of engagement be defined similarly. Expert judgment provided additional conten t-related evidence for the validity of using the survey items to measure the intended variables. The expert reviewers were the four members of my supervis ory committee (two professors of educational psychology and two professors of curriculum and instruct ion) and a Senior Proj ect coordinator at a local, university-affiliated laboratory school. Ea ch reviewer received a list of the survey items, organized by the variables they were inte nded to measure, and they were asked to provide feedback on both the c ontent-appropriatene ss and readability of the items. This feedback was used to revise the list of items on the survey.
125 An additional technical note about survey content is warranted before addressing the formatting of the survey. All variables in the model except two were operationalized by four or more survey items meant to reflect the extent to which the student experiences the underlying construct. Such items are some times called â€œreflectiveâ€ indicators for this reason (Fornell & Bookstein, 1982). Because reflective items are intended to measure the same quality in slightly different ways, re sponses to them are expected to be highly correlated with each other (Jarvis, Mackenzi e, & Podsakoff, 2003) . The two variables measured differently were Previous Mastery Experience and Senior Project Self-Efficacy. For these variables, most of the items (see Appendix A) were not meant to reflect an underlying variable. Instead, they were intended to cumulatively form an overlying variableâ€”hence the term â€œf ormativeâ€ indicators (Fornell & Bookstein, 1982). For example, I conceived Senior Project Self-Effi cacy as a feeling of efficacy additively determined by a studentâ€™s level of confiden ce with the distinct processes of writing, researching, doing hands-on work, and managi ng time. Unlike reflective indicators, formative items need not be highly correlated with each other (Jarvis et al., 2003). It is not hard to imagine, for example, that a st udent might feel highly efficacious in writing but not in time management (see, e.g., Zi mmerman, Bandura, & Ma rtinez-Pons, 1992). Failing to distinguish reflective and formativ e measurement is increasingly seen as a problem for social science research (e.g., Bollen & Lennox, 1991; Diamantopolous & Winklhofer, 2001; Jarv is et al., 2003). The two variables with formative indicators were not measured solely by four formative items (see Appendix A). Each al so was measured by two general questions meant to reflect the cumulative impact of the variable (e.g., â€œOverall, how confident are
126 you about Senior Project?â€). Although these additional items turned out to be extremely important (for reasons I will describe), I in itially included them on the survey only for a technical reason related to st ructural equation modeling. St ructural equation models in which some variables are measured by fo rmative items often have a problem with identification (MacCallum & Browne, 1993), a tec hnical condition ensuring that the model-testing algorithms can converge on a unique set of parameter estimates (Schumacker & Lomax, 2004). The most theo retically appropriate solution to the problem of identification is to include two reflective indicators for a variable that is otherwise defined formatively (Jarvis et al., 2003). This approach is the one adopted here. Formatting choices related to items and re sponse options also were required. For each variable in the model, at least one of the survey items was written from a negative perspective. For example, one of the Seni or Project Engagement questions was, â€œHow often do you complain about the project to ot her people?â€ This standard survey-design practice is meant to induce respondents to pa y careful attention to each question instead of falling into an acquiescent response se t (Cronbach, 1950; Messi ck, 1962). Another formatting technique used to promote attenti on was to vary the response options. While some items required a rating on a 5-point sc ale from â€œStrongly Disa greeâ€ to â€œStrongly Agree,â€ others required a rati ng on a 5-point scale of frequenc y (â€œNeverâ€ to â€œAlwaysâ€) or intensity (e.g., â€œNot at all confidentâ€ to â€œV ery confidentâ€) (see Appendix A). Finally, in the formatted version of the survey (see Appe ndix B), items measuring the same variable were not arranged consecutively. Recall that most of the items were reflective, designed to measure the same underlying content in slig htly different ways. Given this overlap in
127 content and the associated potential for a frustrating sense of repetition for the respondent, it seemed prudent to stagger related items. Finally, to make the survey appear more official, I included the University of Florida wordmark next to the title of the survey (see Appendix B). A copy of the email granting permission to use the University wordmark is provided (see Appendix C). Pilot-Testing of Survey To assess the clarity of the surveyâ€™s language and formatting, I administered it individually to three high sc hool juniors recruited from the general-track English class at a local school. (See Appendix D for the protoc ol approved by the University of Florida Institutional Review Board.) The participantsâ€”a white female, an African-American female, and a Hispanic maleâ€”were chosen randomly from the ten juniors who volunteered to participate and obtained parental permission after hear ing a description of the pilot study. I started by giving each student a brief description of Senior Project and asking if he or she was familiar with the pr ogram. All had heard of it and knew its basic design. I then asked the student to imagin e working on the projec t this year and to answer the questions on that basis, record ing their answers on a â€œbubble sheet.â€ I encouraged the participants to make notes in the margins of the survey if they encountered any unfamiliar words or any instruc tions or questions that were not clear. When students completed the survey, which in all three cases took just under 15 minutes, I asked them to share their not es with me and to indicate whether any of the questions were offensive. The decision to pilot-test the survey with a small set of juniors having no experience with Senior Project requires justification. An ideal pilot test would have included a large number of seni ors actually working on Senior Project. For such a
128 sample, not only would the questions have been more immediately re levant, but the data would have had potential value in refi ning the survey psychometrically (e.g., by eliminating items with low variance). Howe ver, full-scale Senior Project programs are not yet common in Florida or elsewhere; give n the large sample-si ze requirements of the data analysis, I did not want to forfeit one of the accessible Senior Project schools. Nor was pre-testing the survey on a subset of the seniors at a Senior Project school a practical compromise, as it would have revealed the cont ent of the survey prior to the main study. Although the approach used was not ideal, it at least had the virtue of providing a conservative test of readability. The three students offered dire ct and indirect feedback a bout the survey. The direct feedback included suggested changes in word ing (e.g., from â€œevaluatedâ€ to â€œjudgedâ€) and recommendations about formatting (e.g., changi ng boldfaced words to fully capitalized words). Indirect feedback was provided when participants violated instructions. For example, one student bubbled-in her own na me instead of the schoolâ€™s name. This mistake prompted me to ask the student how I could change the instru ctions to reduce the likelihood of this problem. The student sugge sted bullets and capital letters in the instructions, which I subsequently incorporat ed into the survey. After reviewing these potential changes with the chair of my supervisory committee, I designed the final version of the survey (see Appendix B). Participants Recruitment of Schools I recruited schools for the study in two wa ys. First, in mid-August 2005, I posted a brief description of the study, along with a request for part icipation, on th e electronic bulletin board of the nationa l Senior Project Network. Operated by the Partnership for
129 Dynamic Learning, with content pr ovided by the Senior Project Center, this network is an online resource available by subscription for those wishing to share and receive information about best practices in Senior Project. When I first joined the network, I had requested a master list of all subscribing sc hools for the purpose of randomly selecting a sample of these schools. However, the director of the Senior Project Center was unable to provide this list (C. Osher, pers onal communication, August 15, 2005), and she recommended posting the request on the electronic bulletin board. This posting yielded only one response by the end of Septem ber 2005. I therefore began to initiate direct email contact with Senior Proj ect coordinators. I found the names and contact information for these coordinato rs in two ways. Firs t, I used a list of Senior Project schools that SERVE (see Chap ter 2) had published on its website in 2003, before it divested its Senior Project operations. From this list I randomly selected 15 schools. Second, because all of the schools on the SERVE list are in the southeastern United States, I broadened the geographi c scope of the sampling by conducting a Google search for schools with year-long Senior Projects. This search yielded an additional 15 schools. My first contact with the Senior Project coordinators at these 30 schools was a standard email in which I e xplained the purpose of the study and its time frame; noted the modest financial incentive for participating, namely, $40.00 to be earmarked for the schoolâ€™s Senior Project program; and, finally, as ked if the school would be interested in receiving further informati on. Twenty-five coordinators responded, with 20 expressing interest and five declining to participate.
130 For each of the 20 interested schools, I se nt formal letters to the principal to describe the study and request permission to co ntinue working with the schoolâ€™s project coordinator. Five principals declined to participate for various reasons (e.g., the timing of the study vis--vis other school events); the rema ining 15 agreed. After obtaining approval from the principals, I sent a formal letter to each project coordinator to notify him or her of this approval, describe more fully the procedures for administering the questionnaire, and request an estimate of the number of seniors that the school would be able to survey. Characteristics of Schools The 15 schools that agreed to particip ate are demographically diverse (see Appendix E for full information). For exampl e, they are located in 10 states; although most are in the eastern U.S., two are in the West. The schools range from majority white to majority African-American. Further, they vary from small to large (as measured by the student-teacher ratio) and from lowto high-poverty (as measured by the percentage of students receiving free or reduced lunch). Attrition of Schools Two schools that agreed to pa rticipate were not included in the final study (Schools 9 and 11 in Appendix E). In one case the proj ect coordinator informed me, shortly before the time designated for the survey, that the seniors had fallen behind schedule in Senior Project and therefore would be unable to answer most of the questions. In the other case the coordinator did not return the survey s until three weeks after the deadline. Final Sample Information Excluding these two schools, I sent ou t 2,478 surveys and received 1,334 complete or nearly complete answer sheets. During manual checking of each answer sheet (part of
131 the data-screening process detailed later), I found that 25 respondents had responded in a random or otherwise inappropriate mannerâ€” for example, by answering in nonsensical â€œzigzagâ€ patterns or by marking the same answ er for all questions. Therefore, the total number of useable responses was 1,309, re presenting an overall response rate of approximately 53%. The demographic character istics of the sample are described in Table 3-1. Table 3-1 Demographic characteris tics of the full sample (N = 1,309) Demographic variable Options % of sample Sex Male 42.9 Female 52.9 Did not answer 4.2 Race American Indian/Native American 1.0 Asian American/Pacific Islander 2.8 Black/African American 10.5 Hispanic/Latino 6.0 White 73.7 Did not answer 5.0 Post-secondary plans 4-year college 60.5 2-year college/trade school 23.4 Employment 2.3 Military 3.5 Undecided 5.6 Did not answer 4.7 (See Appendix F for subsample sizes and response rates by school.) Procedure Administration of Survey In late January 2006, particip ating Senior Project coor dinators received a box containing instructions and su rvey materials. The survey materials were the 4-page survey, a bubble sheet that stud ents would use to record th eir answers, and student and parent consent forms approved by the University of Florida Institutional Review Board. (For the IRB-approved protocol for the main study, see Appendix G.). All four of these items were enclosed in a 9 envelope, with one envelope for each student. Giving
132 each student a sealable envelope was mean t to ensure confidentiality and encourage honest responding. Project coordinators or th eir proxies (fellow t eachers) passed out the envelopes during a class, each time reading a sc ript that described the nature of the study, reinforced its voluntary nature, and provi ded instructions (s ee Appendix G). During the first two weeks of February 2006, students were asked to take the envelopes home, read the student consent form, and check off a box indicating whether they were or were not interested in the study. After receiving parent permission on the consent form, those students who wished to participate completed the survey at home and returned all materials in the sealed envelope to the survey administrator at the end of the week. (In early correspondence with the project coordinators, I learned that a take-home survey was favored over an in-school survey because it would be less disruptive to school routines.) Students who opted not to participate also were aske d to return the envelopes. All students who did so, includ ing those who did not complete the survey, received a $1.00 gift certificate to a fast-f ood restaurant to thank them for their cooperation. (Three schools did not allow student compensation.) Senior Project coordinators returned the envelopes using the boxes in wh ich the materials had arrived. Screening and Preparation of Data Before submitting the bubble sheets to the University of Florida Academic Technology office for scanning, I screened the data. As indicated earlier, I manually checked each form and set aside those in wh ich response patterns clearly suggested that the student had not taken the survey seriously. In addition, despite in structions at the top of the survey to use a number-2 pencil, four students used pen; I recoded their answers in pencil on new sheets. Forty students filled in the bubbles so carelessly (e.g., marking only a small part of the space or marking outsi de the bubble) that their forms had to be
133 redone. Fifty-two respondents violated the in struction to record the schoolâ€™s name, not their own names, on the bubble sheet. I eras ed each name, replacing it with the schoolâ€™s name. The University of Florida Academic Technology office scanned the bubble sheets and provided an ASCII data file. I imported the data into a spreadsheet and checked the resulting database for accuracy by compari ng a random sample of 20 bubble sheets with the corresponding spreadsheet row. All answ ers were accurately recorded. Finally, I added new columns to the database to accommodate answers for reverse-coded survey items. For example, for the item, â€œSenior Pr oject seems useless for preparing me for life after high school,â€ a response of 1 (â€œStrongly Disagreeâ€) was transformed into a score of 5 to indicate high standing on Utility Value. Data Analysis Overview A â€œmodernâ€ path analysis was used to test the path model described in Chapter 2. Such an analysis uses the algorithms of c ovariance-based structural equation modeling (SEM) to permit testing of many simultaneou s linear relationships among variables (Kline & Klammer, 2001). Testing the path model meant first assessing how well the overall model structure captured the patterns in the data and then determining which variables in the model showed the strongest links to self -efficacy and engagement. Because no study heretofore has examined the simultaneous influence of these variables on engagement, much less in the specific cont ext of Senior Project , I anticipated that changes to the model would be required. Some paths might need to be added and others removed from the model, and some variables might prove unimportant in the context of
134 the others. Even the stronge st literature review cannot anticipate fully how familiar variables will behave in novel research situations. This approach to path analysis is one that Jreskog (1993) called modelgenerating , in contrast to a strictly confirmatory appr oach, appropriate in we ll-developed research fields, and a model-comparison approach, suit able when testing competing accounts of a phenomenon. Great caution is warranted wh en using a model-generating approach. Most importantly, a model should never be m odified based solely on the results of a statistical analysis. Rather, each change should be theoretically defensible (Anderson & Gerbing, 1988; MacCallum, Roznowski, & Necowitz, 1992; McD onald & Ho, 2002). However, it is a common observation in the SE M literature that theoretical justifications for post hoc model modifications are too easy fo r creative and committed researchers (Hancock, 1999; Hox & Bechger, 1998; MacC allum et al., 1992; McDonald & Ho, 2002; Steiger, 1990). Hancockâ€™s (1999) warning on th is point is notable: â€œSlapping oneâ€™s own forehead after inspecting residuals or modi fication indexes [numbers that suggest which model changes would improve da ta-model fit] is not penanc e enough to pretend hindsight is foresightâ€ (p. 166). Moreover, even if the theoretical defense of model changes is solid, the process of â€œlistening to the dataâ€ is â€œinherently susceptible to capitalization on chance in that idiosyncratic characteristics of the sample may infl uence the particular modifications that are performe dâ€ (MacCallum et al., 1992, p. 491). How can researchers using a model-generating approach address these concerns? Cross-validation is among the most wi dely recommended strategies (Anderson & Gerbing, 1988; Breckler, 1990; Holbert & Stephenson, 2002; Hoyle & Panter, 1995; MacCallum et al., 1992). Ideally, data from a completely independent sample, collected
135 at a different time and place, should be used to test the generalizabil ity of the modified model. In practice, however, collecting new da ta is often infeasible (Whittaker, 2003). Thus, a common cross-validation strategy is to split an existing sample randomly, using one part to calibrate the model and the othe r part to validate the revised model (e.g., Draper & Smith, 1981; Floyd & Widaman, 1995; Snee, 1977; Trusty & Dooley-Dickey, 1993). Although the advice to cross-validate m odified models is common, it is rarely followed in practice, according to methods-foc used literature reviews in fields such as psychology (MacCallum et al., 1992) and co mmunications resear ch (Holbert & Stephenson, 2002). Not only was the sample divided for calibra tion and validation of the path model, but a fraction of the sample was reserved for refinement of the scales used to measure each model variable. By reducing the size of the sample used specifically for testing the path model, this further split of the sample was statistically conservative, reducing the power of the analysis to detect significant relationships. However, given this studyâ€™s use of untested measurement scales, th e decision to use part of the sample to refine the scales seemed prudent. Also, keeping the sampli ng fluctuations that influenced scale refinement separate from those that would influence assessment of the relationships among the scale-measured variables can increase confidence in the generalizability of the findings. Splitting the sample in the aforementioned ways had significant consequences for the model-testing given the lower-than-anticip ated total sample size. (Recall that the overall response rate was 53%.) The proposed model is complex, with more variables and paths between variables than most models tested using covariance-based SEM. The
136 large number of parameters was increased furt her by the use of formative indicators for two of the variables. Speci fically, per recommendations in the literature (MacCallum & Browne, 1993; MacKenzie, Podsakoff, & Jarv is, 2005), covariances should be estimated for each possible pair of formative indicators and for each pairing of a formative indicator with an exogenous latent variable. Added to the free parameters already in the path and measurement models, these extra parameters raised the total number of parameters to a level exceeding even the most liberal guideline s for the ratio of respondents to parameters needed for stable estimates (e.g., see Be ntler & Chou, 1987, who found that a 5:1 ratio was adequate for normally distributed data). In sum, the three-way split of the sample was unable to accommodate a full-scale SEM, in which both the measurement and structural components of the model could be tested. Moreover, even if the sample size had been sufficient for such an analysis, a full SEM might be too rigorous a test at th is early stage of model assessment and development. In such an analysis, the m easurement modelâ€”the relations between model variables and the set of indicator sâ€”is tested rigorously (via confirmatory factor analysis) either before or at the same time as the overlying path model (see Anderson & Gerbing, 1988). The problem with this approach in the present context was well expressed by Rigdon (2006a) in a posting to an SEM elec tronic discussion gr oup: â€œProducing good measurement models, with the strict constrai nts that we typically employ, takes multiple rounds of creating and refining items.â€ The multiple rounds of measurement refinement that would have established scales with clear convergent and discri minant validity were precluded by time, expense, and the aforem entioned problem of access to Senior Project schools.
137 I therefore adopted the less ri gorous but still powerful appr oach described earlier as a â€œmodernâ€ path analysis. Instead of representing each model variable using the more purified factors that would emerge from a full-scale SEM, I formed additive indices (Hair, Anderson, Tatham, & Black, 1998) and refined them using tools from classical reliability analysis (e.g., item-total correlati ons and coefficient alpha). Because such tools are inappropriate for use with forma tive indicators (Bollen, 1989), the change to a path analysis required the omission of the formative items. Thus, the two variables previously measured by four formative and two reflective indicators (Previous Mastery Experience and Senior Project Self-Effi cacy) were now represented only by their reflective items. Despite these unfortunate concessions to practicality over ri gor, modern path analysis has several advantages over the traditional path analysis stil l used commonly in psychology research (Kline & Klammer, 2001) . First, it uses â€œfull-informationâ€ algorithms to estimate parameters, analyz ing the entire system of equations simultaneously, rather than the â€œpartial-inform ationâ€ methods of estimation that consider only one dependent variable at a time. This approach generally yields more accurate estimates of the strengths of intervariabl e relationships (Holbert & Stephenson, 2002). Second, unlike traditional path analysis, modern methods allo w the researcher to adjust parameter estimates using the estimated measurement error from classical reliability analysis (Bollen, 1989). The eff ect of such adjustment is to remove some of the â€œnoiseâ€ from the variables, allowing true â€œsigna lsâ€ to be heard with greater fidelity. In the following sections I explain how I executed the path analysis. I begin by describing preliminary steps such as hand ling missing data, refining the measurement
138 scales, and checking the major assumptions as sociated with the covariance-based SEM algorithms. Along the way I present some preliminary descriptive statistics (e.g., multivariate normality) that were essentia l for making methodological decisions. In describing the testing of the path model, I provide basic explanatio ns of some of the major technical concepts (e.g., f it indices) so that readers can follow the basic logic even if they are not familiar with structural equation modeling. Imputation of Missing Data The problem of item nonresponse is a co mmon one (see, e.g., Schafer & Graham, 2002). There are a variety of reasons for nonres ponse, among them discomfort with item content, survey fatigue, and accidental omissi on due to survey formatting. Missing data was not a significant problem for the current da taset. Of the 65 items on the survey, only four had greater than 1% missingness. Th ree of these items were the demographic questions about gender (4.2% missing), race (5.0% missing), and post-secondary plans (4.7% missing). Given the potentially sensitiv e nature of the questi ons and their position at the end of the survey, these (still low) rates of missingness were not surprising. Moreover, because these items were used for descriptive purposes only, they did not affect the analysis. The only other item with a relatively high rate of missingness (4.7%) was item 10 (see Appendix B), which assessed level of self-efficacy for conducting research. The only plausible explanation for why students omitted this item is its position on the survey relative to the location of its corresponding row on the bubble sheet. Item 10 is at the beginning of a block of items on the survey (see Appendix B), but its row on the bubble form is at the end of a block, creating a visual di sconnect. Note that item 10 is a formative item; because I eliminated the fo rmative items from the analysis (for the
139 reasons cited earlier), the relatively high rate of nonres ponse on this item did not compromise the analysis. Although rates of missingness for other items were small, most items had at least some missing data. Traditional ways of d ealing with missing data include listwise deletion, which entails removing respondents wi th missing data from the entire analysis, and pairwise deletion, which preserves more of the data by computing each element of the input covariance matrix separately using all the cases that have complete data for that item (Enders, 2006). Strongly recommended over these methods are two options: either using modified SEM algorithms that work around the missing data or using multiple imputation to fill-in the missing values with statistically plausible values (e.g., Little & Rubin, 1989; Schafer & Graham, 2002). Imputa tion has a number of advantages. By producing a complete database, not only doe s imputation maintain the sample size to prevent loss of statistical pow er (Schafer & Graham, 2002), but it also makes possible the use of the valuable diagnostic and correct ive features (e.g., normality assessment and bootstrapping) of the software I used for the main analyses. I used a freeware package called NOR M (Schafer, 2000) to conduct a singleextraction version of the multiple-imputati on procedure for missing data replacement. (This technique is sometimes referred to as single-point multiple imputation; see, e.g., Irwin & Ziegler, 2005). To estimate the missi ng data, the procedure essentially uses a regression that predicts the missing value ba sed on responses from the same participant on variables correlated w ith the one in question (Cole et al., 2005). Split of Dataset Once I had a complete dataset, I used the SPSS 13.0 â€œselect casesâ€ command to split the 1,309 cases randomly into three approximately equal subsamples. The
140 refinement subsample, used to eliminate poor survey items from the scales measuring each variable, consisted of 436 cases. The same number of cases appeared in the calibration subsample, used for initial tes ting and modification of the path model. Finally, the validation subsample, used to test the generalizability of the modified model beyond the specific dataset used to generate it, included 437 cases. Refinement of Measurement Scales Following recommendations in the scaledevelopment literature (Clark & Watson, 1995; Cortina, 1993; Streiner, 2003), I examined two types of descrip tive statistics that address the psychometric quality of indi vidual items: alpha-if-item-removed and corrected item-total correlat ions. A description and ra tionale for each one follows. Alpha-if-item-removed gives the value of coefficient alpha ( ) for a scale when a particular item is eliminated from the s cale. Alpha provides a lower-bound estimate of the internal consistency (homogene ity) of a set of items; thus, if = .75, we can say that at least 75% of the total score variance is a ccounted for by systematic covariance of the items on the scale (Crocker & Algina, 1986). More intuitively, â€œspeaks directly to the ability of the clinician or researcher to inte rpret the composite score as a reflection of the testâ€™s itemsâ€ (Henson, 2001, p. 178). A higher generally supports gr eater confidence in such an interpretati on. If the value of increases significantly when an item is removed from a scale, it is a sign that the item did not correlate well with the other items intended to measure the same underlying variable. Th ere is no clear rule de fining a â€œsignificantâ€ change in alpha, and the disadvantage of rem oving an itemâ€”potential loss of important theoretical nuance in the vari ableâ€”must be weighed agains t the advantage of having a more internally consistent scale. For exam ple, removing an item that results in an increase in of .01 would seem to have little psychometric value and might omit a
141 valuable piece of the underlying variable. I therefore used alpha-if-item-removed in conjunction with theoretical c onsiderations, item-level descri ptive statistics (e.g., means and standard deviations; see Appendix H), and a sec ond item-quality statistic. The second statistic is the corrected item -total correlation, which is the Pearson correlation between the score on a given item and the sum of the scores on the other items measuring the same variable. One adva ntage of this statistic over alpha-if-itemremoved is related to the fact that is affected by the length of a scale (Streiner, 2003), with longer scales generally yielding higher values even when the scales are not unidimensional (Cortina, 1993). A corollary is that the alpha-if-ite m-removed statistic becomes less sensitive to change when there are many items on a scale. Because most of the scales on the survey have only four or fewer items, this potential problem is not serious. However, for the two longer scales (Academic Engagement and Senior Project Engagement), it seemed prudent to examine co rrected item-total correlations. Item-total correlations that exceeded a medium effect size ( r .30) were considered adequate (see Cohen, 1988). Full results of the initial reliability anal yses are provided in Appendix I. Based on the results, I modified five scales and elim inated one. The first modification applied to the 9-item scale measuring Academic Engageme nt. Two items on the scale had corrected item-total correlations under 0.30 and were associated with small increases in when they were individually remove d. One of these items asked how often the student â€œhas funâ€ in classes (Item 6). Although havi ng fun in oneâ€™s classes manifests an ideal level or type of engagement, enjoyment is secondary to the behavioral and cognitive commitment grounding my operational defini tion of engagement. The other poor item (Item 8R)
142 addressed how often the student â€œcomplain sâ€ about classes. Although griping about classes seems inconsistent with an engaged approach to schoolwork, such complaints may be a nearly inevitable symptom of seniori tis. Consistent with this interpretation, the mean score for this item ( M = 2.92; see Appendix H) was the lowest on its scale. Recall that the item sets measuring Acad emic Engagement and Senior Project Engagement were parallel by design. To ma intain this parallelism, it was necessary to remove from the Senior Project Engage ment scale those it ems (Items 21 and 22R) corresponding to those just removed from the Academic Engagement scale. The internal consistency of Senior Project Engagement was not unduly compromised by this change. In fact, even without these it ems, this scale was among the two most reliable ones on the survey (see Table 3-2). The reliability analysis also suggested re moving an item from the Autonomy scale. The item in question (Item 38R) asked respondent s to rate their level of agreement with the statement, â€œIt feels like t eachers have all the power in se tting rules and deadlines for Senior Project.â€ The corrected item-total co rrelation for this item (.185) fell short of the .30 guideline, and the removal of the item resulted in an increase in from .616 to .684. Why might this item have behaved differen tly than its scale-mates, which addressed seemingly similar issues (e.g., studentsâ€™ percep tion of their freedom in choosing a topic)? Both the mean and standard deviation for this item were low compared to the corresponding statistics for the other items on this scale, sugg esting that students felt little control over basic rules and deadlines for Senior Project but believed, to varying degrees, that they had more input into other decisions.
143 The scale for Sensitivity al so required modification. The corrected item-total correlation of .270 for one of the items (Item 52) was below.30, and the removal of this item yielded an increase in from .659 to .700. Responses to this item, â€œSenior Project deadlines are sensitive to the other demands on me as a senior (for example, college applications or a job),â€ did not correlate strongly with the othe rs in the scale. This result was surprising. Consider, for example, anothe r item (Item 49) that seems to address the same point in only a slightly different way: â€œThereâ€™s enough flexibility in the way Senior Project is set up so that I can still deal with it when my schedule gets crazy.â€ Perhaps the item in question was too vague (What does it mean for deadlines to be â€œsensitiveâ€?), leading respondents to answer it in a neutral fashion. S upport for this interpretation comes from the item-level statistics showing th at the mean score for the item was close to the neutral point and that its removal from th e scale resulted in the smallest decrease in overall scale variance. The final scale modification applied to U tility Value. Removing one of the items led to an increase in from .808 to .825. Although this itemâ€™s corrected correlation with the total exceeded .30, the value of .474 was small comp ared to all other item-total correlations in the scale (e.g., the next smalle st was .629). The item in question (Item 47) addressed studentsâ€™ level of agreement with this statemen t: â€œPeople outside school (for example, adults in the community) might actuall y be interested in my project work.â€ The other items on the scale asked respondents to what extent they believed Senior Project helped them do the following: work on skills th at are important in r eal world, reach some of their future goals, and prepare them for lif e after high school. Item 47 was different in that it asked students to infer something about other people rather than comment on
144 themselves and because it did not directly a ddress personal utility. I included the item because of its close connection with the con cept of authenticity described in Chapter 2 (e.g., Newmann et al., 1992). However, I do not believe that its elimination from the scale compromises the conceptual coherence of Utility Value. Finally, the scale measuring Novelty was eliminated entirely. No item had an itemtotal correlation exceeding .248 (see Appendix I). In addition, the scale reached its highest internal consistency of .418 with all four items includedâ€”an value significantly lower than those for the other final scales (see Table 3-2). The following statements, to which respondents indicated their level of agreement, comprised the scale: With Senior Project, I get a chance to do something new with my skills. (Item 40) Senior Project feels just like everything else we do in school. (Item 45R) Senior Project makes senior year feel different than last year. (Item 48) Even with Senior Project, my seni or year feels boring. (Item 53R) Based on the low reliability, it clearly is po ssible for students to be doing something new with their skills and yet not neces sarily feel that, overall, seni or year feels different as a result. Also, Senior Project may change the complexion of senior year with or without making it feel any more exciting. Removing th e Novelty variable from the model seems warranted given its poor operati onalization and the relatively sm all amount of theoretical and empirical support for its link to engagement (see Chapter 2). Mo reover, even if the concept were operationalized well, the effect s of novelty might be attenuated given the timing of the survey, which took place a pproximately halfway through the yearlong project. Some of the internal-consistency estima tes in Table 3-2 are lower than those typically seen in published literature. The re liability guidelines usually cited come from Nunnally (1967, 1978; Nunnally & Bernstein, 1 994), who recommended different criteria
145 depending on the purposes of the research. Fo r example, for applied settings in which important decisions will be made based on test scores, Nunnally cons istently set the bar Table 3-2 Number of items and reliabi lity estimates for measurement scales Variable No. items Alpha Academic Engagement 7 .768 Senior Project Engagement 7 .805 Advisor Support 4 .743 Peer Support 4 .689 Parent Support 4 .576 Clarity of Expectations 4 .637 Autonomy 3 .684 Sensitivity 3 .700 Utility Value 3 .825 Previous Mastery Experience 2 .681 Senior Project Self-Efficacy 2 .518 high, insisting that a reliability estimate of .90 is the minimu m that should be tolerated. In contrast, for scales used in the early stages of research (as here), Nunnally raised the bar with successive editions of his book. In 1967, he considered the range .50-.60 an acceptable minimum for preliminary researc h, but by 1978, Nunnally stated the minimum of .70 that would become the often-cited standard for research in the social sciences (see Lance, Butts, & Michels, 2006, for discussion). Six of the eleven scales on the survey ha d reliability estimates falling below the .70 guideline. However, there are several reasons not to be unduly concerned. First, as indicated earlier, is sensitive to the number of items on the scale. With only four items on most scales, it is not surpri sing that some of the reliabil ity estimatesâ€”especially those for the two scales with only two item sâ€”would fall short of the .70 standard.2 Second, as Streiner (2003) pointed out, it is a myth that â€œbigger is always betterâ€ when evaluating . 2 The Spearman-Brown prophecy formula (see Crocker & Algina, 1986) can be used to project the reliability of a scale when parallel items are added. If th e number of items in the least reliable scale (Senior Project Self-Efficacy) were do ubled from two to four, an value near the .70 standard would be obtained.
146 Especially with short scales like those in the present study, â€œ[h]igher values may reflect unnecessary duplication of cont ent across items and point more to redundancy than to homogeneityâ€ (p. 102). For these reasons, and because the measurement error of the scales would be accounted for in the analyses, all scales except Novelty were used in the study. For each respondent in the refinement subs ample, sum scores were formed for each variable. These scores were submitted to a Pearson correlation analysis, yielding a matrix of intervariable correlations (see Tabl e 3-3). The table shows that no bivariate correlations exceeded conventional threshol ds for collinearity (redundancy), which generally vary from r = .80 to r = .90 (e.g., Berry & Feldman, 1985; Kline, 2005; Tabachnik & Fidell, 2001). Therefore, the measured variables in the model showed adequate discriminant validity for use as separate constructs in the path model. Assumption Checks Linear relationships The standard algorithms for structural equation modeling assume that the relationships among the variables are lin ear (Schumacker & Lomax, 2004). It was therefore important to inspect scatterplots of the sum scores for each endogenous variable against the sum scores for each exogenous variable. The data from the calibration subsample were used for this purpose. Visual inspection of the scatterplots revealed no obvious nonlinear (e.g., quadratic) trends. In addition, all of the pairwise Pearson correlations between endogenous and exogenous variables (e.g., betw een Senior Project Engagement and Autonomy) were statistic ally significant (see Table 3-3). The relationships in the data appeared to be sufficiently linear for the purposes of SEM analysis.
147 Table 3-3 Correlations between model variab les in the calibration subsample (N = 436) Variable 2 3 4 5 6 7 8 9 10 11 1. Acad Engage .600 .326 .172 .253 .363 .268 .233 .242 .460 .274 2. SP Engage â€” .404 .416 .470 .551 .467 .445 .439 .354 .511 3. Advisor â€” .220 .270 .438 .306 .285 .252 .196 .234 4. Peer â€” .432 .398 .565 .581 .598 .002 .356 5. Parent â€” .473 .474 .465 .516 .083 .290 6. Clarity â€” .547 .557 .506 .227 .431 7. Autonomy â€” .665 .584 .107* .463 8. Sensitivity â€” .610 .071 .472 9. Utility â€” .038 .349 10. Prev Exper â€” .257 11. SP Efficacy â€” Note . Acad Engage = Academic Engagement, SP Engage = Senior Project Engagement, Advisor = Advisor Support, Peer = Peer S upport, Parent = Parent Support, Clarity = Clarity of Expectations, Utility = Utility Value, Prev Exper = Previous Mastery Experience, SP Efficacy = Seni or Project Self-Efficacy. All correlations except those with superscripts are significant at p < .001. *p < .05; p > .05; p > .10. Multivariate normality Maximum likelihood estimation, the default algorithm in structural equation modeling, assumes that the variables in the an alysis display joint multivariate normality. This condition requires that all univariate dist ributions be approximately normal (i.e., the data for each variable are di stributed along a standard bell -shaped curve) and that the joint distribution of any pair of variables be approximately bivariate normal (i.e., the normal distribution for variable x holds for all values of variable y ) (Kline, 2005). When data do not conform to this pattern of joint multivariate normality, there are potentially serious consequences for model-testing if appropriate adjustments are not made. Although parameter estimates themselves are rela tively robust to moderate violations of the normality assumption (Browne, 1984), standa rd errors for these estimates generally are not. Rather, standard errors tend to be underestimated when the data are non-normal;
148 this underestimation increases th e probability of a Type I e rror (i.e., claiming statistical significance when the evidence does not support it) when testing the significance of the parameter (Finney & DiStefano, 2006). Nonnormality also affects the chi-square 2 statistic used to test the overall fit of the model, making it mo re likely that a researcher will reject a correctly specified model (Anderson & Gerbing, 1988). To assess joint multivariate normality, three measures typically are used: univariate skew, univariate kurtosis, and multivariate kurtosis (Mardiaâ€™s normalized coefficient; Mardia, 1970). Skew refers to the degree of asymmetry in a distribution relative to the standard normal curve. Kurtosis refers to the â€œpeakednessâ€ of a di stribution relative to the same standard curve (Kline, 2005). There is no clear consensus on what values of these three indices indicate significant de partures from joint multivariate normality. Approximate guidelines for threshold values of these statistics are offered in the literature (e.g., Finney & DiStefano, 2006). However, th e most conservative approach is to consider the ratio of the skew or kurtosis statistic to its standa rd error (the â€œcritical ratioâ€) and compare the absolute value of this ratio to the standard critical value of 1.96. Values larger than 1.96 suggest statistically signifi cant departures from normality. To compute these three statistics for each variable in the calibration subsample, I used the SEM software package AMOS 6.0 (Arbuc kle, 2005) (see Table 3-4). In deciding whether adjustments for non-normality are warranted, the most important result is the statistic for multivariate kurtosis and its critical ratio (see last row of Table 3-4). For the present data, the large critical ratio i ndicated a significant deviation from joint multivariate normality. As indicated, failing to adjust for this deviation would compromise the testing of overa ll model fit and the testing of individual
149 parameter estimates. Several corrective methods are available (see, e.g., Finney & DiStefano, 2006; West, Finch, & Curran, 1995). Bootstrapping is one of these methods, and it is the one used here because it has been shown to be effective in controlling error rates for both model fit and parameter estimates in moderate to large sample sizes (Fouladi, 1998; Nevitt & Hancock, 2001). Table 3-4 Skew and kurtosis statistics for m odel variables in the calibration subsample Variable Skew Critical ratioskew Kurtosis Critical ratiokurtosis Autonomy -.174 -1.480 -.131 -.558 Clarity -.346 -2.948 -.219 -.931 Utility -.056 -.481 -.976 -4.160 Sensitivity .183 1.562 -.881 -3.755 Advisor -.752 -6.410 .342 1.458 Parent -.399 -3.400 -.352 -1.500 Peer .288 2.452 -.621 -2.648 Prev Exper -.495 -4.219 -.141 -.600 SP Efficacy -.137 -1.169 -.497 -2.119 Acad Engage -.408 -3.477 -.105 -.449 SP Engage -.496 -4.224 -.189 -.805 Multivariate 13.644 8.423 Note . Variable names are abbreviated as in previous tables. Cr itical ratios were computed by dividing each sta tistic by its standard error. The row labeled â€œMultivariateâ€ contains Mardiaâ€™s multivariate kurtosis coefficient. Bootstrapping is a process of resampling in which large subsamples are drawn from the parent sample with replacement (meaning that the same case may appear more than once in a given bootstrap subsample). For each of these subsamples, the statistic of interestâ€”be it the 2statistic of overall model fit or a particular parameter estimateâ€”is computed, forming an empirical sampling dist ribution. This data-b ased distribution, rather than the usual theoretical distributi on that assumes joint multivariate normality, is then used for calculations (Finney & DiSt efano, 2006). AMOS specifically uses the Bollen-Stine bootstrap (Bollen & Stine, 1992) to adjust the p -value associated with the obtained 2statistic. Thus, it does not adjust 2 itself, but instead corrects the statistical significance associated with the statistic. In contrast, the â€œnave bootstrapâ€ technique,
150 used to obtain adjusted standard errors for the parameters, directly adjusts the standard errors themselves (Finney & DiStefano, 2006) . Following the recommendation of Nevitt and Hancock (2001), I used 2,000 bootstrap samp les for both the Bollen-Stine and nave bootstrap adjustments. Independence of cases All variables in the present model were measured at the level of the individual student. None were school-lev el variables, such as school size, that would have shown between-school variance but no within-sc hool variance. Nevertheless, variables measured at an individual level may show school -level effects, such that students within a school tend to be more similar than student s from different sch ools. Such a finding would not be surprising. Not only are students at the same school partially â€œhomogenizedâ€ by their interaction and join t exposure to similar influences (e.g., teachers), but they also are likely to have at titude-relevant similarities associated with the residence-based system of public schoo ling (Kish, 1965). These within-school similarities would not be perf ect, of course. Students in the same classroom may have significantly different perceptions of the sa me â€œobjectiveâ€ conditions (see, e.g., Ames & Archer, 1988). However, it is necessary to a ssess whether the individual-level data show significant within-school correlation. This assessment is necessary because of one of the assumptions underlying the standard algorithms for struct ural equation modeling. Like mo st statistical tests, these algorithms assume that the data are derived from a simple random sample, meaning that cases (respondents) are independe nt of one another. Ideall y, then, two randomly chosen students from the same school should be no mo re alike than two students from different schools. However, for the reasons noted above, such complete independence seems
151 unlikely. Ignoring the data clustering and pro ceeding with standard st atistical approaches that assume independent cases can produce misleading statistical tests (Donner & Klar, 1994; Judd, McClelland, & Culhane, 1995; Kenny & La Voie, 1985; Kish, 1965). Specifically, failing to account for the clusteri ng leads to underestimat ed variances in the analyses, which tends to positively bias significance tests (i.e., make them less conservative). It is rare fo r researchers to check for and acknowledge this issue, much less make statistical adjustments for it (Judd et al., 1995; Kenny & La Voie, 1985). In fact, in the course of my rese arch, I read three studies in highly regarded journals that used structural equation mode ling but apparently did not ac count for potential clustering in the data (McWhirter, Bandalos, & H ackett, 1998; Randhawa, Beamer, & Lundberg, 1993; Vallerand et al., 1997; cf. Madon et al., 2001). The homogeneity of elements within a cluste r (school, in this case) is measured by the coefficient of intracla ss correlation (ICC) (Haggar d, 1958; Kish, 1965). Stated another way, the ICC measures the proportion of variance in a variable that is explained by cluster (class) membership rather than i ndividual-level variation. When cluster sizes are unequal, as they are in this study (see A ppendix F), the formula for the ICC (Haggard, 1958) is: WMS k BCMS WMS BCMS ICC ) 1 ( , where i i ik k k c k2) ( 1 1 The abbreviations BCMS and WMS stand for â€œbetween-class mean squareâ€ and â€œwithin mean square,â€ respect ively. These two numbers come from an analysis of variance (ANOVA), and they indicate the amount of the total variance due to school membership and individual differe nces, respectively. The number of clusters, in this case
152 13, is represented by c in the formula. The number of respondents in the ith cluster is represented by ik , and denotes the summation of all indi cated values over the entire set of clusters. Note that, if responses within a school are just as variable as responses between schools, then BCMS and WMS will be equal, and the ICC will equal zero (Donner & Klar, 1994), in which cas e the data can be treated as independent. Intraclass correlations significantly larg er than zero, however, sugge st that adjustments are necessary. The ICCs for the variables in th e model (see Table 3-5) indicate that the assumption of school-independent data was not tenable for most variables. Making technically appropriate adju stments for the non-independence of observations is difficult in stru ctural equation modeli ng. Recent versions of a flexible statistical modeling program called Mplus (L. Muthn & B. Muthn, 1998-2006) enable users to adjust standard errors and the chisquare test of model fit using its TYPE = COMPLEX command. (Samples wi th clusters are referred to as â€œcomplexâ€ samples.) Unfortunately, the number of clusters in my study (13) is sma ller than the number recommended by the software developers for obtaining highly stable corrected estimates (T. Asparouhov, personal communication, March 30, 2006; L. Muthn, 2006)â€” especially given the relatively small subs ample sizes created by splitting the sample. Thus, using the correction for clustering in the calibrati on and validation subsamples (each N 400) would not have worked well. However, according to one of the lead Mplus programmers, using TYPE = COMPLEX to test the model on the whole sample (N 1300), despite its suboptimal number of cl usters, is the best solution available and should produce â€œrelatively ok resultsâ€ (T. As parouhov, personal communication, March 30, 2006).
153 Table 3-5 Intraclass correlations for model variables Variable Intraclass correlation Autonomy 0.276 Clarity of Expectations 0.088 Utility Value 0.206 Sensitivity 0.297 Advisor Support 0.039 Parent Support 0.137 Peer Support 0.209 Previous Mastery Experience 0.023 Senior Project SelfEfficacy 0.097 Academic Engagement 0.027 Senior Project Engagement 0.094 Primer to Aid Interpretation of Model Analyses Thus far I have described the steps taken pr ior to model testing: imputation of data to create a complete dataset; division of the sample into three parts; refinement of the scales used to measure the variables; and assessment of the linearity, multivariate normality, and independence assumptions underlying the standard statistical procedures. Before describing the actual model testing, I o ffer brief prefatory remarks to help readers unfamiliar with structural equation modeling to understand the methods and results. Recall that the goals of the present study were to test th e overall fit of the model to the data and to assess the relative contribu tions of the variables in explaining Senior Project Self-Efficacy and Seni or Project Engagement. The parameter estimates that address the latter goal should not be interprete d until the model actually fits well. (SEM software will provide parameter estimates for poor-fitting models , but these estimates cannot be interpreted confidently .) The most stringent test of overall model fit is the chisquare 2 statistic described in the previous section. This statistic expresses the difference between the observed covariances in the data and the c ovariances predicted by
154 the model (Kline, 2005). A large differen ceâ€”that is, a statis tically significant 2 valueâ€”suggests that the model does not describe the patterns in the data well. Thus, contrary to the usual logic of statisti cal hypothesis-testing, researchers hope for non significant 2 values, which suggest that there is not strong evidence for rejecting the model. By itself, however, the chi-square test of overall model fit is problematic. First, research has shown that the 2 statistic is very sensitive to large sample sizes (Fan, Thompson, & Wang, 1999; Marsh, Balla, & Ma cDonald, 1988; see discussions in Bollen, 1989; Kline, 2005). In such samples, even small differences between observed and predicted covariances may lead to reject ion of the model. Second, because the 2 test assesses the hypothesis of exact fit between the model and the data, many researchers consider it too stringent given that m odels generally are intended only as approximations of real-world relations (e.g., Ben tler, 1990; Nevitt & Hancock, 2000). These problems with the 2 statistic have prompted methodologists to develop a large number of fit indices that are less sensitive to sample size and more tolerant of errors in model specification. Less ri gorous, these indices generally do not have associated significance tests; instead, they are judged according to guidelines proposed by different researchers. For this reason, a nd because different indices address slightly different questions, methodologi sts recommend that model f it be assessed by reviewing multiple fit indices (e.g., Schumacker & Lomax, 2004; Smith & Langfield-Smith, 2004). Below I describe the five i ndices that I used, along with the general guidelines for judging the level of model-data fit that they represent.
155 Two of the fit indices I used, the Comparative Fit Index (CFI; Bentler, 1990) and the Non-Normed Fit Index (NNFI; Ben tler & Bonett, 1980), assess the relative improvement in fit provided by the model compar ed to a baseline model. This baseline model is one in which the variables are a ssumed to be unrelated to the dependent variable(s) (Kline, 2005). In both cases, valu es approaching 1.0 indicate that the model in question describes the data better and better compared to the independence model. Though similar, CFI and NNFI are computed in different ways, and the NNFI â€œpenalizesâ€ models for their complexity (Kli ne, 2005). Especially during early stages of model development, the general guideline for interpreting CFI and NNFI is that values greater than .90 indicate reasonably good fit (Bentler & Bonett, 1980; Hatcher, 1994). A different fit index, the Root Mean Square Error of A pproximation (RMSEA; Steiger, 1990), essentia lly takes the opposite approach fr om CFI and NNFI. Instead of measuring the degree to which the proposed model is an improvement over an independence model, RMSEA provide s an estimate of the modelâ€™s error of approximation, doing so in a way that penalizes less parsimonious models (Kline, 2005). Because RMSEA measures error, values closer to zero indicate better fit of the model to the data. The general guideline is that values of RMSEA less than .08 indicate â€œreasonableâ€ errors of approxi mation, with â€œgoodâ€ fit sugge sted by values less than .05 (Browne & Cudeck, 1993; MacCal lum, Browne, & Sugawara, 1996). However, as in the case of CFI and NNFI, a slightly relaxed criter ion may be appropriate in early stages of model development. Steiger (1999), for example, suggested that RMSEA < .10 is appropriate in preliminary re search. SEM software genera lly provides a 90% confidence interval for the RMSEA. Therefore, it is possible to compare th e upper and lower bounds
156 of the interval to the guidelines to as sess the strength of ev idence for reasonable approximate fit. Ideally, the upper bound of the interval is lower than the guideline adopted for the study (Kline, 2005). Finally, I used the Goodness of Fit Index (GFI; Jreskog & Srbom, 1981), which is similar to the R2 in traditional regression analysis. Specifically, GFI estimates the proportion of observed covariances in the samp le covariance matrix that are explained by the covariance relationships im plied by the model (Kline, 2005) . Values closer to 1.0 indicate better fit. By convention, values of GFI greater than . 90 indicate acceptable fit (Kline, 2005), although other authors (e.g., Schumacker & Lomax, 2004) suggest a higher threshold of .95. Given that most fit indices are based on 2, it is natural to ask whether the adjustments made to 2 or its probability value when controlling for non-normal data (see earlier section on bootstrapping) can be inco rporated into the fit indices. Nevitt and Hancock (2000) showed that it is possible to make such an adjustment for the RMSEA index; the process is technically di fficult, however, especially when the 2 has a very low p -value. Fortunately, adjusting the fit i ndices for non-normality is unwarranted given the statistically conservative approach I adopted for this study. Any non-normality adjustments to fit indices would lead to more optimistic assessments of model-data fit (West et al., 1995); stated another way, obt aining acceptable fit indices without such adjustments provides even stronger ev idence of approximate model fit. Initial Model Testing Given the aforementioned problems with using Mplus for model-testing on the relatively small subsamples, and given the ease of iterative model modification with the
157 AMOS software, I used AMOS to conduct th e initial model testing. The variancecovariance matrix of the sum scores in th e calibration subsample (see Appendix J) was subjected to maximum-likelihood estimation, with the Bollen-Stin e (1992) bootstrap and nave bootstrap functions of AMOS used to adjust for multivariate non-normality. To provide a correction for measurement error, residual terms were introduced for each variable. The variances of these residual te rms were fixed according to the following formula: 2) 1 (i is , where i is coefficient alpha for variable i , and 2 is is the variance estimate for variable i in the calibration subs ample (Bollen, 1989). To evaluate and refine the model, I examined 2and the fit indices and considered the modification indices provi ded by the AMOS software. A modification index shows the minimum amount by which the 2 statistic would decrease if the parameter in question were freed rather than constr ained (Kline, 2005). (Recall that smaller 2 values indicate better fit of the model to the data.) Consider, for example, that there is no path in the model from Advisor Support to Clarity of E xpectations; in the anal ysis, then, this path is constrained to a value of zero. However, if the patterns in the data suggest that there is a directed relationship between Advisor Support and Clarity of Expectations, then AMOS would indicate so by noting the relationshi p and providing a corre sponding modification index. Notwithstanding some general guid elines in the literature (e.g., Jreskog & Srbom, 1986), there is no widely accepted criterion for deciding when a modification index is important, in part because the decision depends on the magnitude of the obtained 2 statistic. For example, a modification inde x of 10.0 merits closer scrutiny in a model with 2 = 40.0 than in a model with 2 = 400.0. However, regardless of the size of the
158 modification index, no decision about adding a free parameter to a model should be made without theoretical justification. Before finding a model with adequate approximate fit, I proceeded through multiple rounds of model modification and re-testing (described in the Results) using the data in the calibration subsample. As indicated earlier, such post hoc modification is potentially dangerous, even in the presence of what seem to be sensible theoretical reasons for the changes (e.g., MacCallum et al., 1992). The likelihood of capitalizing on chance patterns in the data is too high to affo rd much confidence in the findings of such a process. Therefore, once I found an accep table approximate model, I tested its generalizability to the third subsample. Cross-Validation of Model There are many possible approaches to crossvalidation. One option is to take the final model that was developed using the calibra tion subsample, apply it to the data in the validation sample, and compare the results (e .g., Floyd & Widaman, 1995). This form of validation, however, requires subjective interpretation of the degree of similarity in parameter values and fit statistics and i ndices across the two subsamples. A more rigorous approach to cross-validation is to use a multigroup test to examine the modelâ€™s configural and metric stab ility (invariance) when applied to both subsamples (Vandenberg & Lance, 2000). Configural stability means that the modelâ€™s basic configuration of free and fixed paths captures the patterns of data in the two subsamples equally well. The test of metric stability is stronger, assessing whether more specific features of the modelâ€”in this case, the magnitudes of the path coefficientsâ€”are essentially equal in the two subsamples.
159 AMOS provides a multigroup model test that makes it possible to conduct invariance assessments for cross-validation. The subsamples (calib ration and validation) are treated as two different groups, and the so ftware tests whether equality constraints across the groups yield a good-fitting multigr oup model. If the fit of the multigroup model is poor (as indicated by the same fit indices used for assessing a single-group model), it is an indication that the data in the subsamples are patterned differently enough that the equality constraints are like an ill-f itting straitjacket. Testing for configural stability is an important first step because it serves as a baseline: if the two subsamples do not show the same basic pattern of free and fixed paths, then there is no point in conducting the more stringent metric test of comparable magnitudes for these paths (Vandenberg & Lance, 2000). If both the configural and metric multigr oup models fit the data adequately, then the next step is to determine whether the f it of the more stringent (metric) model is nevertheless inferior to that of the base line configural model (Vandenberg & Lance, 2000). If the metric model does not fit significantly worse than the baseline model, then the cross-validation is an especially strong one. To compare the f it of the two multigroup models, I used three measures. One is th e difference in the Comparative Fit Index ( CFI) for the two models. Cheung and Rensvold (2002) recommended a maximum CFI of .02; that is, they cons idered models to have approxi mately equivalent fit if the CFIs differed by no more than .02. Second, I compared the RMSEA confidence intervals for the two multigroup models, looking for degree of overlap to assess the similarity of fit. Finally, per Vandenberg and Lance (2000) , I examined the difference in chi-square values for the two models 2 , which has an associated test of significance.
160 Model Stability with Cluster Adjustment The final step in the analysis was examining the impact of the cluster (complex) sampling on the fit of the final, cross-validated path model in the full sample. As indicated, AMOS cannot adjust for non-inde pendent cases; th erefore, I switched to the Mplus software. Unfortunately, the TYPE = COMPLEX mode in Mplus does not permit simultaneous use of an adjustment for multivariate non-normality (L. Muthn & B. Muthn, 1998-2006). Consequently, it was necessary to use standard (unadjusted) maximum-likelihood estimation on the variancecovariance matrix for the full sample (see Appendix K). Because of the inconsistenciesâ€”AMOS results were computed using a subsample (N 400) and a correction for non-normality, wh ereas Mplus results were computed using the full sample (N 1300) and a correction for clus teringâ€”it was not possible to conduct a rigorous comparison of the two. All that was possible was an informal comparison of fit indices and of the relati ve ranking of parameter estimates. Large differences along these two dimensions w ould suggest that the cluster sampling profoundly affected the configur al and metric patterns in th e data. In contrast, small differences along these dimensi ons would suggest that the cr oss-validated model captured the patterns in the data well even with the intraschool similarity on some of the model variables. Limitations of Methods Before examining the results of the model-testing, met hodological limitations should be acknowledged. I have noted severa l limitations already in this chapter; I review them here and add to the list so that the results may be viewed with appropriate caution.
161 One of the studyâ€™s limitations is the psyc hometric quality of the survey. Because there are few published tools to measure acad emic engagement (Anderson et al., 2004), and because no study heretofore has examined engagement in the context of Senior Project, it was necessary to create a new instrument. The competing demands of accommodating the large number of variables in the model and writing a survey whose length would not dissuade students from par ticipating (Dillman, 1978) meant that I could not measure each variable with a large number of items. In addition, the large samplesize requirements of the main study and the asso ciated expense of recruiting participants meant that I was unable to conduct extensive pilot-testing. Considering all of these factors, the relatively low internal-consistency estimates that emerged from the reliability analysis were not surprising. Although the modern path analysis can adjust for the effects of this measurement er ror, there is nevertheless a fa ir amount of â€œnoiseâ€ in the data that may complicate the analysis and interpretation. This point was expressed vividly by Martin (1987): Careful operationalization of major constructs and collec tion of high-quality data are the sine qua non of any substantive re search enterprise. No methodologies, no matter how powerful, allow researcher s to study phenomena they have not measured well. One cannot make a silk purse out of a sowâ€™s ear, no matter how powerful the sewing machine. (p. 36) Even if the variables in the model were measured with near-perfect reliability, validity would remain an issue. The present study used self-report scales only. What students report about their feelings and beha viors may not reflect accurately what would be observed if one had direct access to th eir thoughts and behaviors (Campbell, 1982; Kline, Sulsky, & Rever-Moriyama, 2000). One po ssible reason is social desirability bias (e.g., Paulhus, 1984), which may lead participants to answer questions in ways that they believe will gain approval from others, including the researcher. The agree-disagree
162 format used for most of the scales in th is study not only increase s the probability of socially desirable responding, but it also tends to provide less accurate and precise mapping of an underlying construct than w ould a construct-specific scale (Tay, 2005). The counterbalancing problem with constructspecific scalesâ€”indeed the reason I did not use themâ€”is that they tend to break up the â€œflowâ€ of the survey for respondents, especially when the survey measures many different variables. Another validity concern about self-report scales is that respondents may form hypotheses about the underlying logic of the survey and respond accordingly rather than reporting accurately on their feelings and beha viors. In addition, self-reports can be confounded by stable personality traits (e.g ., De Jonge & Slaets, 2005) and temporary moods at the time of testing. Ideally, self-report measures in this study would have been supplemented by teacher reports about each stud ent and direct observations of studentsâ€™ inand out-of-school habits in relation to Senior Projec t (see, e.g.,. Assor & Connell, 1992). Such measures were not feasible, however, given the aim of studying a wide cross-section of students involved with Senior Project. The hope was that a combination of several factorsâ€”including the brevity of the survey, the incentives for participation, and the multiple steps taken to ensure anonymityâ€”would mitigate these problems by encouraging students to respond thoughtfully. Along with the relatively low reliability of the scales, additional characteristics of the data contributed to a complex and less-than -ideal analysis. As shown earlier, the data were neither multivariate normal (see Table 3-4) nor free of cluster-related dependence (see Table 3-5), and I was unable to correct for both data-quality problems in the same analysis. As a result, there was more unfiltered â€œnoiseâ€ in the data than was optimal.
163 However, in light of evidence that much published research neither adjusts for nor even acknowledges these problems (Judd et al., 1995; Smith & Langfield -Smith, 2004), this study represents a good-faith effort to address both issues. The present study also addre sses the aforementioned concern about the infrequent use of cross-validation to assess the stability of models (e.g., MacC allum et al., 1992). Among studies that do attempt cross-validation, the split-sample technique adopted here is common for the reasons previously not ed (e.g., expense and convenience). Nevertheless, it should be acknowledged that the ideal validation sample is one in which the data are collected completely independently from the data in the calibration sample. Such an approach lessens the likelihood of capitalizing on chance occurrences associated with particular times and places of data collection (Whittaker, 2003). If all of the above limitations were eliminated in the pr esent study, the soundness of its conclusions certainly would be enhanced. However, the cross-sectional design of the study is inherently limited in the ex tent to which it can support clear causal inferences, even when the underlying analysis is as pow erful as structural equation modeling (e.g., Smith & Langfield-Smith, 2004). Baumgartner and Homburg (1996) sugg ested that it is more accurate to think of stru ctural models tested on crosssectional data as a set of relationships that can be viewed simultaneou slyâ€”certainly consistent with causal links but open to alternative explan ations that can be ruled out only through the use of longitudinal studies and controlled experiments. Thus, even if the analysis were to show (for example) a strong and positive direct path from Autonomy to Senior Project Engagement, we could not conclude definitiv ely that an increase in decision-making power itself causes greater engagement. The causation may run the other direction, such
164 that students who are more engaged are allowe d greater autonomy by their supervisors, or changes in both variables may be caused by a third, unmeasured variable. Self-efficacy, in particular, is ideally measured before students perform the relevant activities, thereby positioning it better as a variable causally related to engagement (Bandura, 1997; Zimmerman, 2000) Finally, given the preliminar y nature of the present m odel-testing, the use of structural equation modeling, even in a less rigorous path-analytic form (using sum scores instead of scores deri ved from factor analysis), has limitations. In this â€œmodernâ€ path analysis, the fit of the entire model to the observed covariances takes precedence over the interpretation of indi vidual effects or the expl anation of variance in the dependent variable(s) (Kline, 2005). However, in an analysis with a large number of variables, especially in the early stages of research, it is difficult to obtain good fit. Moreover, accounting for variance in Senior Project Self-Efficacy and Senior Project Engagement was an important goal of the pres ent research. These considerations suggest that Partial Least Squares (PLS) analysis (e.g., Chin, 1998) might have been a viable alternative to the covariancebased structural equation modeling used in the present study. PLS analyzes variances rather than covariances, focuses on prediction rather than overall fit, and makes few demands on the data (e.g., multivariate normality). However, unlike the present approach, PLS does not al low the analyst to account for measurement error, and its parameter estimates tend to be biased, in part because its algorithms do not consider all relationships simultaneously (Haenlin & Kaplan, 2004). Also, because PLS is used far less often in behavioral resear ch than is covariance-based SEM, obtaining support during model-testing would have been far more difficult.
165 Chapter Summary The focus of this chapter was the instru ment, procedure, and methods of data analysis used to test the model of Senior Pr oject engagement devel oped in Chapter 2. A survey was administered to over 1,000 seni ors by project coordinators at 13 schools across the country. One part of the sample was used to refine the measurement scales, eliminating items that were not measuring th e underlying variables well. The data were then submitted to a path analysis powe red by the algorithms of covariance-based structural equation modeling. Part of the sample was used for the initial testing and modification of the model, and the remainder of the sample was used for a preliminary cross-validation of the refined model. Finally, I tested the modelâ€™s stability in the face of an adjustment for non-independent cases, wh ich resulted from sampling students within schools rather than randomly.
166 CHAPTER 4 RESULTS A number of descriptive statistics, incl uding intraclass correlations, intervariable correlations, and Mardiaâ€™s coefficient, we re presented already because they were essential for making decisions about methodol ogy (see Chapter 3). The first purpose of this chapter is to present additional descriptiv e statistics that are usef ul given the practical and theoretical goals of this study. Interp retations of these st atistics will provide a foundation for decisions made during the infere ntial analysis. Desc ribing this analysis, which tests the proposed model of Senior Pr oject engagement, is the second purpose of this chapter. The results of the split-sample cross-validation, assessing how well the final model generalizes to a dataset not used in its creation, follow. Finally, this chapter presents the results of the full-sample analys is assessing the stability of the final model when the clustered nature of the data is taken into account. Descriptive Statistics Table 4-1 contains descriptive statistics for each variable based on the full sample. By comparing variable means to scale midpoi nts, and by examining skew statistics for each variable, it is possible to see a profile of relative strengths and weaknesses in this national sample of Senior Project programs . These statistics also lay a foundation for some conjectures about the response patterns that students adopted when completing the survey.
167 Table 4-1 Descriptive statistics for model variables in the full sample (N = 1,309) Variable Min Max Mid Mean SD Skew Critical ratioskew Autonomy 4.00 20.00 12.0012.27 3.29 -.258 -3.806 Clarity 4.00 20.00 12.0013.74 3.35 -.322 -4.761 Utility 3.00 15.00 9.008.81 3.54 -.048 -.715 Sensitivity 3.00 15.00 9.007.73 3.09 .140 2.068 Advisor 4.00 20.00 12.0015.27 3.45 -.695 -10.270 Parent 4.00 20.00 12.0013.80 3.52 -.340 -5.022 Peer 4.00 20.00 12.0010.33 3.67 .264 3.894 Prev Exper 2.00 10.00 6.007.55 1.76 -.571 -8.440 SP Efficacy 2.00 10.00 6.006.47 1.90 -.127 -1.882 Acad Engage 7.00 35.00 21.0025.66 4.09 -.534 -7.892 SP Engage 7.00 35.00 21.0026.86 4.78 -.565 -8.347 Note . Variable names are abbreviated as in previous tables. Min = minimum score on scale, Max = maximum score on scale, Mid = midpoint of scale, Mean = mean observed score on the scale, SD = standard deviation. Variable scores are the sum of responses to individual Likert-style items scored 1-5, with higher scores indicating stronger or otherwise more favorable standing on the variable. One noteworthy observation is that only two variables had positively skewed distributions. For these variable s, the scores concentrated near the lower end of the scale. The positive skew and below-midpoint mean of Sensitivity suggests that Senior Projects in this sample generally had the problem of conflicting with other demands on seniorsâ€™ attention. Similarly, the pattern of positive skew and relatively low mean of Peer Support suggests that respondents typically percei ved their peers as displaying minimal enthusiasm for Senior Project. Unlike Sensit ivity and Peer Support, Utility Value did not have a positively skewed distribution; however , it shares with these two variables (and no others) the property of having a mean less than or near the scale midpoint. This finding, in conjunction with Utility Valueâ€™s relatively large standard devi ation, suggests that students were generally divided in their feelings about the usefulness of Senior Project work with respect to their own liv es and the world outside school. Variables with negative skewâ€”those for wh ich scores tended to cluster at the higher end of the scaleâ€”also are informa tive. Advisor Support had the strongest
168 negative skew, and its mean was well above the scale midpoint. One interpretation of this pattern is that respondent s generally considered their Se nior Project advisors to be very helpful, respectful, and encouraging. An alternative explanation for the strong positive advisor ratings, or at least a contribu ting factor, is social desirability bias. Despite the steps taken to protect their a nonymity, students may have believed that the faculty member collecting the su rveys would see thei r answers. A reasonable response to such a concern would be offering exagge rated praise for their advisors. Other variables showed strong negative skew. The skew of Previous Mastery Experience and its above-midpoint mean suggest that students were likely to describe themselves as being more than moderately successful in earning good grades and completing class projects when they were juni ors. A similar pattern of negative skew and high mean was found for Academic Engageme nt and Senior Project Engagement. Operationally, students tended to rate themselves as showing more than just â€œoccasionalâ€ signs of engagement in their general coursewo rk and Senior Project activities. Several factors may have contributed to the positively-directed responses for these three variables. One is selection bias; those students most likely to invest the time and effort needed to complete the survey may have been the ones who are more successful and engaged in school. Another is impressi on management (e.g., Schlenker, 1980). Respondents may have been motivated to pr esent their achievements and work habits more favorably in order to impress the surv ey administrator or designerâ€”again, despite the advertised anonymity of the survey. The juxtaposition of positively slanted self-r eports of academic and Senior Project engagement with negatively slanted reports of peer engagement merits further comment.
169 It seems reasonable to assume that responde ntsâ€™ friends were among those completing the survey. If friends indeed ha ve similar values because of selection and influence factors (see Chapter 2), why did Peer Support not show the same pattern of negative skew and favorable ratings as Senior Project Enga gement? Likewise, why was the zero-order correlation between these variables not st ronger than the one actually obtained ( r = .406 in the full sample; see Table 4-2)? One e xplanation for the clear separation between ratings of self and peer engagement is that students work hard but are reluctant to admit high effort and engagement to their friends. Research has shown that older students are especially unlikely to explain successes to th eir friends in terms of effort (Harari & Covington, 1981; Juvonen & Murdoch, 1993, 1995). As as result, students may not have an accurate picture of how enga ged their friends actually are. Also notable are the low correlations be tween Previous Mastery Experience and most of the other variables in the model (see Table 4-2). The lowe st correlations were with Peer Support, Utility Value, Sensitivity, and Autonomy. At face value these results suggest that students who have experienced academic success do not necessarily judge their Senior Project programs highly and that , conversely, less su ccessful students do not necessarily judge the project negatively. If Senior Project indeed has the â€œequalizingâ€ power attributed to it by researchers (see Chapter 2), this result makes sense. Finally, it is remarkable that the aforemen tioned variables correlating weakly with Previous Mastery Experienceâ€” Peer Support, Autonomy, Se nsitivity, and Utility Valueâ€” correlate highly with each other. In fact, their intercorrelations are the highest in Table 42. Three of these variables were grouped together previously; r ecall that Peer Support and Sensitivity share the property of be ing positively skewed and having below-
170 Table 4-2 Correlations between model variables in the full sample Variable 2 3 4 5 6 7 8 9 10 11 1. Acad Engage .546 .300 .227 .283 .342 .253 .197 .248 .432 .244 2. SP Engage â€” .406 .406 .475 .504 .454 .373 .416 .341 .442 3. Advisor â€” .275 .307 .412 .303 .228 .271 .188 .232 4. Peer â€” .422 .395 .565 .583 .603 .020 .311 5. Parent â€” .410 .401 .373 .454 .122 .251 6. Clarity â€” .524 .503 .467 .212 .410 7. Autonomy â€” .653 .598 .093* .388 8. Sensitivity â€” .597 .048 .422 9. Utility â€” .028 .286 10. Prev Exper â€” .228 11. SP Efficacy â€” Note . Variables names are abbreviated as in previous tables. All correlations except those with superscripts are significant at p < .001. *p < .01; p > .05; p > .10. midpoint means, with Utility Value sharing th e relatively low mean. Autonomy differs from these variables in having significant negative skew; however, its mean only barely exceeds the midpoint of the scale, suggesti ng a lukewarm judgment of the degree of choice and freedom permitted in Senior Project. Several factors may help to explain the in terconnection of these four variables. First, there is a logical re lationship among Autonomy, Sensiti vity, and Utility Value. A student who experiences freedom in Senior Project (high Autonomy) seems likely to exercise that choice in ways that are persona lly useful (high Utility) and mindful of the many demands of the senior year (high Sens itivity). Although th is relationship may explain some of the shared va riance, it does little to account for the strong relationship of these three variables to Peer S upport. One possibility is that all four of these variables are â€œsafeâ€ ways for students to express neutral or negative feelings a bout Senior Project. If students wish to express middling enthusiasm for the project, they are unlikely to do so by responding negatively about themselves, their advisors, or their parentsâ€”the referents of the other variables in the m odel. Rather, the safest outle ts for an honest expression of
171 any neutral or negative feelings about Senior Project would be impe rsonal variables such as Autonomy, Sensitivity, and Utility Value, as well as the Peer Support variable that allows respondents to acknowledge such feelings in classmates or even to project their own feelings onto peers. Model Testing and Modification This section outlines the series of analyses and model modifications used to find a model that captures the pattern of covarian ces in the data. In the explanation that follows, each step includes a diagram of the m odel tested at that stage. Unlike the model diagrams presented earlier (see Chapter 2) , these diagrams represent variables with rectangles rather than ovals. This change is consistent with the standard practice in structural equation modeling of using recta ngles for observed vari ables and reserving ovals for latent variables (Kline, 2005). Also , to reduce clutter, error terms associated with each variable are omitted. Analytic resu lts associated with these error terms for the final model appear in Appendix L. In a ddition to diagrams, each step includes the following measures of model-data fit, in this order: the chi-square statistic (2) , its degrees of freedom (df), and its p -value; the Goodness of Fit Index (GFI); the Comparative Fit Index (CFI); the Non-Norm ed Fit Index (NNFI); and the Root Mean Square Error of Approximati on (RMSEA), along with its 90% confidence interval in parentheses. A summary of the model-refi nement steps, including a reminder of the guidelines against which the fit indices were compared (see Chapter 3), appears in Table 4-4 at the end of this section.
172 Model 1 Testing The fit of the original model, Model 1 (see Figure 4-1), was poor: 2(df = 43) = 1692.3, p < .001; GFI = .492; CFI = .054; NNFI = -.210; RMSEA = .297 (.285-.309). Advisor Peer Parent SP Efficacy SP Engage Acad Engage Prev Exper Clarity Utility Autonomy Sensitivity Figure 4-1.Path diagram for the original model of Senior Project Engagement. Abbreviations are as follows: Prev Exper = Previous Mastery Experience, Parent = Parent Support, Peer = Peer Support, Advisor = Advisor Support, SP Engage = Senior Project Engagement, Acad Engage = Academic Engagement, Utility = Utility Value, Clarity = Clarit y of Expectations, SP Efficacy = Senior Project Self-Efficacy. To identify the source(s) of the poor fit, I examined the modification indices provided by AMOS. Recall that a modification index (M I) shows the minimum amount by which the modelâ€™s 2 statistic would decrease (indicating bette r model-data fit) if the parameter in question were freely estimated rather than c onstrained to zero. For Model 1, the MIs suggested that freeing cova riances among certain error terms would improve fit significantly. To understand this suggestion, recall that each variable in the m odel has an error component, representing the part of the variable explained by something other than systematic covariance among the items on the scal e. People often think of this error as â€œnoiseâ€ obscuring a â€œsignalâ€ of theoretical interest. Ideally, the error in one variable is
173 unrelated to the error in another. In structural equation modeli ng, it is standard to assume that the errors are unrelated and thus to cons train the error covariances to zero. However, the MIs for Model 1 suggested that there is some source of systematic covariation among the errors that was not captured by the model. In fact, the MIs i ndicated that the error term for every potential â€œactive in gredientâ€ in this study covari es significantly with most of the other error terms. Fo r example, freeing the covariance between the error terms of Peer Support and Utility Value would decrease 2 by at least 188.086; corresponding figures for other error cova riances are 192.092 for Autonomy and Sensitivity and 78.507 for Peer Support and Parent Support. Th ese three examples are among the 37 error covariances suggested by the MIs. The easiest way to address the issue and improve fit is simply to remove the default constraints, allowing the error c ovariances to be freely estimated. As with any data-based model change, there should be a reasonable s ubstantive justificati on for every covariance freed in this way (e.g., Jreskog, 1993). Fo r example, in longitudinal studies measuring the same variables more than once, it is reasonable to expect variables to contain a variance component that is stable across time but cannot be absorbed into the variable proper (Bentler, 1987)â€”hence correlated errors . Except in such clear cases, however, methodologists generally seem to advise agai nst using correlated measurement error to improve model fit. An important reason to avoi d this â€œshotgunâ€ approach is that doing so can mask an underlying structur e that might provide a more informative explanation for patterns of covariati on (Gerbing & Anderson, 1984; Hoyle & Smith, 1994). Such an underlying structure seems especially likely in the case of Model 1 because the MIs indicated not just one or two correlated errors but an extensive network of
174 covariation among the error terms of the exoge nous variables. In such a case, one potentially useful solution is to try modeli ng the systematic covari ation among the errors as a function of a latent factor (e.g., Mars h, 1996; Palmer, Graham, Taylor, & Tatterson, 2002). The latent factor might repres ent a common-method measurement effect (Bagozzi, Yi, & Phillips, 1991; Spector, 1987; Tepper & Tepper, 1993). For example, homogeneous response formats or similar word ing for items used to measure different variables can cause correlated measurement er ror. This kind of systematic covariance attributable to method rather than theoretical constructs is a po tential threat to the internal validity of all correlationa l studies (Campbell & Fiske, 1959; Mitchell, 1985; Tepper & Tepper, 1993). By itself, introducing a method-related fact or does not explain the psychological process underlying the correlated errors. Social desirability is one possible explanation. Faced with a single survey written by an apparent authority and administered by a teacher, a student might shift answers subtly bu t consistently in a di rection more likely to please, or less likely to displease, these power ful others (e.g., Schmitt, 1994). In contrast, if some variables had been measured by self-report and others by teacher report or researcher observation, social desirability would not have been a common denominator for all the variables. Resear ch has shown that agree-disa gree response scalesâ€”exactly the kind used to measure the variables in th is study showing the greatest number of covarying errorsâ€”are especia lly vulnerable to social desi rability (e.g., Tay, 2005). Another interpretation of these covarying e rror terms relates to a pattern described in the section on descriptive statistics. Reca ll that Autonomy, Sensiti vity, Utility Value, and Peer Support were highly intercorrelated. As suggested, these relationships might be
175 explained by a shared property of providing safe outlets for students to express neutral or negative feelings about Senior Project. Of the 37 error covariances suggested by the analysis, the ones having th e largest MIs were, almost without exception, those connecting these four variab les with one another. Ne vertheless, the suggested covariances were not limited to these four variables. In fact, fairly strong error covariances were suggested for several person al variables that woul d be less vulnerable to the â€œsafe outletsâ€ effect. For example, Pa rent Supportâ€™s error term correlated strongly with error terms for Peer Support (MI = 78.507) and Util ity Value (MI = 68.303), and Advisor Supportâ€™s error term co rrelated with error terms for Clarity of Expectations (MI = 71.569) and Peer Support (MI = 55.012). These two explanations for the correlated errorsâ€”social desirability and safe outletsâ€”are not incompatible. In fact, they are complementary. A social desirability bias does not entail that responses will be positive. If that were the case, it would be hard to explain why students generally would say (f or example) that their peers do not have strong â€œbuy inâ€ for Senior Project or that Se nior Project deadlines and requirements are not very sensitive to the other demands on their time and energy. Rather, a social desirability bias would tend to nudge responses to be more positive than they would be in the absence of any desire to please or avoi d displeasing teachers and other authorities. Thus, social desirability bias can co-exist with a safe-outlets bias that encourages students to express neutral or negative feelings freel y when the referent of the question is not some person important to them (advisor, parent, or self). The plausibility of this interpretation is enhanced by an observation made by one of the Senior Project coordinators who ad ministered the survey at her school:
176 I really wish you could have done the survey in May or June when the kids have completed the experience and can reflect more honestly. Right now many of them are still in the stress of it all and this does affect how they view the program. (W. Wilson, personal communication, February 9, 2006) What this observation suggests is that st udents in the middle of the projectâ€”as all students in this study wereâ€”may have had a ge neral feeling about their work that would color all of their judgments. This general fe eling, in turn, would be run through a social desirability filter as part of the process of answering an individual question about Senior Project. To capture this complex explanation for the extensive error covariances, I introduced a latent variable called â€œMethod.â€ Model 2 Testing Introducing the Method fact or (see Figure 4-2) impr oved the model-data fit significantly: 2(df = 34) = 302.46, p < .001; GFI = .885; CFI = .846; NNFI = .751; RMSEA = .135 (.121-.149). However, overall fi t still failed to reach conventionally acceptable standards. Although parameter estimat es at this stage of model-fit are not entirely trustworthy, it is us eful to note that the Method f actor loadings of all seven exogenous Senior Project variables were signif icant, with most above .700 (standardized) and the lowest at .397. These estimates meet common guidelines for minimally acceptable factor loadings (e.g., Hair et al., 1998). At least tentatively, then, these findings suggest that the Met hod factor is capturing meani ngful common variance in the Senior Project variables.
177 Parent Peer Advisor SP Efficacy SP Engage Acad Engage Prev Exper Clarity Utility Autonomy Sensitivity Method Figure 4-2.Path diagram for Senior Project Engagement model after adding a latent variable (â€œMethodâ€). Variables are abbreviated as in previous figures. As in the previous iteration, I examined the modification indices (MIs) to identify where the model was failing to capture the patterns in the data. By far the largest MI (86.719) suggested a relationship between A cademic Engagement and Previous Mastery Experience. This relationship could be inco rporated into the mode l (thus improving fit) by modeling a direct path from Previous Ma stery Experience to Academic Engagement. Note that, in Models 1 and 2, there is a direct path from Previous Mastery Experience to Senior Project Self-Efficacy. This path represents the hypothesis that students who had greater success with project-re lated skills in the past are likely to feel more efficacious about Senior Project in the present. However, when the formative items were removed from the survey, this hypothe sis no longer was tested with adequate specificity. Consider the remaining items fo r Previous Mastery Experience, which began with the stem, â€œHow successful were you last year at â€: Doing big, complicated pr ojects for your classes? Getting good grades (or positive feedback) on the work you did in school?
178 Although the first item has a clear connection to Senior Projec t, the items as a pair are more clearly linked to pr evious success in school generally . If the model had a â€œgeneral academic self-efficacyâ€ variable, its relations hip to this measure of Previous Mastery Experience probably would be strong. In th e absence of such a variable, and given the strong connection of self-efficacy to engage ment (see Chapter 2), Academic Engagement is the â€œnext best thing.â€ Therefore, it is a ppropriate to model the covariance as a path representing the hypothesis that students who were more successful in school as juniors would tend to be more engaged in their senior-year coursework. Model 3 Testing Introducing this new path to the model (see Figure 4-3) improved model-data fit. However, the 2 statistic and fit indices suggested that the major relationships in the data still were not well captured by the model: 2(df = 34) = 209.40, p < .001; GFI = .918; CFI = .899; NNFI = .837; RMSEA = .109 (.095-.123). The small set of MIs produced in this analysis was dominated, both in frequency and magnitude, by suggested relationships between Academic Engagement and the exogenous vari ables in the study. For example, adding direct paths to Academ ic Engagement from Parent Support, Peer Support, and Clarity of Expect ations would yield relativel y large minimum changes in the 2 statistic (MIs = 27.578, 29.712, and 32.731, respectively). One interpretation is that st udentsâ€™ responses to questions about their Senior Project perceptions and experiences refl ected their views about their general academic experience as seniors. For instance, questions about peer engagement in Senior Project likely measured the degree to which oneâ€™s peer s support academic effort more generally.
179 Parent Peer Advisor SP Efficacy SP Engage Acad Engage Prev Exper Clarity Utility Autonomy Sensitivity Method Figure 4-3.Path diagram for Senior Project Engagement model after removing Prev Exper SP Efficacy and adding Prev Exper Acad Engage. Variables are abbreviated as in previous figures. Similarly, questions about the clarity of Se nior Project expecta tions probably invited comment (not necessarily conscious) on the more general degree of clarity in academic expectations. In turn, th ese general impressions about the senior-year academic experience may be colored by an underlying psychological orientat ion towards senior yearâ€”chronic senioritis for some, incipient se nioritis for others, and high enthusiasm for the senior year among a rare few. If this in terpretation is accurate, th en it is no surprise that there is significant c ovariance between general Academic Engagement and the Senior Project variables. One way to model the preceding interpretati on is to add individual paths from the Senior Project variables to Academic Engagement. For example, a positive path from Peer Support to Academic Engagement woul d suggest that peer support for academics (Senior Project included) contributes to grea ter engagement of oneself in academics. However, adding seven direct paths to the model would be unparsimonious. If Academic
180 Engagement were a major focus of this study, these additional paths mi ght be warranted. However, recall that Academic Engagement is like a covariate in this model, providing a baseline against which to test the effects of the â€œactive i ngredientâ€ variables on Senior Project Engagement. Ideally, then, there w ould be a more parsimonious way to model the systematic covariance between Acad emic Engagement and the seven exogenous Senior Project variables. The modification indices suggested one wa y to do so. Among the MIs provided in this analysis, the la rgest one (MI = 40.990) indicated a relationship between Academic Engagement and the Method factor, which in cludes all seven exoge nous Senior Project variables. Adding Academic Engagement to the Method factor de mands a rethinking of that factorâ€™s interpretation. To this point th e factor has been concep tualized as a studentâ€™s general feeling about Senior Project filtered through so cially desirable response tendencies. Given the preceding argument (i.e ., that responses to questions about Senior Project may have been â€œcontaminatedâ€ by fee lings and beliefs about senior-year work more generally), it is reasonable to reinterpre t the factor as repres enting a general feeling about the senior-year academic experience, again filtered through socially desirable response tendencies. Interprete d this way, the Method factor should be reflected not only in studentsâ€™ responses to ques tions about Senior Pr oject, but also in their answers to questions about how engaged they are in gene ral coursework. At this point it seemed appropriate to change the name of the factor from Method to General Academic Experience.
181 Model 4 Testing Adding Academic Engagement to the Ge neral Academic Experience factor (see Figure 4-4) improved the fit of the model to the data: 2(df = 33) = 166.99, p < .001; GFI = .932; CFI = .924; NNFI = .873; RMSEA = .096 (.082-.111). Advisor Peer Parent SP Efficacy SP Engage Acad Engage Prev Exper Clarity Utility Autonomy Sensitivity General Academic Experience Figure 4-4.Path diagram for Senior Project E ngagement model after adding Acad Engage to the General Academic Experience (f ormerly Method) factor. Variables are abbreviated as in previous figures. Before considering potential refinement of th e model, it seemed prudent to consider how strongly Academic Engagement was associat ed with the General Academic Experience factor. Given the still-inadequate fit, the parameter estimate addressing this question had to be treated with due caution. Nevertheless, it was notable that the loading of Academic Engagement on the Method factor wa s relatively small (standardized = .298), falling short even of the liberal guideline of .30 as a minimally acceptable loading (Hair et al., 1998). This result, however, was not entirely une xpected given the unusual structure of the current model. The highly atypical feature of Model 4 is the presence of a path from a
182 variable outside a factor (Pre vious Mastery Experience) to a variable within a factor (Academic Engagement). Ordinarily, observe d variables within a structural equation model have paths entering them from only two sources: the fact or and the error. Because Previous Mastery Experience at least tent atively showed a st rong relationship to Academic Engagement (standardized = .446 in Model 3 and .430 in the current model), there was less unaccounted-for variance in the la tter variable to be explained by the latent variable. For model refinement, there were few modi fication indices. Those remaining were dominated by suggested covariance relationshi ps between Advisor Support and Clarity of Expectations. Recall that A dvisor Support is supposed to reflect how encouraging and helpful students perceive their advisors to be , and Clarity of Expectations is a studentâ€™s perception of how transparent the project dead lines and expectations are. Certainly one way for an advisor to be supportive is to pr ovide clear directions and guidelines. (Of course, clear expectations may exist in the absence of a strongly supportive advisorâ€”for example, if the overall Senior Project coordinator is especially well-organized.) Therefore, including a direct path from A dvisor Support to Clarity of Expectations seemed a justifiable addition to the model. Model 5 Testing Introducing this new path (see Figure 4-5) improved model-data fit: 2(df = 32) = 142.54, p < .001; GFI = .941; CFI = .937; NNFI = .891; RMSEA = .089 (.074-.104). Although the exact fit of the model remained inadequate (indicated by the statistically significant value of 2), the fit indices collectively appro ached acceptable levels. At this point, I considered the statistical significance of the unstandardized parameter estimates.
183 Advisor Peer Parent SP Efficacy SP Engage Acad Engage Prev Exper Clarity Utility Autonomy Sensitivity General Academic Experience Figure 4-5.Path diagram for Senior Projec t Engagement model after adding Advisor Clarity. Variables are abbrev iated as in previous figures. Only four paths in Model 5 were statistically insignificant: those leading to Senior Project Self-Efficacy from Advisor Support ( = .007, p = .787) and Parent Support ( = .006, p = .803), and those leading to Senior Project Engagement from Sensitivity ( = -.106, p = .133) and Utility Value ( = .035, p = .577). (These four paths were consistently small throughout the model-refinement process.) These insignificant paths are discussed below. In the case of Senior Project Self-Eff icacy, neither support from the advisor nor parent â€œbuy inâ€ to the project we re significant influences. It is useful to recall that the empirical support for these hypotheses (see Chapte r 2) was qualified, especially in the case of Parent Support, by the observation that these variables tend to have larger effects on self-efficacy for younger students. For olde r students, with their greater accumulation of experience as learners and their greater need for psychological independence, the more important influence on self-efficacy is previous experience with the task (Bandura, 1997). Given these observations, the non-significance of paths from Advisor Support and Parent
184 Support to Senior Project Self-Efficacy is not fundamentally disruptive to the model. Therefore, these paths were remove d for the next model iteration. More surprising was the finding that Sensitiv ity and Utility Value had no effect on Senior Project Engagement when the effects of the other variables were controlled. Utility Value, in particular, was expected to show a significant effect, given both the robust support for its importance for high school students (see Chapter 2) and the observation that a major symptom of senioritis is the feeling that high school is no longer useful (see Chapter 1). The explanation fo r these non-significant paths may lie in the shared variance between Utility Value and Sensitivity, on the one hand, and Autonomy, on the other. The pairwise Pearson correla tions between these variables are among the highest in the calibration da taset (see Table 3-3), with Utility Value and Autonomy correlated at .584 and Sensitivity and Aut onomy correlated at .665. These strong associations make sense: students who feel free to make significant choices in Senior Project (high Autonomy) would te nd to select topics and de sign applications that would be useful to them (high Utility Value), a nd they would tend to structure and schedule their work so that it is sensitive to the other demands on their time and energy (high Sensitivity). Although Sensitivity and Utility Value are not redundant with Autonomyâ€” for example, a student who is employed 20 hour s per week may perceive a lot of choice in the project (high Autonomy) yet still feel that the project is inconvenient (low Sensitivity)â€”they seem to share an active ingred ient that makes the inclusion of all three variables in the model unnecessary. Model 6 Testing Upon removing the paths from Advisor Support and Parent Support to Senior Project Self-Efficacy and eliminating Sensitiv ity and Utility Value (see Figure 4-6), fit
185 improved: 2(df = 19) = 60.90, p < .001; GFI = .970; CFI = .961; NNFI = .927; RMSEA = .071 (.051-.091). While the 2 statistic remained significant, all fit indices for the final model were satisfactory, suggesting that th e basic patterns of covariation in the data were well captured by the model. Advisor Peer Parent SP Efficacy SP Engage Acad Engage Prev Exper Clarity Autonomy General Academic Experience Figure 4-6.Path diagram for Senior Projec t Engagement model after removing Parent SP Efficacy and Advisor SP Efficacy and eliminating Utility and Sensitivity. Variables are abbreviated as in previous figures. Because this model has acceptable fit, pa rameter estimates are more trustworthy, and it is productive to examine these estimates to draw preliminary conclusions about the most important â€œactive ingredientsâ€ in explai ning Senior Project E ngagement. Table 4-3 contains estimates for the major parameters in the model. (Parameter estimates related to error terms in the model appear in Appendix L.) Model 6 Interpretation Before interpreting the main results, a description of standardized estimates is in order. When path coefficients are standard ized, it is possible to compare the strength of influence among variables that are measured on different scales. St andardized estimates
186 Table 4-3 Maximum-likelihood parameter esti mates and bootstrap-adjusted standard errors for Model 6 based on the calibration subsample Parameter Unstandardized Estimate SE Standardized Estimate Direct effects on Seni or Project Engagement Parent .241**** .051 .192 Peer .135** .054 .114 Advisor .151*** .050 .120 Clarity .127** .059 .097 Autonomy .121* .063 .089 Acad Engage .326**** .041 .293 SP Efficacy .394**** .092 .171 Direct effects on Seni or Project Self-Efficacy Clarity .160**** .027 .282 Autonomy .153**** .028 .260 Other direct paths Prev Exper Acad Engage .990**** .096 .421 Advisor Clarity .146*** .046 .152 Loadings on General Academic Experience Parent 1.036**** .129 .556 Peer 1.486**** .162 .749 Advisor .847**** .114 .455 Clarity 1.000 â€” .558 Autonomy 1.262**** .139 .732 Acad Engage .724**** .117 .343 Note . SE = bootstrapped standard erro r for the unstandardized estimate. This loading was fixed to 1.000 in order to assign a scale to the latent variableâ€”a necessary step for the model to meet the stat istical requirement of identification (Kline, 2005). *p < .06; **p < .05; ***p < .01; ****p < .001. are interpreted as the number of standard de viations that one variable changes when the other increases by one standard deviation, holding all other variables in the model constant. For example, the standardized coefficient of .192 for the Parent Support Senior Project Engagement path indicates that, with all ot her variables controlled, an increase of one standard deviat ion in the Parent Support score predicts an increase of .192
187 standard deviations in the Senior Project Engagement sc ore. All subsequent path coefficients referenced in the text are standardized. Although the General Academic Experience factor is not the primary interest of this study, it is instructive to note the relative sizes of its variab le loadings. The two highest loadings were for Peer Support ( = .749) and Autonomy ( = .732). Prior to being removed from the model, Utility Value and Se nsitivity had factor loadings in the same range. These results are consistent with th e hypothesized meaning of the factor as an underlying evaluation of the senior-year academ ic experience, Senior Project included. As safe targets for the neutral or negative feelings that are not uncommon when students are in the thick of a demanding project, P eer Support and Autonomy (and Utility Value and Sensitivity before they were eliminated) would be expected to reflect most strongly an underlying state of senioritis. In c ontrast, Advisor Supportâ€”whose negative skew, high mean, low variability, and relatively lo w zero-order correlations with the other variables suggest an especially strong infl uence of social desirabilityâ€”had a weaker loading ( = .455). The path coefficients related to Senior Pr oject Engagement are of primary interest. The largest link to Senior Project Engageme nt is from the â€œinactive ingredientâ€ of Academic Engagement ( = .293). This result suggests that , other things being equal, the strongest influence on studentsâ€™ level of engage ment in Senior Project was their level of engagement in general academic work. Th e second largest path to Senior Project Engagement, from Parent Support ( = .192), suggests that parental involvement and encouragement were relevant in explaini ng a studentâ€™s behavioral and emotional investment in project work. Close in magnitude ( = .171) was the effect of Senior
188 Project Self-Efficacy on project engagement. Thus, when other variables are controlled, students with greater confidence in their ability to succeed on Senior Project were likely to be more engaged in the project. Although a ll the other direct effects on Senior Project Engagement were statistically significant (see Table 4-3), th eir magnitudes were so small that their substantive meaning is minimal. In fact, even the three largest effects (noted above) are modest, collectively ex plaining just over 15% of the variance in Senior Project Engagement. (The proportion of variance in an outcome explained by a given variable is computed by squaring its sta ndardized path coefficient.) The other major paths of interest in the or iginal model are thos e leading to Senior Project Self-Efficacy. After th e refinement of the original model, only two of the five paths remained: one from Clarity of Expectat ions and one from Autonomy. Both path coefficients were significant and relatively substantial ( = .282 for Clarity of Expectations and = .260 for Autonomy), collectivel y explaining approximately 15% of the variance in Senior Project Self-Efficacy. Substantively, the result for Clarity of Expectations suggests that a studentâ€™s sens e of self-efficacy for Senior Project was enhanced when expectations about the nature, quality, and timing of the work were clear. The result for Autonomy indicates that self -efficacy tended to be higher when a student perceived a high level of choice in the project. Two additional paths appear in the model. One, from Previous Mastery Experience to Academic Engagement, was the largest direct effect in the path model ( = .421). With other factors held constant, students w ho reported greater past success with projects and with their courses more generally tended to be more engaged in their senior-year coursework. Like the path from Academic Engagement to Senior Project Engagement,
189 this path belongs in the cate gory of â€œinactive ingredientsâ€; a Senior Project coordinator cannot directly manipulate a studentâ€™s prior success in school with the hope of producing greater engagement in classes and, in turn, Se nior Project. Finally, the path from Advisor Support to Clarity of Expectations ( = .152) suggests that ad visors played a role in clarifying and reinforc ing deadlines and othe r stipulations established at the program level. Summary of Model-Refinement Table 4-4 summarizes the development of th e model from its original to its final form. The change to the model that made the greatest difference in improving modeldata fit was adding the latent variable to capture a general orient ation to senior-year work. Without this factor, the mode l could not account for the extensive interrelationships among the error covariances. Cross-Validation of Model Results As described in Chapter 3, a multigroup stru ctural equation model test was used to determine how well Model 6 fit the data in th e validation subsample, which was not used for testing and refining the model. Per Va ndenberg and Lance (2000), the first step was to examine the fit of a baseline multigroup mode l stipulating that the same pattern of free and fixed parameters held across the tw o subsamples. The fit was adequate: 2(df = 38) = 189.95, p < .001; CFI = .930; RMSEA = .068 (.058-.077). (The inflation of the 2 statistic relative to its value in Model 6 is expected given that the sample size for this multigroup model is essentially twice the size of the calibrati on subsample.) Because the two subsamples evidenced similar basic configur ations, it was appropriate to proceed to a
190 Table 4-4 Summary of the models and the model-refinement process Iteration 2 (df) GFI (> .90) CFI (>.90) NNFI (>.90) RMSE A (<.10) Modifications to model Model 1 1692.30 (43) .492 .054 -.210 .297 Added Method factor. Model 2 302.46 (34) .885 .846 .751 .135 Replaced Prev Exper SP Efficacy by Prev Exper Acad Engage. Model 3 209.40 (34) .918 .899 .837 .109 Included Acad Engage in Method factor. Model 4 166.99 (33) .932 .924 .873 .096 Added Advisor Clarity. Model 5 142.54 (32) .941 .937 .891 .089 Removed Advisor SP Efficacy and Parent SP Efficacy. Removed Sensitivity and Utility from model. Model 6 60.90 (19) .970 .961 .927 .071 â€” Note . df = degrees of freedom. In parenthese s below each fit index is the guideline used in this study to determine acceptable fit. The arrow ( ) in the â€œModifications to modelâ€ column indicates a directed path in the model. more stringent invariance test, in which para meter estimates were constrained to be equal across the two groups. This multig roup model also fit adequately: 2(df = 43) = 192.05, p < .001; CFI = .932; RMSEA = .063 (.054-. 072). This result suggests that the magnitudes of the path coefficients we re similar between the two subsamples. The strength of the metric similarity was assessed by comparing the fit of the two multigroup models (Vandenberg & Lance, 2000). If the fit of the metric model is not significantly worse than the fit of the configural model, greater confid ence is warranted in the generalizability of the path coefficients found in the init ial model-testing. Table 4-5 presents a direct comparison of the two models.
191 Table 4-5 Model-fit and -comparison statistics for multigroup SEMs used to test crosssubsample generalizability of Se nior Project Engagement model Model Fit Model Comparison Invariance test 2(df) CFI RMSEA (90% CI) CFI 2(df, p ) Configural 189.95(38).930 .068(.058-.077) â€” â€” Metric 192.05(43).932 .063(.054-.072) .002 2.1(5, .835) Note . 90% CI = 90% confidence interval. CFI = difference in CFI between the two multigroup models. 2 (df, p ) = difference in 2 between the two models (degrees of freedom, p -value for the 2 difference) The difference in CFIs ( CFI) for this model comparison was .002, well within the 0-.02 range recommended by Cheung and Rensvold (2002). The RMSEA confidence intervals overlapped almost completely, lending further support to the metric similarity between the two subsamples. Fina lly, even the difference in the 2 statistics for the two multigroup models was nonsignificant, 2 (df = 5) = 2.1, p = .835, indicating strong generalizability of Model 6 to the validation subsample. Model Stability with Cluster Adjustment Results As indicated in Chapter 3, it was necessary to check the stability of the model after incorporating a sta tistical adjustment for the non-indepe ndence of the cases due to student clustering within schools (see Table 4-6). The results of the testing using the Mplus correction for complex sampling on the full sample indicated that the model c ontinued to be a good approximation of the relationships in the data: 2(df = 19) = 170.62, p < .001; CFI = .957; NNFI = .918; RMSEA = .078. In addition, the sizes and rela tive rankings of the path coefficients, both within and across effect categories (e.g., e ffects on Senior Project Engagement), were
192 essentially the same as those found in the model developed and cross-validated on the subsamples (see Table 4-6). Table 4-6 Maximum-likelihood parameter es timates and complex sample-adjusted standard errors for Model 6 based on the full sample Parameter Unstandardized estimate SE Standardized Estimate Direct effects on Seni or Project Engagement Parent .254 .030 .193 Peer .084 .027 .067 Advisor .154 .033 .114 Clarity .167 .020 .121 Autonomy .132 .050 .094 Acad Engage .393 .064 .340 SP Efficacy .463 .091 .189 Direct effects on Seni or Project Self-Efficacy Clarity .161 .017 .286 Autonomy .137 .023 .239 Other direct paths Prev Exper Acad Engage .900 .045 .395 Advisor Clarity .152 .035 .156 Loadings on General Academic Experience Parent 1.043 .206 .586 Peer 1.283 .228 .692 Advisor .755 .084 .434 Clarity 1.000 â€” .590 Autonomy 1.270 .266 .763 Acad Engage .703 .117 .347 Note . SE = standard error adjusted for th e non-independence of cases using Mplus. This loading was fixed to 1.000 in order to assign a scale to the latent variableâ€” necessary for the model to meet the statistical requirement of identification (Kline, 2005). Chapter Summary The original model developed to explain st udent engagement in Senior Project did not capture the patterns of covariance in the data. Given the early stages of research on Senior Project and the psychometric quality of the survey, this result was not surprising. Changes to the model were made based on m odification indices whose implications were
193 theoretically sensible and consistent with the descriptive characteristics of the data. The major structural change to the model was di ctated by the finding of extensive and strong covariance relationships among the error terms. This structural change was the addition of a latent variable that essentially repres ented a general judgment al orientation towards the senior year (consistent with the litera ture on senioritis), adjusted by a social desirability filter. At the end of this itera tive process of refineme nt, a model fitting the data well was obtained. The generalizability of this model to a subsample that had not been used for model refinement was checked and found to be excellent. Moreover, the model fit well when applied to data from the full sample with a correction for the clustering in the data. The main findings can be described as follows: Students repo rting a history of greater academic success were likely to show higher levels of engagement in their senioryear courses. This general level of engagement, in turn, was clearly and positively associated with degree of engagement in Se nior Project. Studentsâ€™ engagement in the project also was affected by the degree of interest and encouragement from parents and the level of self-efficacy the student felt with re spect to the project. Students were more likely to feel efficacious when they experien ced high levels of freedom and choice in a context of clear expectations. Advisors a pparently were key in providing this clarity.
194 CHAPTER 5 CONCLUSION The previous chapter provide d technical interpretations of the major effects found in the analysis. For example, the path co efficient of .192 between Parent Support and Senior Project Engagement technically means that an increase of one standard deviation in the Parent Support measure was associat ed with an increase of .192 standard deviations in the measure of Senior Proj ect Engagement. One purpose of the present chapter is to expand such narrow interpre tations by considering the theoretical and practical implications of the findings. Thes e implications must be viewed with due caution because of the studyâ€™s limitations, disc ussion of which forms the second part of this chapter. Finally, the implications and limitations of the study provide a basis for suggestions for future research. Interpretation and Implicat ions of Main Findings The Role of Inactive Ingred ients in Promoting Engagement The original and final models contain two â€œinactiveâ€ ingredients, so called because they essentially lie outside the control of people who design and administer Senior Project programs. These two i ngredients are Previous Master y Experience, defined as the studentâ€™s self-perceived hist ory of academic success; and Academic Engagement, the studentâ€™s level of investment in senior-year coursework generally. The paths originating from these variables represented the largest ef fects in the final mode l. Previous Mastery Experience had a strong dire ct link to Academic Engagement, which in turn had a relatively strong direct path to Senior Proj ect Engagement. These paths indicate that
195 successful students tended to be more engaged academically and that students with higher levels of academic engagement tended to comm it more time and effort to the specific task of Senior Project. These findings are not surprising. Reca ll that Previous Mastery Experience and Academic Engagement were included in the model precisely because they were expected to provide a baseline for eval uating the effects of the active ingredients. Nevertheless, corroborating the importance of these variables is useful. One practical implication for Senior Project personnel is that additional encouragement, support, and incentives may be needed to engage students who have str uggled academically in the past and those who (struggling or not) are minimally engaged in th eir daily coursework. At the same time, however, the effects of previous success a nd general academic engagement are not so large that they negate the oft-mentioned â€œequalizingâ€ potential of Senior Project (Darling-Hammond et al., 1995; DeFao, 2005; Pa rizek & Kavasan, 2000; see Chapter 2). In other words, there is plenty of varian ce in Senior Project Engagement above and beyond that explained by oneâ€™s academic hi story and general engagement in academicsâ€” plenty of room for seniors to â€œcome to lif eâ€ because of engaging features of a selfselected project. In addition to its practical implications, th e finding that prior success is associated with general academic engagement bears on theo retical issues described in Chapter 2. The path from Previous Mastery Experience to Academic Engagement in the final model does not represent a truly direct effect; that is, some explanatory mechanism(s) clearly must intervene between prior academic out comes and present investment of time and energy in academic work. One possible mech anism is self-efficacy. Thus, success with a
196 skill would breed confidence in applying the sk ill, which in turn would promote greater persistence, emotional involvement, and thoughtfu lness in future situations involving that skill (e.g., Bandura, 1997). Indeed, this causa l chain was the backbone of the original model: Previous Mastery Experience (related to Senior Project skills) Senior Project Self-Efficacy Senior Project Engagement (see Figur e 4-1). However, recall that this central chain in the original model had to be altered (see Figure 4-3) because eliminating the formative items from the survey change d the definition of the mastery and selfefficacy variables. For this reason, the current study cannot address the proposition that self-efficacy mediates the connection between previous success and present engagement â€”part of self-efficacy theory (Bandura, 1997) and expectancy value approaches (Wigfield & Eccles, 2000). This shortcoming of the study with respect to theory testing is at least partly compensated by the stimulus it provides for th eory building. Because this study indicates a connection between prior success and pres ent engagement but does not demonstrate that self-efficacy mediates this connection, there is a theoretical gap inviting speculation about possible mechanisms. For example, students who have experienced greater academic success are more directly familiar w ith the material and psychological rewards of such success. This familiarity, in turn, ma y foster a sense of attainability. Rewards that seem abstract, in contrast, are likely to have minimal incentive power. (The relation may not be linear. The power of rewards to stimulate greater engagement may increase with familiarity but eventually plateau or even decline. Work on the partial reinforcement effect in learning theory is consistent with this position [e.g., Nevin, 1988]). Another possible mechanism for th e link between previous achievement and
197 present engagement is the familiar â€œPygma lion effectâ€ (Rosenthal & Jacobson, 1968). Students with a history of academic success may be more engaged in the present because they receive more attention, encouragem ent, and challenge from their teachers. In sum, even so-called inactive ingredient s provide important points of leverage for increasing student engagement in Senior Pr oject. For example, although Senior Project advisors cannot change the academic history of an advisee, they can design appropriate interventions if they understand some of the more proximal reasons that prior success promotes engagementâ€”reasons such as enhanced self-efficacy, greater familiarity with rewards, closer attention and encouragemen t, or any number of other complementary mechanisms. In this way, prior academic success and general academic engagement do provide â€œactive sitesâ€ for catalyzing enga gement. Nevertheless, they do seem less manipulable than the other variables in the model, which are the subject of the next section. The Role of Active Ingredients in Promoting Engagement The active ingredients exhibiting the largest direct effects on engagement in Senior Project were Parent Support and Senior Pr oject Self-Efficacy. Efficacy, in turn, was affected by the degree of perceived autonomy an d by the clarity of project expectations. Support from the project advisor was a meani ngful source of clear expectations. These findings, described in the following subsect ions, are theoretically and practically valuable. Support from parents Recall that Parent Support, in the context of the other variables, had no significant relation to Senior Project Self-Efficacy. Th is path was therefore removed during model refinement (cf. Figures 4-5 and 4-6), suggesting that Pare nt Support does not contribute
198 to engagement indirectly via enhanced self-effic acy. However, the dire ct effect of Parent Support on Senior Project Engagement was amo ng the largest effects in the model. One interpretation of this findingâ€”the one with the clearest implic ations for people who coordinate Senior Projectsâ€”is that current parent involvement and interest in Senior Project supports student engage ment. Before exploring that interpretation, however, an alternative interpretation should be consider ed. Perhaps the items measuring Parent Support actually were tapping in to the cumulative effects of past educational involvement by parents, which is known to promote engagement (e.g., Hickman et al., 1995; Keith et al., 1993; Simon, 2001). Alt hough the Parent Support items were written specifically to reflect parentsâ€™ current involvement with Seni or Project (see Appendix A), students may have interpreted the items more generally. The favored interpretation, that studentsâ€™ responses were veridi cal representations of their parentsâ€™ current le vel of involvement, has a majo r practical implication: the importance of parent outreach in Senior Project, as in other high school academic endeavors (e.g., Simon, 2001). Research reviewed in Chapter 2 showed that overt parent support tends to decline as students progress through school. This trend can be overcome if parents know that teache rs desire and expect their involvement (Hoover-Dempsey & Sandler, 1995) and if they know how to be productively involved (Eccles & Harold, 1993; Epstein, 1986). In the case of Senior Pr oject, such information and expectations might be communicated via a parent meeting at the beginning of the year and a periodic Senior Project parent newsletter. Further, parents might be encouraged to attend the student oral presentations, a nd every effort should be made to schedule the presentations at times that are sensitive to parentsâ€™ work schedules. The potential cost of not keeping
199 parents apprised in these ways during pla nning and implementation of Senior Projects was evident in the parent protests describe d in Chapter 2 (Archer, 2005; Tomsho, 2005). An additional cost, revealed by the model testing, is lower student engagement once the project is in place. Accompanying these practical im plications are several theore tical ones. Of the four main theoretical frameworks grounding this study, none offers a t horough articulation of how parents promote academic engagement. In self-efficacy and expectancy value approaches, parents appear main ly in the role of helping ch ildren to develop a sense of confidence (e.g., Bandura, 1997; Wigfield & Eccl es, 2000). However, as indicated at the beginning of this subsection, this indirect pa th to promoting engagement was not evident in the present study. A natural question, then, is through what mechanism(s) Parent Support might have contributed to Senior Project Engagement. Modeling the relation between Parent Support and Senior Project Enga gement as a direct path does not mean that encouragement and interest from parents translates into greater engagement without mediation from some accessible and potentially manipulable behavioral or psychological process. Rather, it means only that possible intervening processes were not explicitly modeled in this study. The model of parent involvement developed by Hoover-Dempsey and Sandler (1995, 2005) suggests several such mechanisms by which student outcomes are affected. They include reinforcement, modeling, encour agement, and instruction. Some of the Parent Support survey items in this study tapped into exactly these mechanisms (see Appendix A). For example, parents who ask que stions about Senior Project (Item 26) are modeling both how to talk and think about academic matters and the value of doing so.
200 Parents who convey clearly that they like a studentâ€™s projec t idea (Item 30) are offering reinforcement for present engagement and encouragement for future engagement. This study therefore provides support for Hoover-Dem psey and Sandlerâ€™s model in a novel contextâ€”novel not only for its specificity (c f. Senior Project with the general-academic focus of Hoover-Dempsey and Sandlerâ€™s work ) but also for its focus on an age group thought to be, and often perceiving themselves to be, very independent from their parents. Self-efficacy: Its e ffects and causes Self-efficacy was an active ingredient in explaining Senior Project Engagement. Students who felt more confident and less worr ied about their success in Senior Project generally reported greater engagement. This result reinforces the central place that feelings of competence and c ontrol occupy in all four theoretical models grounding the present study. Although people certainly differ in their to lerance for frustration and failure (e.g., Clifford, 1988), few people are so inured to the resulting negative affect that they will engage in a task persistently and enthusiastically if they do not firmly believe that they can execute the requisite skills. The finding that efficacy is linked to engage ment in Senior Project has a number of practical implications. First, project coordinators and advi sors would do we ll to ensure that students have or develop confidence in thei r ability to execute th e essential skills of Senior Project. The self-efficacy literat ure (e.g., Bandura, 1997; Lopez & Lent, 1992) provides compelling evidence that the primary source of such confidence is previous experiences of mastery with those skills. Ther efore, students are more likely to have high confidence with respect to Se nior Project if the previous high school curriculum has provided opportunities to practice the essential skills. Such practice appears to be rare. For instance, the 2005 High School Survey of Student Engagement of over 80,000
201 students found that more than one-third of respondents reported having written no papers longer than 5 pages in the cu rrent school year (HSSSE, 2005b) . Other necessary skills, such as information literacy and oral fluency, are emphasized even less in the high school curriculum. A possible challenge in applyi ng this lesson is provi ding prior experiences that build efficacy without sacrificing the poten tially engaging feeling that one is having novel academic experiences in the senior ye ar. (Unfortunately, the poor psychometric quality of Novelty in this study made it im possible to determine whether a sense of newness was in fact associated with greater engagement.) Even if students have prior experiences with the essential skills of Senior Project, they will begin senior year with a wide rang e of levels of project -related efficacy. Thus, Senior Project coordinators and advisors face th e challenge of how to raise levels of selfefficacy to productive levels for all students in the senior year. Insight into how they might do so is provided by the othe r major findings of this study. First, recall that Clarity of Expectations had a relatively strong dire ct effect on Senior Project Self-Efficacy. Thus, students who understood what was expect ed of them were more likely to feel confident about doing good work on Senior Projec t. This finding is consistent with the commonsense notion that people are likely to feel more efficacious about hitting a clear target than a fuzzy oneâ€”a notion embodied cl early in both self-efficacy theory (e.g., Bandura, 1986) and Self-Determination Theory (e.g., Connell & Wellborn, 1991; Vallerand et al., 1997). Therefore, the significant Clarity of Expectations Senior Project Self-Efficacy path found in this study provides additional support for these theories in the novel context of Senior Project.
202 At a practical level, one way to create a cl ear target for Senior Project would be to show students examples of highand low-qua lity work and to discuss reasons for the evaluations. Students would need exposure to multiple exemplars of quality so that they understand the many forms that quality can take (Sadler, 1989). In addition, students should receive the grading rubric early in th e process so that they know exactly what criteria will be used to judge their work. Another way to promote clear expectations and thereby enhance self-efficacy is to provide frequent and timely feedback (Schunk, 1983, 1984; Schunk & Pajares, 2002). The need for such feedback is especi ally acute for longterm assignments like Senior Project, for which it can be difficult to sustain a clear focus and sense of momentum. Providing this kind of feedback is the job of the advisor. Therefore, in hindsight it is not surprising that the analysis suggested a relatively large path from Advisor Support to Clarity of Expectations. An advisor rated as helpful and encouraging certainly is likely to be one who offers the feedback and stru cture needed to keep the target clear. Expectations may be â€œobjectivelyâ€ clearâ€”t horoughly articulated in a Senior Project manual, in letters to parents, and in cl assroom sessions devoted to examining past projects with rubric in handâ€” but the advisor is in the be st position to facilitate the ongoing, subjective sense of clarity that enhanc es the studentâ€™s self-efficacy. Practically, then, Senior Project programs should ensure th at advisors have the training and the time needed to support students indi vidually. Time may be the mo re precious resource. Most teachers are busy with myriad job responsibil ities. Asking them to supervise student projects without formally building time into th eir schedules is a reci pe for resentment and low-quality advising. Because time set aside fo r faculty investment in Senior Project co-
203 opts time that might be used for other valuab le purposes, schools that wish to start a new program or improve an existing one must invol ve the faculty in serious discussions about the benefits and opportunity costs of the program. Support from advisors is one part of a set of variables that dir ectly and indirectly explained the project self-efficacy of student s in this study. The other major effect on Senior Project Self-Efficacy came from Autono my. Students who reported that they felt a lot of freedom to make choices in the progr am tended to feel more efficacious. This finding supports the contention in Self-Deter mination Theory that perceived competence at a skill increases only when the learning o ccurs in an environment where the student feels autonomous and respected (e.g., Williams & Deci, 1996). An additional theoretical implication, given that both Autonomy and Clarity of Expectations showed effects on self-efficacy, is support for the hypothesis that freedom and structure can be complementary rather than opposit ional (see Chapter 2). Ind eed, students often need help making choices. Miller (1972) expressed this point well in describing his experiences supervising independent studi es for high school students: Although students clamor for freedom to do their own thing, when freedom is extended, many youngsters are frequently at a loss to make choices. Instead of developing their own ideas the tendency is to turn to the faculty either instinctively or fearfully and ask, â€œW hat do you suggest?â€ (p. 77) The practical challenge of ensuring that autonomy and structure are in fact complementary is considerable in Senior Pr oject. Research in several New Hampshire high schools (Duff, 2004) revealed a tension between allowing students to design their own projects, on the one hand, and ensuri ng that they push themselves beyond their â€œcomfort zones,â€ on the other. For example, athletes who proposed coaching clinics for neighborhood youth encountered re sistance from faculty, especi ally when the students
204 had prior coaching experience. To the Senior Project advisors a nd coordinators, such projects seemed too â€œsafeâ€â€”taking advantage of existing self-efficacy but doing little or nothing to expand it. One key to resolving th e tension may be to reward and otherwise encourage risk-taking in the project (see, e.g., Clifford, 1991). Students who aim high, work hard, and nevertheless fall short should not have to fear negative repercussions in their project evaluations. Summary of Implications The implications of this study emerged primarily from an interplay between theory and practice. Consider, for example, the finding that prior academic success predicted current academic engagement. The prima facie practical implication of this result is that students with a history of academic difficulty are likely to need special attention and encouragement in order to engage in Seni or Project and thereby reap its potential benefits. However, considering the theoretic al implications of th e resultâ€”specifically, the potential mechanisms for the effectâ€”gener ated ideas that enriched the practical implications. In particular, Senior Project c oordinators and advisors should consider how to guard against advisor behavi or that might lead to self-f ulfilling prophecies for students, as well as how to make less successful students more directly familiar with the rewards that accompany school success. The same interplay of theory and practice was evident in deriving implications from the finding that Parent Support â€œdirectly â€ enhanced student engagement in Senior Project. The basic practic al lesson is that pare nt outreach related to the project is likely to pay dividends. What that outre ach might look like was again illuminated by theoretical speculation about m echanismâ€”that is, consideration of exactly how parent support might promote student engagement.
205 The finding that self-efficacy had a relatively strong direct effect on Senior Project Engagement suggested that proj ect advisors and coordinators should pay special attention to keeping project-relevant confidence high for all students. Beyond the curricular implication that students should have experi ence with project skill s (e.g., research and oral presentation) well in advance of senior year, this study suggested several others. Because both Autonomy and Clarity of Expect ations showed significant effects on Senior Project Self-Efficacy, special care should be take n to ensure that students have a balance of freedom and structure that enables them to make conse quential decision s about their own work within parameters that provide clear targets and room for risk-taking. A major source for such parameters is the project advisor. Schools therefore should provide teachers with time and support in the regular schedule to make advising a priority. By virtue of their relatively large path coefficients, the paths considered in this section merit special attention in designing and administering Seni or Project. Is it a corollary implication that vari ables exhibiting smaller effects in this study (e.g., Peer Support Senior Project Engagement), as we ll as variables that ultimately were removed from the model (e.g., Utility Value) , are unimportant? Not necessarily. The same caution warranted in applying the posit ive findings of this studyâ€”caution rooted partly in the methodological li mitations already described in Chapters 3 and 4â€”should be observed in dismissing small effects. The purpos e of the next section is to review these limitations and describe some additional ones. Collectively these limitations serve as a reminder that the present study is a preliminary ex ploration rather than an effort to test or refine a well-established th eory in a novel context.
206 Limitations of the Study Methodological Limitations Causal indeterminacy As indicated previously (see Chapter 3), the non-experimental and cross-sectional nature of this study necessarily limits the strength of potential cau sal inferences. Any prior uses of words such as â€œaffectâ€ and â€œinf luenceâ€ therefore should be regarded more as linguistic conveniences than as presumptions th at causality was clearly established. For instance, the finding of a relatively strong dire ct path from Clarity of Expectations to Senior Project Self-Efficacy certainly is cons istent with the intuit ively appealing claim that changes in studentsâ€™ self-efficacy for th e project can be caused by changes in the degree to which they understand what is bei ng asked of them. However, it also may be possible to explain the connecti on as follows: students with a clearer understa nding of the project may tend to be more academically ab le, and academically able students would tend to have higher self-efficacy for a school-b ased project. Under such a scenario, increasing the clarity of expectations might not increase self-efficacy because no change would occur in underlying academic ability. Su ch alternative explanations, even if they seem less plausible than the desired causal in terpretation, are reminders of the costs to internal validity associated with the convenience and rela tive ease of cross-sectional correlational research (e.g., Cook & Campbell, 1979). Variable contamination Additional validity concerns raised in Chap ter 3 relate to limitations in the quality of variable measurement. The questionnaire items per se were one source of potential problems. First, the survey did not under go the multiple rounds of pilot-testing and refinement that ultimately would have result ed in an excellent questionnaireâ€”one that
207 measured the relatively large number of variab les in this study in a way that balanced psychometric quality (e.g., high reliability for each scale) with sens itivity to how many items respondents could handle without becomi ng fatigued, irritated, or both. Second, although none of the intervariable correlations reached levels high enough to signal severe multicollinearity (i.e ., redundancy), a more convinc ing demonstration that the variables were clearly differen tiated would require a confirma tory factor analysis. The result of these weaknesses in measurement wa s a â€œnoisyâ€ set of data in which patterns among the true signals were more difficult to find. The quality of variable measurement also may have been limited by the specific format and timing of data collect ion. Recall that all data were in the form of self-report responses on a single survey. Especially wh en administered at a single point in time, such questionnaires are vulner able to systematic distorti on from respondentsâ€™ temporary moods and other biases (e.g., social desirability). Su ch factors would introduce construct-irrelevant variance to the measures of each variable that would be shared among the different variables. The follo wing comment from one of the project coordinators is worth repeating: I really wish you could have done the survey in May or June when the kids have completed the experience and can reflect more honestly. Right now many of them are still in the stress of it all and this does affect how they view the program. (W. Wilson, personal communication, February 9, 2006) The choice of a February administration was deliberate. The inte rest in this study was the influence of studentsâ€™ perceptions of project characteri stics and strength of support on current levels of engagement. A lthough it is true that surveying students in medias res is vulnerable to the kind of stress-re lated bias that worried the project coordinator, waiting to survey students until the projects were completed would have
208 been susceptible to its own set of potential bi ases. Not only are retrospective self-reports open to the effects of mood at the time of recall (e.g ., Blaney, 1986), but they are vulnerable to memory biases such as the te ndency to misremember information in ways that support decisions one already has made (Mather, Shafir, & Johnson, 2000). For instance, students who opted not to work very hard on Senior Project might justify their low engagement by â€œrememberingâ€ features of the project more negatively than they were actually experienced at the time. Although it sidesteps such memory biases , asking seniors about their current impressions of different aspects of Senior Pr oject appears to have been no less vulnerable to contamination from common, construct-irrele vant factors. Indeed, with the clarity of hindsight, it seems obvious that ju st such a potential factor was described at length in Chapter 1: the general mindset of senioritis. Schools with Senior Projects and other curricular innovations may have students with less severe cases of senioritis, but it probably would be nave to believe that su ch a culturally entrenched condition would disappear entirely. Two examples from the literature help to demonstrate this point. According to Humes (2003), counselors at hi ghly regarded Whitney High Sc hool tried to organize a senior retreat to the mountains that would include discussions and seminars on issues such as alcohol and sexâ€”issues for which re turning graduates said they were unprepared in college. While some seniors openly mock ed the idea of the retr eat, most responded with passive-aggressive lassitude. One vetera n teacher remarked, â€œEvery year, itâ€™s the same; the seniors think they know everything there is to knowâ€ (p . 321). Hoover (2003) similarly described a studentâ€™ s reaction to the idea of a major self-directed project:
209 Ms. Floyd, the student who is heading to Randolph-Macon [College], says that students are less likely to sh rug off projects in which they feel they are making a difference in the world. Stil l, she doubts that senioritis is entirely curable: â€œYou need to be a little bit free before you go off to college.â€ (p. A31) Part of being â€œa little bit freeâ€ may be approaching an optional questionnaire with less serious and careful thought than the author would wish. Recall that, after 25 bubble forms were removed because of non-sensical patterns of answers, 40 forms had to be recoded due to careless marking and an additi onal 56 had violated instructions written clearly at the top of the survey. This problem rate of just over 7% of the final sample may not be alarmingly high, but it does suggest the possi bility of some degree of â€œautopilotingâ€â€”a lack of consistent attention to the specifics of each survey question. Such a mindset, in turn, would engender th e consistency of response represented by the General Academic Experience factor, filtered th rough a social desirability bias, that was introduced to account for the larg e number of correlated error terms in the initial model. Because the possibility of a general factor was not considered a priori , these correlated error terms were an unexpected nui sance. They complicated the process of identifying engagement-promoting ingredient s of Senior Project and required the post hoc explanation reviewed in the preceding pa ragraph. With greater forethought, I would have included items on the survey that serv ed as checks on social desirability and on general feelings about the se nior-year academic experience. The absence of such checks is a definite limitation of this study, making th e introduction of the late nt variable a datadriven attempt to â€œmop upâ€ otherwise-unexplai ned patterns of covari ance. Even if the explanations offered are plausible, it is useful to recall the words of Hancock (1999): â€œSlapping oneâ€™s own forehead after inspecting residuals or modification indexes is not penance enough to pretend hindsight is foresightâ€ (p. 166).
210 Is this limitation at all mitigated by the encouraging results of the split-sample cross-validation? Recall that the multigroup invariance test showed that the final model indeed applied well to a set of data that was not used in the modelâ€™s derivation. The most conservative conclusion warranted by such a te st is that the model is robust against charges that the story it tells is merely a f unction of computational errors, data tabulation mistakes, or idiosyncrasies of a particular subsample (e.g., Barrett, 2003; Mulaik, 2003). This replication does not imply, however, that the final model is the model of Senior Project engagement. Another researcher usi ng a different protocol (e.g., timing of data collection and source of data) to study Senior Project might fi nd a very different model. For example, a study with a cleaner operationa lization of the same variables or slightly different timing might not need to introduce a common factor. If so, the factor in the present study would repres ent a nonsubstantive quirk of the present sample and methodologyâ€”a practical and theo retical nuisance indeedâ€”rat her than representing the complex substantive psychology for which I have argued. (The limitations of this kind of cross-validation are elaborated in the follo wing contributions to the SEMNET discussion group: Gregorich, 1998; Mu laik, 2004; Rigdon, 2006b.) Conceptual Limitations One-way paths Among the conceptual limitations of this st udy is the absence of reciprocal effects in the model. For example, interest a nd encouragement from parents may stimulate greater student engagement, but the reverse also is plausible: students who are more engagedâ€”for a variety of reasons other than parent supportâ€”are perhaps more likely to stimulate their parents to become involve d (Epstein, 1984; Hoover-Dempsey & Sandler, 1995; Schunk, 1989). Similarly, student enga gement may determine advisor support as
211 much as vice versa (e.g., Skinner & Belmont , 1993). Not including bidirectional paths in the model, despite theoretical and empirical support for their plausibility, was not an oversight but a purposeful decision grounded in technical concerns. First, structural models with feedback loops often have problems with iden tification (see Chapter 3) and with convergence of the estimation algorithm to a stable solution (Kline, 2006). Kline noted that such difficulties may be partly res ponsible for the scarcity of structural models with feedback loops in the behavioral sciences. Second, the original model already contained a large number of parameters to be estimated, and adding reciprocal paths would have further strained the ratio of samp le size to parameters. Even if justified, however, omitting reciprocal paths represents a limitation in the modelâ€™s potential to tell a complete story of Senior Project engagement. Assumption of causal homogeneity Another conceptual limita tion of the study is its a ttempt to generalize across populations that may be causally heterogeneous . Mulaik (2004) e xpressed this concern as follows: It [cross-validation] will support models th at are â€œaveragingsâ€ of what are distinct models in a mixture of popul ations. This is the case of lack of causal homogeneity in your subject population. If you have large enough samples you may be able to fit a model to this mixture of models, and then cross validate these well to a second random sample from the mixed population. . . . In reality, you need to consider disentangling the subject populations a nd developing separate models for each. This possibility may apply to the present st udy. Demographic data , including gender and race, were collected on the survey but were us ed only for descriptive purposes; they were not part of the model. This decision was de liberate, rooted mainly in concerns about model parsimony. However, for at least one of the potential engagement-promoting
212 ingredients in this study, some literature suggests that certain groups might be more affected than others. The ingredient in question is Advisor Support. The academic engagement of minority students may be especially dependent on the quality of their relationships with teachers. For example, Garber (2002) interv iewed nine high school students in an urban school who were nominated by teachers as â€œresis tantâ€ to learning. Qu alitative analysis of the transcripts showed that the students spen t a lot of time talking about how teachersâ€™ personalities and caring affected their attitudes and participation in school. Viadero (2001) described focus-group work in a sim ilar high school serving mostly low-income minority students. She noted that, when st udents were asked how school motivates them to work hardâ€”even with senioritis thr eateningâ€”they did not mention curricular innovations; instead, they referred to personal relationships with teachers and counselors. Finally, in focus groups with over 100 student s in a California high school with nearly 50% Hispanic enrollment, Hudl ey et al. (2003) found race-cor related differences in the traits that students tended to value. Whereas white students were more likely to mention teachersâ€™ classroom styles (e.g., â€œhe tells you everything that will be on the testâ€), Hispanic students were more likely to describe interpersonal factors (e.g., â€œshe is always there for youâ€). Another individual-difference variable not included in the study was studentsâ€™ general attitude towards the senior year. Although a latent variab le representing this attitude was introduced to capture the c ovariance among the major exogenous variables in the study, at no point was the value of this variable used as an explanatory factor. Thus, the engagement and self-efficacy res ponses of students with serious cases of
213 senioritisâ€”cynical and dismissive of near ly everything associated with high schoolâ€” may be explained by a different pattern of vari ables than are the enga gement responses of students with less serious cases. Along similar lines, different sources of senioritis may be associated with distinct explanatory models. For students whose feelings of frustration are grounded mostly in a percep tion that high school requirements do not respect their status as young adults, the de gree of autonomy offered by Senior Project may be the most important determinant of how engaged they are. In contrast, for students whose senioritis is rooted more in fatigue and a sense of being unable to fulfill all of their commitments, the sensitivity of the projectâ€™s parameters to these other commitments may be foremost in their minds. By not including race or the preceding indi vidual-difference variables as a potential explanatory variable in the model, or at le ast disaggregating the da ta and running separate analyses, this study potentially trades a more precise and discriminating set of models for one that averages across differen t subpopulations. Such an aver age is not useless; it just demands that anyone intending to apply the re sults of such a study to improve a Senior Project program should be especially mindful of local factors and experiences. Of course, even the result s of a particularly well-designed study with seemingly crystal-clear findings should be approached with the same local sensitivity rather than being applied mechanically. Omitted variables Race may not be the only variable w hose omission limits the theoretical and practical significance of the present findings. No te that the effect sizes in the final model were small (albeit statistically significant) , leaving large propor tions of unexplained variance in the major endogenous variables, Senior Project Engagement and Self-
214 Efficacy. The â€œnoisyâ€ measurement alrea dy described in this section probably contributed to the small effect sizes. However, it also is possible that the study failed to include some important variab les whose effects on self-effi cacy and engagement would have been larger. One set of possibilities is suggested by th e relatively large intr aclass correlations found for some of the variables. Recall wh at these correlations indicated: students chosen at random from the same school were li kely to be more similar on many variables than students chosen at random from differe nt schools. Although th e analysis corrected for this â€œnuisanceâ€ feature of the data, school -level variables were not included in the modelâ€”in part because of th e previous argument that people can perceive the same objective school conditions very differently (hence the need onl y for person-level variables), in part because adding school-l evel variables to the model would have compounded the complexity of th e analysis, and in part because there was no clear theory to support their inclusion. However, practical considerations suggest that certain structural, school-level features might have been produc tive additions to the model. Some schools, for example, provide a period during the academic day wh en students can work on Senior Project; others do not. Teacher-student ratio, insofar as it determines the amount of time advisors have to spend with each senior, may be impor tant. School location, as an index of the availability of community resources (e.g ., professionals to interview and shadow, university facilities to accommodate unusual equipment needs), also may influence how engaged students are in Senior Project. Th e potential effects of th ese variables on selfefficacy and engagement could not be determined in the present analysis.
215 Another variable that did not appear in th e model is perceived level of challenge. I had encountered this variable in my review of the engagement literature, especially in the work of Bronfenbrenner (1979), Newm ann (e.g., Newmann et al., 1992), and Csikszentmihalyi and colleagues (e.g., Cs ikszentmihalyi, Rathunde, & Whalen, 1993; Shernoff et al., 2003). However, I omitted perc eived challenge from the model in order to keep the number of variables manageable. This decision raises an important question: Why exclude this particul ar variable, especially when two variables that were included in the original modelâ€”Novelty and Sensitivityâ€”h ad less theoretical and empirical support in the literature? The basis for my decision was my experien ce as a Senior Project coordinator and a teacher of high school seniors. Among the many cases of senior disengagement with which I have dealt, few were due to boredom ro oted in the feeling that oneâ€™s skills were not being challenged and expanded. Rather, they were grounded in exactly the variables that I retained in the model: a sense that the work had no value outside of school, a feeling that they were not being treated as autonomous adults, a sense of being overwhelmed by too many demands at the same time, or a sense of â€œbeen there, done that.â€ My experience may be atypical, howev er, and therefore it is a shortcoming of the present study that I did not include pe rceived challenge in the model. Although engagement was indeed included in the model, its conceptualization and measurement may have been deficient, thus permitting an alternative interpretation of the findings. Recall that two items were removed from the engagement scales because of poor psychometric properties; one item asked students how often they enjoyed their work, and another asked how often they complained about it. Eliminating these items
216 yielded a scale that might be interpreted more as a measure of an accountability mindset and less as a measure of true engagement. The engagement scales still contained items about pride in oneâ€™s work and ease of concen trationâ€”feelings that seem to transcend such a performance orientation (see, e.g ., Ames & Archer, 1988)â€”and yet a larger number of the items may reflect just such an orientation (see Appe ndix A). For example, students may meet all their de adlines, be persistent, spend time outside school on their work, and check their work but have no deep psychological investment in it. This more superficial orientation certainly is a pos itive first step; most studies showing positive effects of engagement on academic performan ce use exactly the kinds of items in the current study. However, ultimately most educat ors want students to take the next step of truly caring about and enjoyi ng the process of exercising their skills and knowledge. Suggestions for Future Research Several recommendations for future resear ch on Senior Project engagement follow directly from the limitations noted in the previous section and in Chapter 3. One recommendation is to collect data using met hods other than student self-report. Although such reports may be the best way to access st udent attitudes and beliefs about features of Senior Project, other sources may be especi ally useful in gathering information about engagement. Teacher reports are one. Mult i-day observations of student work sessions by researchers are another. An additional sugge stion for future research follows from the conceptual limitation of averaging across pot entially causally-dis tinct subpopulations. More deliberate sampling, perhaps with ove rsampling of minority students, would be necessary to determine whether a one-size-fits-all model is appropriate. Other recommendations follow from the limitations of cross-sectional, correlational research. For example, quasi-experimental designs in which variables such as parent
217 outreach and clarity of expect ations are manipulated within or between schools would be useful in supporting causal conclusions. Perh aps more feasible (at least politically) are longitudinal studies using corre lational techniques in which variables are measured at multiple points in time. Among the benefits of such research woul d be the ability to model explicitly the reciprocal influence of variables such as Advisor Support and Senior Project Engagement. Longitudi nal studies also might enable better understanding of the trajectory of engagement as it relates to different project pha ses (e.g., research vs. handson application) and to different â€œstress pointsâ€ in the senior year (Zuker, 1997). The notion of â€œstress pointsâ€ serves as a reminder that the senior year of high school is a period of transiti onâ€”often away from home and friends and towards greater freedom and uncertainty. It is essential to understand the psychology of this transition so that programs such as Senior Project work with the grain rather than against it. Therefore, the literature on senior-year ps ychology described in Chapter 1, vivid and important but mostly anecdotal and athe oretical, should be supplemented by more systematic and large-scale research. For exam ple, Sizerâ€™s (2002) qua litative analysis of interviews with over 100 senior s across the country (described in Chapters 1 and 2) might be used to generate standard survey questions or interview protocol s that could be used more widely. Collaborations between edu cational psychologists and school-based counseling psychologistsâ€”the latter being more di rectly familiar with the daily stresses and complaints of high school senior sâ€”might prove similarly productive. Ultimately, the value of this recommended research, especially in an era of accountability, depends on the extent to whic h engagement in Senior Project can be shown to pay dividends in the form of demons trable, highly valued skills and attitudes.
218 Therefore, an important direction for future research is measuring and documenting the intellectual growth for which Senior Project cl aims responsibility. Do research skills, and information literacy more generally, improve as a result of Senior Project? There are instruments to measure such skills (e.g., Wise, Cameron, Yang, & Davis, 2005). Does the rigorous writing requirement of the project truly enhance studentsâ€™ preparation for the kind of writing they are likely to do in colle ge? Does Senior Proj ect contribute to the development of dispositions and skills th at are associated w ith good self-directed learners, such as the ability and desire to ask questions, to work independently, and to tolerate ambiguity and frustration? Here, too, there are standardized instruments to measure such outcomes (e.g., Guglielmino, 1977; see Delahaye & Choy, 2000, for a test review). Answering these questions is not prerequis ite to addressing the question driving the present study: What aspects of studentsâ€™ experiences of Senior Project are most conducive to high levels of engagement? Ind eed, the project is unlik ely to have any of these positive long-term outcomes unless senior s are invested in the work behaviorally, cognitively, and emotionally. This studyâ€™s conc lusions about the ingredients that inspire such engagement are less clear-cut than hoped. Nevertheless, the present work has provided a substrate for further research on Se nior Project and other potential remedies for senioritis by arguing for the issueâ€™s educ ational importance and by suggesting a set of potential active ingredients for any such remedy.
219 APPENDIX A SURVEY ITEMS AND RESPONSE OPTIONS BY VARIABLE Note: (R) indicates that the item is worded negatively and therefore was reverse-coded prior to analysis Engagement Scales Item stem: How often do you Response options: Never (1)-Rarely (2)-Occasi onally (3)-Most of the time (4)-Always (5) Academic Engagement Meet deadlines for your class assignments Quit when the work gets difficult (R) Check your work to make su re there are no mistakes Put maximum effort into your class work Spend time outside of school on class work Have fun in your classes Find it easy to concentrate in your classes Complain about your classes to other people (R) Feel proud of the work you do for classes Senior Project Engagement Meet all the deadlines for Senior Project Quit when the work gets difficult (R) Put maximum effort into Senior Project work Work very carefully on Senior Project Spend time outside school on the project Enjoy your work on Senior Project Find it easy to concentrate on project work Complain about the project to other people (R) Feel proud of the work you do on the project Task-Characteristics Scales Item stem: How much do you agree or disagree with Response options: Strongly Disagree (1) . . . Strongly Agree (5) Utility Value Senior Project helps me work on skills that are important in the â€œreal world.â€ Senior Project seems useless for preparing me for life after high school. (R) I believe that the work Iâ€™m doing for Seni or Project will help me reach my future goals. People outside school (for example, adults in the community) might actually be interested in my project work.
220 Autonomy I had total freedom in choosing my Senior Project topic. Senior Project gives me a lot more choices than I usually get in school. It feels like teachers have all the power in setting rules and deadlines for Senior project. (R) I feel good about how much control I have had in making decisions about my project. Sensitivity Senior Project adds too much work on top of the other stress of senior year. (R) Senior Project deadlines are sensitive to the other demands on me as a senior (for example, college applications or a job). The people who run Senior Project at my school seem to understand the challenges of senior year. Thereâ€™s enough flexibility in the way Senior Project is set up so that I can still deal with it when my schedule gets crazy. Clarity of Expectations Senior Project deadlines are clearly laid out so that I know when things are due. I understand clearly how my Senior Project work will be judged. There are plenty of chances to get feedback on my project so that I know if Iâ€™m doing ok. I feel lost about what I am supposed to do on Senior Project. (R) Novelty Even with Senior Project, my senior year feels boring. (R) Senior Project makes senior year feel different than last year. With Senior Project, I get a chance to do something new with my skills. Senior Project feels just like everything else we do in school. (R) Support Scales Item stem: How much do you agree or disagree with Response options: Strongly Disagree (1) . . . Strongly Agree (5) Advisor Support My faculty advisor (mentor) for Senior Project treats me with respect. I get a lot of encouragement from my facu lty advisor (mentor) for Senior Project. When I need help with something related to Senior Project, my advisor (mentor) is hard to find. (R) My Senior Project advisor (mentor) giv es me the help that I need to be successful. Peer Support My senior friends work hard on Senior Project. My friends and I sometimes help each other with Senior Project. Most of my friends think that Senior Project is a waste of time. (R) Overall, my close friends seem to enjoy Senior Project work.
221 Parent Support My parents/guardians never ask me questi ons about my Senior Project work. (R) My parents/guardians think that my Senior Project is a great idea. I have received help on Senior Project fr om a parent/guardian (for example, help contacting an expert or getting supplies I needed). My parent/guardian expects me to do the best work I can on Senior Project. Efficacy-Related Scales Senior Project Self-Efficacy Item stem: How confident do you feel about Response options: Not at all confident (1) . . . Very confident (5) Formative indicators: Finding all the research you need for Senior Project Writing the Senior Project paper Working on the practical (hands-on) part of the project Managing your time (for example, meeting deadlines) Reflective indicators: Overall, how confident are you about Senior Project? (same response options as above) How often do you worry about your ability to do good work on Senior Project? (R) [Response options: Never (1) . . . Always (5)] Previous Mastery Experience Item stem: How successful were you last year at Response scale: Not at all successful (1) . . . Very successful (5) Formative indicators: Finding the research you needed for class assignments Writing papers for your classes Doing practical and creative things related to your interests (for example, writing music, building something, etc.) Managing your time (for example, meeting deadlines) inside and outside school Reflective indicators: Doing big, complicated projects for your classes Getting good grades (or positive feedback) on the work you did in school
APPENDIX B FORMATTED SURVEY
227 APPENDIX C PERMISSION TO USE UNIVERSITY WORDMARK Subject: Re: Permission to use logo From: "Debbie Gay" Date: Mon, September 26, 2005 10:36 am To: email@example.com Priority: Normal Options: View Full Header | View Printable Version | View Message details No, I just wanted to make sure we were talking about the same logo!! Please accept this e-mail as permission to use the Seal as outlined below. I have attached a copy for your convenience. Please let me know if you have any questions or need anything else. Thank you for your cooperation with this office. Sincerely, Debbie Gay Licensing Manager/UF >>> < firstname.lastname@example.org > 9/26/2005 9:30 AM >>> Yes, I think I am referring to the University Seal (and not to anything with a gator on it). Should I be contacting someone else? Thanks, Bryan > Are you referring to the University Seal?? > >>>> < email@example.com > 9/26/2005 8:47 AM >>> > Hi Debbie, > > I am writing to request permission to use the official UF logo (the > circular one) on the front page of a survey that I will be > administering to high school students for my dissertation. I am in > the College of Education, and my survey asks questions about high > school seniors' level of engagement in their coursework and in a > special program called Senior Project. My dissertation supervisor > is Dr. Elizabeth Bondy, firstname.lastname@example.org . > > If you need more information, please email me. > > Thanks very much. >
228 APPENDIX D PILOT STUDY PROTOCOL APPROVED BY THE UNIVERSITY OF FLORIDA INSTITUTIONAL REVIEW BOARD 1. TITLE OF PROTOCOL: Assessing Readability and Other Features of a Survey Related to Student Engagement 2. PRINCIPAL INVESTIGATOR(s): Bryan Duff Graduate student, School of Teaching and Learning (College of Education) 7822 SW 56th Ave., Gainesville, FL 32608 Home phone: 338-8209 Email: email@example.com 3. SUPERVISOR (IF PI IS STUDENT): Dr. Elizabeth Bondy Professor, School of Teaching and Learning (College of Education) 2202 Norman Hall Phone: 392-0751 x247 Email: firstname.lastname@example.org 4. DATES OF PROPOSED PROTOCOL: From Dec. 5 to Jan. 15. 5. SOURCE OF FUNDING FOR THE PROTOCOL: Personal funds. 6. SCIENTIFIC PURPOSE OF THE INVESTIGATION: To assess whether the instruc tions and questions in a survey that I have developed for my dissertation can be understood w ithout difficulty by a variety of high school seniors. (The survey is incl uded here as an attachment.) To determine the average amount of time th at it takes for hi gh school seniors to complete the survey. 7. DESCRIBE THE RESEARCH METHODOLOGY IN NON-TECHNICAL LANGUAGE. I will sit down with each st udent individually in a room at the high school (e.g., the library) and follow this sc ript: â€œIâ€™m putting together a survey for high school seniors and want to see if Iâ€™ve written it well. Iâ€™d like for you to complete the survey at a normal pace. If there are any questions that donâ€™t make sense, or any words that youâ€™ve never seen before, please make a note in the margin of the survey. When youâ€™re finished, Iâ€™d like to ask you a few questions about what the survey was like. Do you have any questions for me before starting? Before we start, I want to make sure you know your
229 rights as a research participant, so Iâ€™d appr eciate it if you would read this form [informed consent].â€ Participants then will complete the survey while I am sitting in the room. After the student completes the survey, I will ask the following questions, in order: Did you make any notes in the margin about questions or words that didnâ€™t make sense? Was the layout of the survey easy to follow? Were any of the questions difficult to answer? Was there any time during the survey when you wanted to stop answering? Did any of the questions offend you? 8. POTENTIAL BENEFITS AND ANTICIPATED RISK. (If risk of physical, psychological or economic harm may be involv ed, describe the steps taken to protect participant.) No more than minimal risk. 9. DESCRIBE HOW PARTICIPANT(S) WILL BE RECRUITED, THE NUMBER AND AGE OF THE PARTICIPANTS, AND PROPOSED COMPENSATION (if any): I will ask a teacher at P.K. Yonge De velopmental Research School who teaches English to high school juniors if he would be willing to pass out consent forms in his classes. If he agrees, and if the research director at the school (Dr. Brian Marchman) approves it, I will ask the teacher to read the following script at the beginning of each class: â€œA local graduate student is looki ng for high school juniors who would be willing to sit down in the morning or afternoon and l ook at a survey. He wants feedback about how clear the survey is. The researcher is offering $10 for 25-45 minutes. This survey has nothing to do with this cl ass and your and your parentsâ€™ decision about whether you wish to participate will have no effect on your grade. In fact, I will put this folder at the back of the room, and you will have until the end of the week to put your signed form in there, if you wish. I will not look inside th e folder. If you are interested, your parents will need to read and sign this form by the end of the week.â€ I will return at the end of the week to pick up the fo lder. I will randomly select three of the parental-consent forms and pr oceed to contact the student by telephone to arrange a time to meet at the school. Participants will receive $10. 00 for their participation. 10. DESCRIBE THE INFORMED CONSENT PROCESS. INCLUDE A COPY OF THE INFORMED CONSENT DOCUMENT (if applicable). The informed-consent document for parents, which will be passed out to students by the teacher, is attached. Immediately before participants sit down to take the survey and comment on it, they will be asked to read and sign a consent form that reiterates and reinforces their rights as participan ts. That form is also attached.
230 APPENDIX E DEMOGRAPHIC CHARACTERISTICS OF RECRUITED SCHOOLS Table E-1 Recruited schoolsâ€™ location, size, racial composition, and average socioeconomic status School ID State Locale Studentteacher ratio % Minority % Free/ reduced lunch 1 CA Urban fringe 27.2 29.0 12.3 2 VT Rural 9.2 2.0 37.0 3 NC Rural 16.6 60.0 54.0 4 PA Urban fringe 13.2 14.7 4.0 5 MA Rural 8.9 4.0 6.0 6 MD Urban fringe 15.6 16.6 11.5 7 FL Urban fringe 17.6 24.4 25.0 8 FL Mid-size central city 18.9 12.0 4.8 9 NC Rural 13.3 <1.0 69.0 10 FL Urban fringe 17.8 36.0 20.0 11 NH Urban fringe 11.0 3.6 2.0 12 TN Rural 19.5 8.3 5.9 13 PA Urban fringe 14.1 7.0 3.3 14 NV Mid-size central city 18.9 24.0 21.0 15 FL Mid-size central city 16.5 37.0 15.5 Note . Except in the case of School 12, all information was obtained in fall 2005 from the Common Core of Data published by the Nationa l Center for Education Statistics on its website, www.nces.ed.gov/ccd . Information for School 12 was obtained by calling the appropriate dist rict office.
231 APPENDIX F SURVEY RETURN DATA BY SCHOOL Table F-1 Number of surveys and re sponse rates for participating schools School ID Subsample size Response rate (%) 1 157 79 2 24 40 3 28 65 4 181 53 5 40 67 6 171 49 7 105 60 8 78 65 10 96 48 12 148 49 13 123 49 14 91 30 15 67 84 Note . Schools 9 and 11 were not included in the study for reasons described in the text. Subsample size is the number of useable su rveys returned by a gi ven school. Response rate is the subsample size divi ded by the total number of su rveys sent to the school.
232 APPENDIX G MAIN STUDY PROTOCOL APPROVED BY THE UNIVERSITY OF FLORIDA INSTITUTIONAL REVIEW BOARD 1. TITLE OF PROTOCOL: Engagement in Senior Project Work as a Function of Support and Task Variables 2. PRINCIPAL INVESTIGATOR: Bryan Duff Graduate student, School of Teaching and Learning (College of Education) 7822 SW 56th Ave., Gainesville, FL 32608 Home phone: 338-8209 Email: email@example.com 3. SUPERVISOR (IF PI IS STUDENT): Dr. Elizabeth Bondy Professor, School of Teaching and Learning (College of Education) 2202 Norman Hall Phone: 392-0751 x247 Email: firstname.lastname@example.org 4. DATES OF PROPOSED PROTOCOL: November 2005 to March 2006 5. SOURCE OF FUNDING FOR THE PROTOCOL: Personal funds. 6. SCIENTIFIC PURPOSE OF THE INVESTIGATION: To determine the set of factors that expl ains the greatest proportion of variance in high school seniorsâ€™ engagement in Senior Project work. Ultimately, the goal is to provide useful information to people w ho coordinate high school Senior Project programs. 7. DESCRIBE THE RESEARCH METHODOLOGY IN NON-TECHNICAL LANGUAGE. The attached survey will be administered to high school seniors by the schoolâ€™s Senior Project coordinator at some point dur ing the first two weeks of February 2006. At the beginning of the week, the Senior Projec t coordinator will pass out an envelope to students that contains the survey, a scantron sh eet, an informed consent document for self and parents, and a $1.00 McDonalds gift cer tificate (in schools whose districts allow student compensation). The coordinator will read the following script: â€œA student at the University of Florid a is interested in your opinions about Senior Project. He has a survey that he has asked us to administer. We would like you to take this envelope home and read the informed-consent
233 form. If you wish to participate, please check the appropriate box on the consent form and then, if you are not yet 18, have at least one of your parents or guardians read the consent form. If they approve and sign the form, then you may fill out the survey and return it to me in the sealed envelope. If you do not wish to particip ate, that is just fine. You are NOT required to participate. If thatâ€™s what you decide, please check the box indicating that you do not wish to part icipate, and then return everything in the sealed envelope to me by the end of the week. (In districts that allow student compensation): Whether you participate or not, you are still welcome to take the gift certificate that is included in the packet.â€ 8. POTENTIAL BENEFITS AND ANTICIPATED RISK. (If risk of physical, psychological or economic harm may be involv ed, describe the steps taken to protect participant.) No more than minimal risk. No benef its other than the compensation described below. 9. DESCRIBE HOW PARTICIPANT(S) WILL BE RECRUITED, THE NUMBER AND AGE OF THE PARTICIPANTS, AND PROPOSED COMPENSATION (if any): I have contacted Senior Project c oordinators at 15 high schools around the country, all of whom have e xpressed interest in particip ating. Permission has been obtained from all principals at these schools. Seniors in high school are typically 17 a nd 18 years old. Seniors who are 18 may fill out the survey after completing the student consent form. Those who are 17 will be required to obtain a parent si gnature on the parental-conse nt form before they can participate; these 17-year-old students also w ill sign the student consent form as an assent form. I do not know exactly how many seniors will agree to participate in the study. However, a reasonable estimate (based on convers ations with Senior Project coordinators at the various schools) would be 100 seniors at each school, for a total of 15(100) = 1500 seniors. All students who take the envelope home , whether or not they actually complete the survey, will be allowed to take the $1.00 Mc Donalds gift certificate. In addition, each school that agrees to participate will receive $40.00, which I will ask to be earmarked to support the Senior Project program at the school. 10. DESCRIBE THE INFORMED CONSENT PROCESS. INCLUDE A COPY OF THE INFORMED CONSENT DOCUMENT (if applicable). The informed-consent documents for parents and students are attached.
234 APPENDIX H ITEM-LEVEL DESCRIPTIVE STATISTICS Table H-1 Item-level means and standard deviations for the full sample Variable Item M SD Variable Item M SD Acad Engage 1 4.25 .747SP Engage 15 4.22 .958 2R 4.02 .839 16R 4.04 1.017 3 3.28 1.036 17 3.78 .993 4 3.73 .823 18 3.79 1.022 5 3.45 1.067 19 3.73 1.011 6 3.32 .962 20 4.01 1.006 7 3.36 .902 21 3.10 1.295 8R 2.92 1.038 22R 2.89 1.281 9 3.58 .863 23 3.30 1.033 Utility 37 3.03 1.350Autonomy 38R 2.06 1.145 44 2.89 1.358 42 3.97 1.214 47 3.06 1.265 50 3.32 1.204 51R 2.88 1.407 55 2.92 1.267 Sensitivity 39 3.03 1.339Clarity 41 4.02 1.159 43R 2.04 1.222 46R 3.35 1.236 49 2.65 1.341 54 3.07 1.244 52 2.94 1.287 56 3.29 1.208 Novelty 40 3.31 1.339Advisor 25 4.45 .935 45R 3.55 1.279 28 3.48 1.209 48 3.84 1.216 32R 3.65 1.244 53R 3.41 1.349 34 3.69 1.176 Peer 27 3.14 1.126Parent 26R 3.41 1.367 29 2.67 1.295 30 3.08 1.394 31 2.25 1.262 33 3.14 1.442 36R 2.27 1.406 35 4.18 1.073 SP Efficacy 10 3.74 .993Prev Exper 57 3.88 .990 11 3.54 1.085 58 3.82 1.005 12 3.88 1.048 59 3.78 1.071 13 3.50 1.160 60 3.66 1.098 14 3.59 1.039 61 3.61 1.033 24R 2.89 1.258 62 3.94 .990 Note . M = mean, SD = standard deviation. Vari able names are abbreviated as in previous tables. Item numbers correspond to their order on the formatted survey. R indicates that the item was reverse-coded. Item scores range 1-5, with higher numbers indicating stronger or otherwise more favorable standing on the unde rlying variable.
235 APPENDIX I RELIABILITY STATISTICS FOR MEASURE REFINEMENT Table I-1 Alpha-if-item-remove d and corrected item-total correlations for item-level data in the refinement subsample Variable ( ) Item iTr Variable Item iTr Acad Engage (.745) 1 .718 .465 SP Engage (.823) 15 .813 .456 2R .724 .414 16R .824 .353 3 .710 .489 17 .793 .632 4 .696 .599 18 .787 .685 5 .725 .414 19 .794 .623 6 .746 .279 20 .812 .467 7 .719 .445 21 .797 .592 8R .753 .256 22R .821 .421 9 .707 .522 23 .803 .546 Advisor (.743) 25 .701 .515 Peer (.689) 27 .645 .440 28 .663 .572 29 .683 .378 32R .754 .421 31 .556 .576 34 .604 .668 36R .600 .509 Parent (.576) 26R .494 .371 Clarity (.637) 41 .587 .387 30 .547 .309 46R .610 .357 33 .472 .398 54 .547 .444 35 .504 .372 56 .522 .479 Autonomy (.616) 38R .684 .185 Sensitivity (.659) 39 .535 .518 42 .575 .357 43R .593 .440 50 .421 .556 49 .513 .548 55 .450 .514 52 .700 .270 Utility (.808) 37 .724 .696 Novelty (.418) 40 .395 .189 44 .718 .708 45R .314 .266 47 .825 .474 48 .358 .224 51R .758 .629 53R .331 .248 Prev Exper (.681) 61 â€” .517 SP Efficacy (.518) 14 â€” .356 62 â€” .517 24R â€” .356 Note . R next to an item indicates that the item was reverse-coded before analysis. is the value of coefficient alpha for the scale when all items are included. is the value of coefficient alpha for the scale when the given item is removed. iTr is the corrected itemtotal correlation. Not applicable because a one-item scale has no internal consistency.
APPENDIX J INPUT MATRIX FOR INITIAL MODEL TESTING
237Table J-1 Variance-covariance matrix for the calibration subsample Variable 1 2 3 4 5 6 7 8 9 10 11 1. Acad Engage 15.743 8.6993.7343.6463.843 4.1072.901 2.561 2.9572.949 1.811 2. SP Engage 8.699 19.7975.9937.3716.973 6.7826.171 4.794 6.1782.292 3.453 3. Advisor 3.734 5.99311.9674.5373.692 4.6773.009 2.082 3.6000.869 1.292 4. Peer 3.646 7.3714.53713.6025.427 5.2456.750 6.483 8.6270.211 2.223 5. Parent 3.843 6.9733.6925.42711.998 3.9854.127 3.152 4.8830.701 1.348 6. Clarity 4.107 6.7824.6775.2453.985 11.1105.282 5.031 5.2380.809 2.593 7. Autonomy 2.901 6.1713.0096.7504.127 5.28210.276 6.562 7.0780.230 2.424 8. Sensitivity 2.561 4.7942.0826.4833.152 5.0316.562 9.490 6.4010.187 2.562 9. Utility 2.957 6.1783.6008.6274.883 5.2387.078 6.401 12.6540.164 1.777 10. Prev Exper 2.949 2.2920.8690.2110.701 0.8090.230 0.187 0.1642.771 0.423 11. SP Efficacy 1.811 3.4531.2922.2231.348 2.5932.424 2.562 1.7770.423 3.607 Note . Variables are abbreviated as in previous tables.
APPENDIX K INPUT MATRIX FOR COMPLEX SA MPLE-ADJUSTED MODEL TESTING
239Table K-1 Variance-covariance matrix for the full sample Variable 1 2 3 4 5 6 7 8 9 10 11 1. Acad Engage 16.693 10.6514.2273.4114.0734.6893.409 2.4833.5903.1081.887 2. SP Engage 10.651 22.8306.6887.1177.9918.0757.155 5.4977.0342.8664.006 3. Advisor 4.227 6.68811.8903.4803.7224.7633.438 2.4293.3071.1441.518 4. Peer 3.411 7.1173.48013.4715.4574.8606.832 6.6117.8400.1302.162 5. Parent 4.073 7.9913.7225.45712.4044.8464.651 4.0565.6670.7571.677 6. Clarity 4.689 8.0754.7634.8604.84611.2505.794 5.2065.5521.2522.609 7. Autonomy 3.409 7.1553.4386.8324.6515.79410.856 6.6456.9800.5412.425 8. Sensitivity 2.483 5.4972.4296.6114.0565.2066.645 9.5346.5330.2612.472 9. Utility 3.590 7.0343.3077.8405.6675.5526.980 6.53312.5460.1761.919 10. Prev Exper 3.108 2.8661.1440.1300.7571.2520.541 0.2610.1763.1040.761 11. SP Efficacy 1.887 4.0061.5182.1621.6772.6092.425 2.4721.9190.7613.595 Note . Variables are abbreviated as in previous tables.
240 APPENDIX L MODEL RESULTS FOR ERROR TERMS Path coefficient from the corresponding error term Unstandardized estimate SE Standardized estimate Parent 1.254 .048 .831 Peer 1.193 .063 .663 Advisor 1.759 .065 .890 Clarity 1.265 .051 .767 Autonomy 1.179 .059 .682 Prev Exper 1.524 .052 1.000 Acad Engage 1.669 .059 .840 SP Efficacy 1.000 â€” .550 SP Engage 1.000 â€” .559 Note . Variables are abbreviated as in previous tables. indicates a direct path from the error term to the given variable. SE = bootstrapped standard error. This path was fixed to 1.000 in order to sa tisfy the identification condition (Kline, 2005).
241 REFERENCES ACT, Inc. (2006). Reading between the lines: What the ACT reveals about college readiness in reading . Retrieved June 4, 2006, from the American College Testing Program website: http://www.act.org/path/pol icy/pdf/reading_report.pdf . Adelman, C. (1999). Answers in the tool box: Academi c intensity, attendance patterns, and bachelorâ€™s degree attainment . Washington, DC: U.S. Department of Education. Adelman, C. (2006). The toolbox revisited: Paths to degree comp letion from high school through college . Washington, DC: U.S. De partment of Education. Ainley, J. (1994, April). Multiple indicators of high school effectiveness . Paper presented at the meeting of the American Educational Research Association, New Orleans, LA. Albrecht, T. L. & Adelman, M. B. (1987). Communicating social support . Newbury Park, CA: Sage. Alvermann, D. E. (2002). Effective literacy instruction for adolescents. Journal of Literacy Research, 34 (2), 189-208. Amen, J., & Reglin, G. (1992). High school seniors tell why they are â€œstressed outâ€. Clearing House, 66 (1), 27-29. Ames, C., & Archer, J. (1988). Achievement goals in the classroom: Studentsâ€™ learning strategies and motivation processes. Journal of Educational Psychology, 80 (3), 260-267. Anderson, A. R., Christenson, S. L., Sincla ir, M. F., & Lehr, C. A. (2004). Check & Connect: The importance of relationships for promoting engagement with school. Journal of School Psychology, 42 , 95-113. Anderson, J. C., & Gerbing, D. W. (1988). Structural equation m odeling in practice: A review and recommended two-step approach. Psychological Bulletin, 103 (3) , 411423. Andrews, H. A. (2004). The dual-credit phenomenon: Challenging secondary school students across 50 states. Stillwater, OK: New Forums Press.
242 Arbuckle, J. L. (2005). AMOS (Version 6.0) [Computer software]. Chicago: Small Waters. Archer, J. (2005, April 13). R.I. downplays tests as route to diplomas: Students must demonstrate their knowledge, skills. Education Week, 24 (31), 1, 24-25. Arms, M. (1980). What has happened to the se nior project: A view from the cockpit. Independent School, 39 (4), 19-22. Arnett, J. J. (1997). Young peopleâ€™s concep tions of the transition to adulthood. Youth & Society, 29 , 1-23. Arnett, J. J. (1998). Learning to stand al one: The contemporary American transition to adulthood in cultural and historical context. Human Development, 41 , 295-315. Arnett, J. J. (2000). Emergi ng adulthood: A theory of development from the late teens through the twenties. American Psychologist, 55 (5), 469-480. Arnone, M. P. (2003). Using instructional design strategies to foster curiosity (ERIC Digests). Syracuse, NY: ERIC Clearinghouse on Information and Technology. (ERIC Document Reproduction Service No. ED479842) Arnstein, R. L. (1980). Th e student, the family, the university, and transition to adulthood. In S. C. Feinstein, P. L. Giov acchini, J. G. Looney, A. Z. Schwartzberg, & A. D. Sorosky (Eds.), Adolescent psychiatry: D evelopmental and clinical studies, vol. 8 (pp. 160-172). Chicago: Un iversity of Chicago Press. Arter, J., & McTighe, J. (2001). Scoring rubrics in the cl assroom: Using performance criteria for assessing and im proving student performance. Thousand Oaks, CA: Corwin Press. Assor, A., & Connell, J. P. (1992). The validity of studentsâ€™ self-reports as measures of performance affecting self-appraisals. In D. H. Schunk & J. L. Meece (Eds.), Student perceptions in the classroom (pp. 25-50). Hillsdale, NJ: Lawrence Erlbaum. Assor, A., Kaplan, H., & Roth, G. (2002). Choice is good, but re levance is excellent: Autonomy-enhancing and suppressing t eacher behaviors predicting studentsâ€™ engagement in schoolwork. British Journal of Ed ucational Psychology, 72 , 261278. Atkinson, J. W. (1964). An introduction to motivation . Princeton, NJ: Van Nostrand. Bachman, J. G., & Schulenberg, J. (1993). Ho w part-time work intensity relates to drug use, problem behavior, time use, and sa tisfaction among high school seniors: Are these consequences or merely correlates? Developmental Psychology, 29 (2), 220235.
243 Bagozzi, R. P., Yi, Y., & Phillips, L. W. (1991). Assessing construct validity in organizational research. Administrative Science Quarterly, 36 , 421-458. Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84 (2), 191-215. Bandura A. (1986). Social foundations of thought and ac tion: A social cognitive theory . Englewood Cliffs, NJ: Prentice Hall. Bandura, A. (1997). Self-efficacy: The exercise of control . New York: W.H. Freeman and Company. Bandura, A., & Cervone, D. (1983). Self-e valuative and self-efficacy mechanisms governing the motivational e ffects of goal systems. Journal of Personality and Social Psychology, 45 (5), 1017-1028. Bandura, A., & Schunk, D. H. (1981). Cultivating competence, self-efficacy, and intrinsic interest through pr oximal self-motivation. Journal of Personality and Social Psychology, 41 (3), 586-598. Barrett, P. (2003, April 22). Re: Replica ting, model testing, and fixing and freeing parameters post hoc (P2) [Thr ead 96]. Message posted to http://www.aime.ua.edu/cgi-bin/wa?A1=ind0304&L=semnet . Baumgartner, H., & Homburg, C. (1996). A pplications of structural equation modeling in marketing and consumer research: A review. International Journal of Research in Marketing, 13 (2), 139-161. Bearden, L., Spencer, W., & Moracco, J. (1989). A study of hi gh school dropouts. The School Counselor, 37 , 113-120. Belton, L. (1996). What our teachers should know and be able to do: A studentâ€™s view. Educational Leadership, 54 (1), 66-68. Bentler, P. M. (1987). Drug use and persona lity in adolescence and young adulthood: Structural models with nonnormal variables. Child Development, 58 , 65-79. Bentler, P. M. (1990). Comparative fit indices in stru ctural models. Psychological Bulletin, 107 , 238-246. Bentler, P. M., & Bonett, D. G. (1980). Significance tests and goodness of fit in the analysis of covariance structures. Psychological Bulletin, 88 , 588-606. Bentler, P. M., & Chou, C. (1987). Pract ical issues in structural modeling. Sociological Methods and Research, 16 , 78-117.
244 Berndt, T. J., Laychak, A. E., & Park, K. (1990). Friendsâ€™ influence on adolescentsâ€™ academic achievement motivation: An experimental study. Journal of Educational Psychology, 82 (4), 664-670. Berry, W. D., & Feldman, S. (1985). Multiple regression in practice . Beverly Hills, CA: Sage. Birman, B.F., & Natriello, G. (1978). Perspective on absenteeism in high schools. Journal of Research and Development in Education, 11 (4), 29-38. Bishop, J. H. (1989). Why the apathy in American high schools? Educational Researcher, 18 , 6-10, 42. Black, P., & Wiliam, D. (1998). As sessment and classroom learning. Assessment in Education: Principles, Policy & Practice, 5 (1), 1-56. Blaney, P. H. (1986). Affect and memory: A review. Psychological Bulletin, 99 (2), 229246. Blumenfeld, P.C., Soloway, E., Marx, R.W., Kr ajcik, J.S., Guzdial, M., & Palincsar, A. (1991). Motivating project-based learni ng: Sustaining the doing, supporting the learning. Educational Psychologist, 26 (3&4), 369-398. Boesel, D. (2001, April). Student attitudes toward high school and educational expectations . Paper presented at the meeting of the American Educational Research Association, Seattle, WA. Bollen, K. A. (1989). Structural equations with latent variables . New York: John Wiley. Bollen, K., & Lennox, R. (1991). Conventiona l wisdom on measurement: A structural equation perspective. Psychological Bulletin, 110 (2), 305-314. Bollen, K., & Stine, R. A. (1992). Bootstra pping goodness-of-fit measures in structural equation models. Sociological Methods and Research, 21 , 205-229. Bonesteel, M. D., & Sperry, S. L. (2002). Building a better bridge. Principal Leadership (High School Ed.), 2 (9), 39-43. Bouffard-Bouchard, T. (1990). Influence of self-efficacy on performance in a cognitive task. The Journal of Social Psychology, 130 (3), 353-363. Bouffard-Bouchard, T., Parent, S., & Larive, S. (1991). Influence of self-efficacy on self-regulation and performance among juni or and senior high -school students. International Journal of Behavioral Development, 14, 153. Boyer, E. L. (1983). High school: A report on secondary education in America . New York: Harper Colophon Books.
245 Bracey, G. W. (2000). The TIMSS â€œFinal Yearâ€ study and report: A critique. Educational Researcher, 29 (4), 4-10. Bracey, G. W. (2004). Looking at adolescents through international assessments. In T. Urdan & F. Pajares (Eds.), Educating adolescents: C hallenges and strategies (pp. 131-148). Greenwich, CT: Information Age Publishing. Breckler, S. J., (1990). Applications of covariance structure m odeling in psychology: Cause for concern? Psychological Bulletin, 107 , 260-273. Brimm, J. L., Forgety, J., & Sadler, K. ( 1978). Student absenteeism: A survey report. NASSP Bulletin, 62 (415), 65-69. Bronfenbrenner, U. (1979). The ecology of human development: Experiments by nature and design . Cambridge, MA: Harvard University Press. Brookhart, S. M. (1997). A theoretical framew ork for the role of classroom assessment in motivating student effort and achievement. Applied Measurement in Education, 10 (2), 161-180. Brophy, J. (1987a). Synthesis of research on st rategies for motivation students to learn. Educational Leadership, 45 (2), 40-48. Brophy, J. (1987b). On motivating students. In D. C. Berliner & B. V. Rosenshine (Eds.), Talks to teachers (pp. 201-248). New York: Random House. Brown, B. B. (1990). Peer groups and peer cultures. In S. S. Feldman & G. R. Elliott (Eds.), At the threshold: The developing adolescent (pp. 123-146). Cambridge, MA: Harvard University Press. Brown, B. B. (1993). School culture, school politics, and the acad emic motivation of U.S. students. In T. M. Tomlinson (Ed.), Motivating students to learn (pp. 63-98). Berkeley, CA: McCutchan Publishing. Brown, I., Jr., & Inouye, D. K. (1978). Lear ned helpless through m odeling: The role of perceived similarity in competence. Journal of Personality and Social Psychology, 36 , 900-908. Browne, M. W. (1984). Asymptotically dist ribution-free methods for the analysis of covariance structures. British Journal of Mathematic al and Statistical Psychology, 37 , 62-83. Browne, M. W., & Cudeck, R. (1993). Alte rnative ways of assessing model fit. Sociological Methods and Research, 21 , 230-258. Bruner, J. (1966). Toward a theory of instruction . Cambridge, MA: Harvard University Press.
246 Butler, R. (1988). Enhancing and undermining intrinsic motivation: The effects of taskinvolving and ego-involving evaluati on on interest and performance. British Journal of Educational Psychology, 58 , 1-14. Butler, R., & Nisan, M. (1986). Effects of no feedback, task-related comments, and grades on intrinsic motivation and performance. Journal of Educational Psychology, 78 (3), 210-216. Campbell, D. T., & Fiske, D. W. (1959). Convergent and discrimi nant validation by the multitrait-multimethod matrix. Psychological Bulletin, 56 , 81-105. Campbell, J. (1982). Editorial: Some remarks from the outgoing editor. Journal of Applied Psychology, 75 , 547-560. Carey, K. (2004). A matter of degrees: Improving gradua tion rates in four -year colleges and universities (A report by the Education Trust). Retrieved April 23, 3006, from http://www2.edtrust.org/N R/rdonlyres/11B4283F-104E-4511B0CA1D3023231157/0/highered.pdf . Carey, N., & Farris, E. (1996). Parents and schools: Partners in student learning (NCES Publication No. 96-913). Washington, DC : U.S. Government Printing Office. Carnegie Corporation Council on Adolescent Development (1989). Turning points: Preparing American youth for the 21st century . New York: Author. Cavanagh, S. (2004, November 24). Governi ng board looks to mark eting to sell NAEP to seniors. Education Week, 24 (13), 26. Chadwell, D. R. (1991). Show what you know. The American School Board Journal, 178 (4), 34-35. Chapman, J. (2004, November 6). Stressed out : Senior year isnâ€™t ne cessarily the time of your life. The Charleston Gazette , p. 6C. Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Modeling, 9 , 223-255. Chin, W. W. (1998). The Partial Least Square s approach to structural equation modeling. In G. A. Marcoulides (Ed.), Modern business research methods (pp. 295-336). Mahwah, NJ: Lawrence Erlbaum. Chmelynski, C. (2003, November 11). Districts try various strategies to make senior year meaningful. School Board News, 23 . Retrieved April 22, 2006, from http://www.nsba.org/site/doc_sbn.as p?TRACKID=&VID=58&CID=1336&DID=3 2410 . Clark, R. (1983). Family life and school achievement: Why poor black children achieve and fail. Chicago: University of Chicago Press.
247 Clark, L. A., & Watson, D. (1995). Construc ting validity: Basic issu es in objective scale development. Psychological Assessment, 7 (3), 309-319. Clifford, M. M. (1988). Failure tolerance a nd academic risk-taking in tento twelveyear-old students. British Journal of Educ ational Psychology, 58 , 15-27. Clifford, M. M. (1991). Risk taking: Theoretical, empirical, and educational considerations. Educational Psychologist, 26 , 263-297. Cohen, J. (1988). Statistical power analysis for the behavioral sciences . Hillsdale, NJ: Lawrence Erlbaum Associates. Cole, J. C., Motivala, S. J., Khanna, D., Lee, J. Y., Paulus, H. E., & Irwin, M. R. (2005). Validation of a single-factor structure a nd the scoring protocol for the Health Assessment Questionnaire-Disability Index (HAQ-DI). Arthritis Care and Research, 53 , 536-542. Coleman, J. S. (1961). The adolescent society: The social life of teenag ers and its impact on education . New York: Free Press of Glencoe. Coleman, J. S., Hoffer, T., & Kilgore, S. (1982). High school achievement: Public, Catholic, and private schools compared . New York: Basic Books. Colorado School-to-Career Partnership (1999). What works! Colorado High School Senior Survey 1999: Initial results . (ERIC Document Re production Service No. ED431887) Combs, B. (1995). Reshaping change: The im plementation of the senior project in one English class. Dissertation Abstracts International, 56 (09), 3528. (University Microforms No. 9544904) Conan, N. (2005, April 27). Senioritis. Talk of the Nation . National Public Radio. Transcript retrieved July 7, 2005, from LexisNexis Academic. Conley, D. T. (2001). Reth inking the senior year. NASSP Bulletin, 85 (625), 26-41. Conley, D. T. (2003). Understanding university success: A report from Standard for Success, a project of the Association of American Universities and The Pew Charitable Trusts . Eugene, OR: Center for Educational Policy Research. Connell, J. P. (1990). Context, self, and ac tion: A motivational analysis of self-system processes across the life span. In D. Cicchetti & M. Beeghly (Eds.), The self in transition (pp. 61-98). Chicago: Univ ersity of Chicago Press. Connell, J. P., & Ryan, R. M. (1987). Self-Regulatory Style Ques tionnaire: A measure of external, introjected, identified, and in trinsic reasons for initiating behavior . Manual, University of Rochester.
248 Connell, J.P., Spencer, M.B., & Aber, J.L. ( 1994). Educational risk and resilience in African-American youth: Context, self , action, and outcomes in school. Child Development, 65 , 493-506. Connell, J. P., & Wellborn, J. G. (1991). Competence, autonomy, and relatedness: A motivational analysis of self-system proce sses. In M. R. Gunnar & L. A. Sroufe (Eds.), Minnesota Symposium on Child Psychology, Vol. 23 (pp. 43-77). Hillsdale, NJ: Lawrence Erlbaum. Cook, T. D., & Campbell, D. T. (1979). Quasi-experimentati on: Design and analysis issues for field settings . Chicago: Rand McNally. Cortina, J. M. (1993). What is coeffici ent alpha? An examination of theory and applications. Journal of Applied Psychology, 78 , 98-104. Crocker, L., & Algina, J. (1986). Introduction to classical and modern test theory . Fort Worth, TX: Harcourt Brace. Cronbach, L. J. (1950). Further eviden ce on response sets an d test design. Educational and Psychological Measurement, 10 , 3-31. Crooks, T. J. (1988). The impact of clas sroom evaluation practices on students. Review of Educational Research, 58 (4), 438-481. Crosnoe, R. (2001). Academic orientation a nd parental involvemen t in education during high school. Sociology of Education, 74 , 210-230. Crosnoe, R., Erickson, K.G., & Dornbusch, S.M. (2002). Protective functions of family relationships and school f actors on the deviant behavi or of adolescent boys and girls: Reducing the impact of risky friendships. Youth & Society, 33 , 515-544. Csikszentmihalyi, M. (1990). Flow: The psychology of optimal experience . New York: Harper & Row. Csikszentmihalyi, M. (1997). Finding flow: The psychology of engagement with everyday life . New York: Basic Books. Csikszentmihalyi, M., & Larson, R. (1987). Va lidity and reliability of the ExperienceSampling Method. Journal of Nervous and Mental Disease, 175 , 525-536. Csikszentmihalyi, M., Rathunde, K. R., & Whalen, S. (1993). Talented teenagers: The roots of success and failure . Cambridge, England: Cambridge University Press. Cushman, K. (1994). College admi ssions and the esse ntial school. Horace, 10 (5). Retrieved June 4, 2006, from the website of the Coalition of Essential Schools: http://www.essentialschools.or g/cs/resources/v iew/ces_res/134 .
249 Cushman, K. (2003). Fires in the bathroom: Advice for teachers from high school students . New York: The New Press. Cusick, P. A. (1973). Inside high school: The studentâ€™s world . New York: Holt, Reinhart, and Winston. Cyr, K. (2001, January 26). Curing senioritis or curing childhood? Retrieved June 4, 2005, from http://20below.mainetoday.com/vi ews/opinion/senioriti s012601.shtml. Darling-Hammond, L., Ancess, J., & Falk, B. (1995). Authentic assessment in action: Studies of schools and students at work . New York: Teachers College Press. Darling-Hammond, L., & Bransf ord, J. (Eds.) (2005). Preparing teachers for a changing world: What teachers should learn and be able to do . San Francisco: Jossey Bass. Darling-Hammond, L., Chajet, L., & Robertso n, P. (1996). Restructuring schools for high performance. In S. H. Fuhrman & J. A. Oâ€™Day (Eds.), Rewards and reform (pp. 144-192). San Francisco: Jossey Bass. Davidson, A., & Phelan, P. (1999). Student sâ€™ multiple worlds: An anthropological approach to understanding stude ntsâ€™ engagement with school. In T. C. Urdan (Ed.). Advances in motivation and achievement (pp. 233-273). Stamford, CT: JAI. Davies, M., & Kandel, D. B. (1981). Pare ntal and peer influences on adolescentsâ€™ educational plans: Some further evidence. American Journal of Sociology, 87, 363-387. Davis, P. (2006, April 13). Schools s eek remedy for eradicating senioritis. The Providence Journal , p. D1. Dean, P. J., Dean, M. R., & Rebalsky, R. M. (1996) Employee perceptions of workplace factors that will most improve their performance. Performance Improvement Quarterly, 9 (2), 75-89. Deci, E. L. (1975). Intrinsic motivation . New York: Plenum. Deci, E. L., & Ryan, R. M. (1985). The ge neral causality orient ations scale: Selfdetermination in personality. Journal of Research in Personality, 19 , 109-134. Deci, E. L., & Ryan, R. M. (1987). Th e support of autonomy and the control of behavior. Journal of Personality and Social Psychology, 53 (6), 1024-1037. Deci, E. L., Schwartz, A. J., Sheinman, L., & Ryan, R. M. (1981). An instrument to assess adults' orientations toward co ntrol versus autonomy with children: Reflections on intrinsic motivati on and perceived competence. Journal of Educational Psychology, 73, 642-650.
250 Deci, E. L., Vallerand, R. J., Pelletier, L. G., & Ryan, R. M. (1991). Motivation and education: The self-determination perspective. Educational Psychologist, 26 (3&4), 325-346. DeFao, J. (2005, June 9). High school seniors design course s; Innovative program designed to bypass 2nd-semester slump. The San Francisco Chronicle , p. B5. De Jonge, P., & Slaets, J. (2005). Response sets in self-report data and their associations with personality traits. European Journal of Psychiatry, 19 (4), 209-214. Delahaye, B., & Choy, S. (2000). Review of Self-Directed Learning R eadiness Scale. In J. Maltby, C. A. Lewis, & A. Hill (Eds.), Commissioned reviews of 250 psychological tests (Vol. 2). Lewiston, NY: The Edwin Mellen Press. Deslandes, R., & Bertrand, R. (2005). Motiv ation of parent invol vement in secondarylevel schooling. The Journal of Educational Research, 98 (3), 164-175. Deslandes, R., Royer, E., Bertrand, R., & Tu rcotte, D. (1997). School achievement at the secondary level: Influence of parenting style and parent involvement in schooling. McGill Journal of Education, 32 (3), 191-207. Dewey, J. (1913). Interest and effort in education . Cambridge, MA: Houghton Mifflin. Dewey, J. (1904/1965). The relation of theory to practice in edu cation. In M. L. Borrowman (Ed.), Teacher education in America: A documentary history (pp. 140171). New York: Teachers College Press. DeWitt, D. M., & Joyce, K. (2001). Merging the community and curriculum. Principal Leadership, 2 (1), 33-35. Diamantopolous, A., & Winklhofer, H. M. ( 2001). Index construction with formative indicators: An alternative to scale development. Journal of Marketing Research, 38 , 269-277. Dielman, T. E. (1994). Correction for the de sign effect in school-based substance use and abuse prevention research: Sample size requirements and analysis considerations. In A. Cazar es & L. A. Beatty (Eds.), National Institute on Drug Abuse Research Monograph 139 (pp. 115-126). Rockville, MD: National Institute on Drug Abuse. Dillman, D. A. (1978). Mail and telephone surveys: The total design method . New York: Wiley. Dillon, D. R. (1989). Showing them that I wa nt them to learn and that I care about who they are: A microethnography of the social organization of a secondary low-track English classroom. American Educational Research Journal, 26 , 227-259.
251 Donner, A., & Klar, N. (1994). Methods fo r comparing event ra tes in intervention studies when the unit of a llocation is a cluster. American Journal of Epidemiology, 140 (3), 279-289. Dâ€™Orio, W. (2002, October). Searching for the cure to senioritis. District Administration [Electronic version]. Retrieved June 4, 2005, from http://www.districtadminist ration.com/page.cfm?p=20 . Dowson, M., & Cunneen, T. (1998, April). School improvement that works: Enhancing academic achievement through motivational change: A longitudinal qualitative investigation. Paper presented at the meeting of the American Educational Research Association, San Diego, CA. Draper, N., & Smith, H. (1981). Applied regression analysis (2nd ed.). New York: Hafner. Dreis, J., & Rehage, L. ( 2006). Let seniors lead. Educational Leadership, 63 (8), 38-42. Duckworth, K., Fielding, G., & Shaugnessey, J. (1986). The relationship of high school teachersâ€™ classroom testing practices to st udentsâ€™ feelings of efficacy and efforts to study . Eugene, OR: Center for Educati onal Policy and Management, Oregon University. Duff, B. (2004). A new model for The Derryfield School â€™s Independent Senior Project . Unpublished masterâ€™s project, Univers ity of New Hampshire, Durham, NH. Dunn, K. (2001). The fourth-year itch. Teacher Magazine, 12 (7), 12-14. Eccles, J. S. (1994). Unde rstanding womenâ€™s educationa l and occupational choices. Psychology of Women Quarterly, 18 , 585-609. Eccles, J., Adler, T. F., Futterman, R., Goff , S. B., Kaczala, C. M., Meece, J., et al. (1983). Expectancies, values and academic behavior. In J. T. Spence (Ed.). Achievement and achievement motives (pp. 75-146). San Francisco: W. H. Freeman. Eccles, J.S., Early, D., Frasier, K., Belansky, E., & McCarthy, K. (1997). The relation of connection, regulation, and support for au tonomy to adolescentsâ€™ functioning. Journal of Adolescent Research, 12 (2), 263-286. Eccles, J. S., & Harold, R. D. (1993). Parent-school involvement during the early adolescent years. Teachers College Record, 94 (3), 568-587. Eccles, J. S., & Harold, R. D. (1996). Fam ily involvement in childrenâ€™s and adolescentsâ€™ schooling. In A. Booth & J. F. Dunn (Eds.), Family-school links: How do they affect educational outcomes? (pp. 3-34). Mahwah, NJ: Lawrence Erlbaum.
252 Eccles, J., Midgley, C. (1989) . Stage/environment fit: De velopmentally appropriate classrooms for young adolescents. In R. E. Ames & C. Ames (Eds.), Research on motivation and education , Vol. 3 (pp. 139-186). New York: Academic Press. Eccles, J., & Midgley, C. (1990). Changes in Academic Motivation and Self-Perception During Early Adolescence. In R. Montemayor (Ed.), Early adolescence as a time of transition (pp. 1-29). Beve rly Hills, CA: Sage. Eccles, J., Midgley, C., & Adler, T. F. ( 1984). Grade-related ch anges in the school environment: Effects on achievement motiv ation. In J. Nicholls (Ed.), Advances in motivation and achievement: The development of achievement motivation , Vol. 3 (pp. 283-331). Greenwich, CT: JAI Press. Eccles, J. S., Midgley, C., Wigfield, A., Buch anan, C. M., Reuman, D., Flanagan, C., et al. (1993). Development duri ng adolescence: The impact of stage-envi ronment fit on young adolescentsâ€™ experiences in schools and in families. American Psychologist, 48 (2), 90-101. Eccles, J. S., & Wigfield, A. (1995). In the mind of the actor: The structure of adolescentsâ€™ achievement task valu es and expectancy-related beliefs. Personality and Social Psychology Bulletin, 21 (3), 215-225. Egelson, P., Harman, S., & Bond, S. (2002, April). A preliminary study of Senior Project programs in selected southeastern high schools. Paper presented at the meeting of the American Educational Research Association, New Orleans, LA. Elawar, M. C., & Corno, L. ( 1985). A factorial experiment in teachersâ€™ written feedback on student homework: Changing teacher beha vior a little rather than a lot. Journal of Educational Psychology, 77 (2), 162-173. Emerick, L. J. (1992). Academic underachievement among the gifted: Studentsâ€™ perceptions of factors that reverse the pattern. Gifted Child Quarterly, 36 (3), 140146. Enders, C. K. (2006). Analyzing structural eq uation models with missing data. In G. R. Hancock & R. O. Mueller (Eds.), Structural equation m odeling: A second course (pp. 313-344). Greenwich, CT: Information Age Publishing. Engle, S. (2003). College freshmen spend less time studying and more time surfing the net, UCLA survey reveals . Retrieved June 3, 2006, from the UCLA Graduate School of Education and Information Studies site: http://www.gseis.ucla.edu /heri/02_press_release.pdf . Entwistle, D. R. (1990). Schools and the adoles cent. In S. S. Feldman & G. R. Elliott (Eds.), At the threshold: The developing adolescent (pp. 197-224). Cambridge, MA: Harvard University Press.
253 Epstein, J. L. (1983). The influence of frie nds on achievement and affective outcomes. In J. L. Epstein & N. Karweit (Eds.), Friends in school: Patterns of selection and influence in secondary schools (pp.163-174). New York: Academic Press. Epstein, J. L. (1984). A longitudinal study of school and family effects on student development. In S. A. Mednick, M. Harway, & K. M. Finello (Eds.), Handbook of longitudinal research: Birth and childhood cohorts , Vol. 1 (pp. 381-397). New York: Praeger. Epstein, J. L. (1986). Parentsâ€™ reactions to teacher practices of parent involvement. The Elementary School Journal, 86 (3), 277-294. Epstein, J. L. (1987). Parent involvement: Wh at the research says to administrators. Education and Urban Society, 19 , 19. Epstein, J. L. (1992). School and family partnerships. In M. Alkin (Ed.), Encyclopedia of Educational Research , 6th ed. (pp. 1139-1151). New York: MacMillan. Epstein J. L., & McPartland, J. M. (1976). The concept and measurement of the quality of school life. American Educational Research Journal, 13, 15-30. Erickson, F., & Shultz, J. (1992). Studentsâ€™ ex perience of the curriculum. In P. Jackson (Ed.), Handbook of research on curriculum (pp. 465-485). New York: MacMillan. Fan, X., Thompson, B., & Wang, L. (1999). The effects of sample size, estimation methods, and model specificat ion on SEM fit indices. Structural Equation Modeling, 6 , 56-83. Fazio, T. J., & Ural, K. K. (1995). The Princeton peer leader ship program: training seniors to help firs t-year students. NASSP Bulletin, 79 (568), 57-60. Fehrmann, P. G., Keith, T. Z., & Reimers, T. M. (1987). Home influence on school learning: Direct and indirect effects of parental involve ment on high school grades. Journal of Educational Research, 80 (6), 330-337. Feldlaufer, H., Midgley, C., & Eccles, J. S. (1988). Student, teacher, and observer perceptions of the classroom environment before and afte r the transition to junior high school. Journal of Early Adolescence, 8, 133-156. Ferry, T. R., Fouad, N. A., & Smith, P. L. (2000) . The role of family context in a social cognitive model for career-related ch oice behavior: A math and science perspective. Journal of Vocational Behavior, 57 , 348-364. Finn, J. D. (1989). Withdrawing from school. Review of Educational Research, 59 (2), 117-142. Finn, J. D. (1993). School engagement and students at risk (Report No. NCES-93-470). (ERIC Document Reproduction Service No. ED362322)
254 Finn, J. D., & Cox, D. (1992). Participati on and withdrawal among fourth-grade pupils. American Educational Research Journal, 29 (1), 141-162. Finn, J. D., Pannozzo, G. M., & Achilles, C. M. (2003). The â€œwhyâ€™sâ€ of class size: Student behavior in small classes. Review of Educational Research, 73 (3), 321368. Finn, J. D., Pannozzo, G. M., & Voelkl, K. E. (1995). Disruptive and inattentivewithdrawn behavior and achieve ment among fourth graders. The Elementary School Journal, 95 , 421-434. Finn, J. D., & Rock, D. A. (1997). Academic success among students at risk for school failure. Journal of Applied Psychology, 82 (2), 221-234. Finn, J. D., & Voelkl, K. E. (1993). School ch aracteristics related to student engagement. Journal of Negro Education, 62 (3), 249-268. Finney, S. J., & DiStefano, C. (2006). N onnormal and categorical data in structural equation modeling. In G. R. Hancock & R. O. Mueller (Eds.), Structural equation modeling: A second course (pp. 269-312). Greenwich, CT: Information Age Publishing. Firestone, W. A., & Rosenblum , S. (1988). Building commitme nt in urban high schools. Educational Evaluation and Policy Analysis, 10 (4), 285-299. Fischetti, J., Dittmer, A., & Hohmann, M. (1993). Overcoming friendly fire: Restructuring high schools to meet the needs of all students. Equity & Excellence in Education, 26 (3), 60-64. Florida Senateâ€™s Committee on E ducation (2005, September). Interim project report 2006-115: High school reform . Retrieved April 22, 2006, from http://www.flsenate.gov/data/Publications /2006/Senate/reports/interim_reports/pdf/ 2006-115ed.pdf . Floyd, F. J., & Widaman, K. F. (1995). Factor analysis in the deve lopment of clinical assessment instruments. Psychological Assessment, 7 , 286-299. Fornell, C., & Bookstein, F. L. (1982). Tw o structural equation models: LISREL and PLS applied to consumer exit-voice theory. Journal of Marketing Research, 19 , 440-452. Fouladi, R. T. (1998, April). Covariance structure analysis techniques under conditions of multivariate normality and nonnormalityâ€”Mod ified and bootstrap test statistics. Paper presented at the meeting of the Amer ican Educational Research Association, San Diego, CA.
255 Fredricks, J. A., Blumenfield, P. C., & Pa ris, A. H. (2004). School engagement: Potential of the concept, state of the evidence. Review of Educational Research, 74 (1), 59-109. Furrer, C., & Skinner, E. (2003). Sense of re latedness as a factor in childrenâ€™s academic engagement and performance. Journal of Educational Psychology, 95 (1), 148-162. Garber, S. H. (2002, April). â€œHearing their voicesâ€: Percep tions of high school students who evidence resistance to schooling . Paper presented at the meeting of the American Educational Research Association, New Orleans, LA. Geocaris, C. (1996/1997). Increasing student engageme nt: A mystery solved. Educational Leadership, 54 (4), 72-75. Gerbing, D. W., & Anderson, J. C. (1984). On the meaning of with in-factor correlated measurement errors. Journal of Consumer Research, 11 , 572-580. Gilbert, T. F. (1978). Human competence: Engineering worthy performance . New York: McGraw-Hill. Gilbert, R. N., & Robins, M. (1998). Welcome to our world: Realities of high school students . Thousand Oaks, CA: Corwin Press. Godowsky, S. H., Scarbrough, M. A., & Steinwed el, C. (1992). The senior project: An exhibition of achievement. In J. B. Podl (Ed.), The process of planning backwards: Stories from three schools (pp. 2-5). Providence, RI: Coalition of Essential Schools. Goffman, E. (1967). Interaction ritual . New York: Pantheon. Gonzalez, A. R., Doan Holbein, M. F., & Qu ilter, S. (2002). High school studentsâ€™ goal orientations and their relationship to perceived parenting styles. Contemporary Educational Psychology, 27 , 450-470. Good, T. L., & Brophy, J. E. (1997). Looking in classrooms (7th ed.). New York: Longman. Goodenow, C. (1993). Classroom bel onging among early adolescent students: Relationships to motivation and achievement. Journal of Early Adolescence, 13 (1), 21-43. Goodlad, J. I. (1984). A place called school . New York: McGraw-Hill. â€œGov. Guinn proclaims May 16-20 â€˜Rate Your Future Weekâ€™.â€ (2005, May 16). [Press release from the office of NV Governor Kenny C. Guinn.] Retrieved September 15, 2005, from http://gov.state.nv.us/pr/2005/PR_2005-0516RateYourFutureWk.htm .
256 Gregorich, S. (1998, September 25). Cross validation [Thread 32]. Message posted to http://www.aime.ua.edu/cgi-bin/wa?A1=ind9809&L=semnet . Grolnick, W. S., & Slowiaczek, M. L. (1994) . Parentsâ€™ involvement in childrenâ€™s schooling: A multidimensional conceptual ization and motivational model. Child Development, 65 , 237-252. Guglielmino, L. M. (1977). Development of the Self-Directed L earning Readiness Scale (Doctoral dissertation, Un iversity of Georgia). Dissertation Abstracts International, 38 , 6467A. Guthrie, J. T., & Wigfield, A. (2000). Engage ment and motivation in reading. In M. L. Kamil, P. B. Mosenthal, P. D. Pearson, & R. Barr (Eds.), Handbook of reading research , Vol. 3 (pp. 403-422). Mahwah, NJ: Lawrence Erlbaum. Haenlin, M., & Kaplan, A. M. (2004). A be ginnerâ€™s guide to Pa rtial Least Squares analysis. Understanding Statistics, 3 (4), 283-297. Haffey, K. R. (1995). Comba ting senioritis: The effectiven ess of an incentive program on attendance for high school seniorsâ€”A four-year perspective. Dissertation Abstracts International, 56 (05), 1598. (University Microforms No. 9528790) Haggard, E. A. (1958). Intraclass correlation and the analysis of variance . New York: The Dryden Press. Hair, J. F., Jr., Anderson, R. E., Ta tham, R. L., & Black, W. C. (1998). Multivariate data analysis (5th ed.). Upper Saddle River, NJ: Prentice Hall. Hallinan, M. T., & Williams, R. A. (1990). Studentsâ€™ characteristics and the peerinfluence process. Sociology of Education, 63 , 122-132. Hancock, G. R.. (1999). A sequential Sc heff-type respecification procedure for controlling Type I error in exploratory st ructural equation model modification. Structural Equation Modeling, 6 (2), 158-168. Harari, O., & Covington, M. V. (1981). React ions to achievement behavior from a teacher and student perspective: A developmental analysis. American Educational Research Journal, 18 , 15-28. Hardre, P. L., & Reeve, J. (2003). A motiva tional model of rural studentsâ€™ intention to persist in, versus dr op out of, high school. Journal of Educational Psychology, 95 (2), 347-356. Harter, S. (1981). A new self-report scale of in trinsic versus extrinsic orientation in the classroom: Motivational and in formational components. Developmental Psychology, 17, 300-12.
257 Harter, S. (1996). Teacher and classmate influences on scholastic motivation, selfesteem, and level of voice in adolescents. In J. Juvonen & K. R. Wentzel (Eds.), Social motivation: Understanding childrenâ€™s school adjustment (pp. 11-42). Cambridge, England: Cambridge University Press. Hartup, W. W. (1993). Adolescents and their friends. In B. Laursen (Ed.), Close friendships in adolescence: New di rections for child development , Vol. 60 (pp. 322). San Francisco: Jossey Bass. Hatcher, L. (1994). A step-by-step approach to using the SAS(r) system for factor analysis and structural equation modeling . Cary, NC: SAS Institute. Hebel, S. (2003, October 31). Virginia governor wants to inject college into the senior year of high school. The Chronicle of Higher Education , p. A22. Henderson, J., Winitzky, N., & Kauchak, D. (1996). Effective teaching in Advanced Placement classrooms. Journal of Classroom Interaction, 31 , 29-35. Henson, R. K. (2001). Understanding intern al consistency reliability estimates: A conceptual primer on coefficient alpha. Measurement and Evaluation in Counseling and Development, 34 , 177-189. Herring, H. B. (2001, June 13). One the ev e of extinction: 4 years of high school. New York Times , p. B9. Hickman, C. W., Greenwood, G., & Mill er, M. D. (1995). High school parent involvement: Relationships with achievement, grade level, SES, and gender. Journal of Research and Development in Education, 28 (3), 125-133. Hidi, S. (1990). Interest and its contri bution as a mental re source for learning. Review of Educational Research, 60 (4), 549-571. High School Survey of Student Engagement (2005a). Getting students ready for college: What student engagement data can tell us . Retrieved April 19, 2006, from http://ceep.indiana.edu/hssse/pdf/college_prep_hssse05.pdf . High School Survey of Student Engagement (2005b). What we can learn from high school students . Retrieved April 19, 2006, from http://ceep.indiana.edu/h ssse/pdf/hssse_2005_report.pdf . Hoffman, L. M. (2002). Why high schools donâ€™ t change: What students and their yearbooks tell us. The High School Journal, 86 (2), 22-37. Hoffman, L. M. (2005). Beyond high stakes testing: Rural high school students and their yearbooks. The Qualitative Report, 10 (1), 55-86.
258 Holbert, R. L., & Stephenson, M. T. ( 2002). Structural equation modeling in the communication sciences, 1995-2000. Human Communication Research, 28 (4), 531-551. Holland, J. L. (1992). Making vocational choices: A theory of careers (2nd ed.). Englewood Cliffs, NJ: Prentice Hall. Hood, A., & Egelson, P. (2005, April). A formative evaluation of a high school student assessment process. Paper presented at the meeting of the American Educational Research Association, Montreal, Canada. Hoover, E. (2003, June 27). Fighting â€œsenioritis.â€ The Chronicle of Higher Education , pp. A30-31. Hoover-Dempsey, K. V., Battiato, A. C., Walker, J. M., Reed, R.P., DeJong, J., & Jones, K. P. (2001). Parental involvement in homework. Educational Psychologist, 36 (3), 195-209. Hoover-Dempsey, K. V., & Sandler, H. M. ( 1995). Parental involveme nt in childrenâ€™s education: Why does it make a difference? Teachers College Record, 97 (2), 310330. Hoover-Dempsey, K. V., & Sandler, H. M. (2005, March 22). Final performance report for OERI Grant #R305T010673: The social c ontext of parental in volvement: A path to enhanced achievement . Retrieved November 1, 2005, from http://www.vanderbilt.edu/Pea body/family-school/Reports.html . Hox, J. J., & Bechger, T. M. (1998). An introduction to structural equation modeling. Family Science Review, 11 , 354-373. Hoyle, R. H., & Panter, A. T. (1995). Writing about structural equation models. In R. H. Hoyle (Ed.), Structural equation modeling: Concepts, issues, and applications (pp. 158-176). Thousand Oaks, CA: Sage. Hoyle, R. H., & Smith, G. T. (1994). Form ulating clinical research hypotheses as structural equation models: A conceptual overview. Journal of Consulting and Clinical Psychology, 62 (3), 429-440. Hudley, C., Daoud, A., Polanco, T., Wright-C astro, R., & Hershberg, R. (2003, April). Student engagement, school climate, and future expectations in high school . Paper presented at the Biennial Meeting of the Society for Research in Child Development, Tampa, FL. Hughes, E., & Orr, W. T., Jr. (1989). Seni or surveysâ€”A valuable planning tool for administrators. NASSP Bulletin, 73 (515), 45-51. Humes, E. (2003). School of dreams: Making the grade at a top American high school . New York: Harcourt.
259 Irwin, M. R., & Ziegler, M. R. (2005). Sleep deprivation poten tiates activation of cardiovascular and catecholamine res ponses in abstinent alcoholics. Hypertension, 45 (2), 252-257. Jamison, K. A. (2003). The effects of trained facilitati on of learning-oriented feedback on learner engagement, performance, self-efficacy, and enjoyment . Unpublished doctoral dissertation, Virginia Polytechnic Institute and State University. Retrieved June 4, 2006, from http://scholar.lib.vt.edu/ theses/available/etd-04272004220646/unrestricted/JamisonKDissertation.pdf . Jarvis, C. B., Mackenzie, S. B., & Podsakoff, P. M. (2003). A critical review of construct indicators and measurement model missp ecification in marketing and consumer research. Journal of Consumer Research, 30 , 199-218. Jason, L. A., & Burrows, B. (1983). Tran sition training for high school seniors. Cognitive Therapy and Research, 7 (1), 79-92. Jayson, S. (2005, April 19). With â€œsenioritisâ€ the diagnosis, the search for a cure is on. USA Today , p. 5D. Jester, A., & Ismail, R. H. (2006, February 20). State wants high school grads ready for next step: Too many need remedial work in college. The Lexington Herald-Leader . Retrieved from Lexis-Nexis Academic. Johnson, J., & Farkas, S. (1997). Getting by: What American teenagers really think about their schools (A Report from Public Agenda). New York: Public Agenda. Johnson-Reid, M., Davis, L., Saunders, J., Williams, T., & Williams, J. H. (2005). Academic self-efficacy among African Amer ican youths: Implications for school social work practice. Children and Schools, 27 (1), 5-14. Jreskog, K. G. (1993). Testing structural equation models. In K. A. Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 294-316). Newbury Park, CA: Sage. Jreskog, K. G., & Srbom, D. (1981). LISREL V: Analysis of linear structural relationships by the method of maximum likelihood . Chicago: National Educational Resources. Jreskog, K. G., & Srbom, D. (1986). LISREL VI: Analysis of linear structural relationships by maximum likelihood, inst rumental variables, and least squares methods (4th ed.). Mooresville, IN: Scientific Software. Judd, C. M., McClelland, G. H., & Culhane, S. E. (1995). Data analysis: Continuing issues in the everyday analys is of psychological data. Annual Review of Psychology, 46 , 433-465.
260 Juvonen, J., & Murdock, T. B. (1993). How to promote social a pproval: Effects of audience and achievement outcome on pub licly communicated attributions. Journal of Educational Psychology, 85 , 365-376. Juvonen, J., & Murdock, T. B. (1995). Gradelevel differences in the social value of effort: Implications for self-presentation tactics of early adolescents. Child Development, 66 , 1694-1705. Kadyszewski, H. (2003, March 4). Serious se nioritis? Fewer stude nts say courses are meaningful. The Christian Science Monitor , p. 16. Kanevsky, L., & Keighley, T. (2003). To produce or not to produce? Understanding boredom and the honor in underachievement. Roeper Review, 26 (1), 20-28. Kantrowitz, B., & Wingert, P. (2000, December 11). Curing senioritis. Newsweek, 136 (24), 60-61. Karweit, N. (1973). Rainy days and Mondays . (ERIC Document Reproduction Service No. ED086927) Kastner, L. S., & Wyatt, J. (2002). The launching years: Strate gies for parenting from senior year to college life . New York: Three Rivers Press. Keith, T. Z., Keith, P. B., Troutman, G. M., Bickley, P. G., Trivette, P. S., & Singh, K. (1993). Does parental invol vement affect eighth grad e studentsâ€™ achievement? Structural analysis of national data. School Psychology Review, 22 , 474-496. Keith, T. Z., Reimers, T. M., Fehrmann, P. G ., Pottebaum, S. M., & Aubey, L. W. (1986). Parental involvement, homework, and TV time: Direct and indirect effects on high school achievement. Journal of Educational Psychology, 78 , 373-380. Keller, B. (2003, April 23). Less than awesome. Education Week, 22 (32), 3-6. Keller, J. M. (1987). Development and use of the ARCS model of motivational design. Journal of Instructional Development, 10 (3), 2-10. Kelly, K. (2001). Seeking a cu re for senior-year slump. Harvard Education Letter, 17 (4), 1-4. Kemmery, R. J., & Cook, H. J. (2002) . Bridge to the â€œreal world.â€ Principal Leadership (High School Ed.), 3 (3), 44-48. Kenny, D. A., & La Voie, L. (1985). Separating individual and group effects . Journal of Personality and Social Psychology, 48 (2), 339-348. Kessler, R. (1999/2000). Initiation: Saying good-bye to childhood. Educational Leadership, 57 (4), 30-33.
261 Kessler, S. (1996). The â€œSenior Passageâ€ course . In L. C. Mahdi, N. G. Christopher, & M. Meade (Eds.), Crossroads: The quest for contemporary rites of passage (pp. 185-197). Peru, IL: Open Court Publishing. Kirsch, A. (1997, May 13). High schools put some spring into senior year. The Christian Science Monitor , p. 15. Kirst, M. W. (2001). Overcoming the high school senior slump: New education policies (Report No. K-16-R-01-01). (ERIC Document Reproduction Service No. ED455720) Kish, L. (1965). Survey sampling . New York: John Wiley & Sons. Klem, A. M., & Connell, J. P. (2004). Rela tionships matter: Linking teacher support to student engagement and achievement. Journal of School Health, 74 (7), 262-273. Kline, R. B. (2005). Principles and practice of st ructural equation modeling (2nd ed.). New York: Guilford Press. Kline, R. B. (2006). Reverse arrow dynamics: Formative measurement and feedback loops. In G. R. Hancock & R. O. Mueller (Eds.), Structural equation modeling: A second course (pp. 43-68). Greenwich, CT: Information Age Publishing. Kline, T., & Klammer, J. (2001). Path m odel analyzed with ordinary least squares multiple regression versus LISREL. The Journal of Psychology, 135 (2), 213-225. Kline, T., Sulsky, L. M., & Rever-Moriyama, S. D. (2000). Common method variance and specification errors: A practi cal approach to detection. The Journal of Psychology, 134 (4), 401-421. Kohn, A. (1993). Choice for children: Why and how to let students decide. Phi Delta Kappan, 75 (1), 8-20. Kopp, T. (1982). Designing boredom out of instruction. Performance and Instruction, 21 (4), 23-27. Kuhlthau, C. C. (1985). A process approach to library skills instruc tion: An investigation into the design of the library research process. School Library Media Quarterly, 13 (1), 35-40. Laffey, J. M. (1982). The assessment of involvement with school work among urban high school students. Journal of Educational Psychology, 74 (1), 62-71. Lance, C. E., Butts, M. M., & Michels, L. C. (2006). The sources of four commonly reported cutoff criteria: What did they really say? Organizational Research Methods, 9 (2), 202-220.
262 Langdon, C.A. & Vesper, N. (2000). The sixth Phi Delta Kappa poll of teachersâ€™ attitudes towards th e public schools. Phi Delta Kappan, 81 (8), 607-611. Larson, R., & Verma, S. (1999). How children and adolescents spend time across the world: Work, play, and developmental opportunities. Psychological Bulletin, 125, 701. Lee, V. E., & Smith, J. B. (1994). High sc hool restructuring and student achievement: A new study finds strong links. Issues in Restructuring Schools (Report No. 7), 1-18. Lee, V. E., & Smith, J. B. (1995). Effects of high school restructuring and size on early gains in achievement and engagement. Sociology of Education, 68 , 241-270. Lepper, M. R. (1988). Motivational consid erations in the study of instruction. Cognition and Instruction, 5 (4), 289-309. Linnenbrink, E. A., & Pintrich, P. R. (2003). The role of self-efficacy beliefs in student engagement and learning in the classroom. Reading & Writing Quarterly, 19 , 119137. Linton, T. E., & Pollack, E. W. (1978). Bore dom transcended: Adolescent survival in the suburban high school. The High School Journal, 62 (2), 69-72. Lippard, J. F. (2000). The senior project: Combating â€œsenioritisâ€ and ensuring student accountability . Report for the Massachusetts Ch arter School Fellowship Program. Retrieved March 12, 2004, from http://www.masscharterschools.org/ fellowships/docs/000064/index.html . Little, R., & Rubin, D. (1989). The analysis of social science data with missing values. Sociological Methods and Research, 18 , 292-326. Locke, E. A., & Latham, G. P. (1990). A theory of goal-setting and task performance . Englewood Cliffs, NJ: Prentice-Hall. Lombardi, K. S. (2003, May 25). When senioritis strikes. The New York Times , p.WC1. Lopez, L. (2004, October 4). Senior Project: Effectiven ess study in South Carolina . [Pilot year report (2003-2004) submitted to SERVE.] Unpublished report obtained from Dr. Paula Egelson on July 26, 2005. Lopez, F. G., & Lent, R. W. (1992). Sources of mathematics self-efficacy in high school students. Career Development Quarterly, 41 (1), 3-11. Lord, M. (2001, May 28). More calculus? Toss the frisbee! U.S. News & World Report, 130 (21), 44. Lorenz, S. (1999). The senior proj ect: The best idea I ever stole. NASSP Bulletin, 83 (609), 77-85.
263 Lott, J. G. (1995). When kids da re to question th eir education. Educational Leadership, 52 (7), 38-42. Lowe, M. A. (2003). Inside Taylor High: The phenom enology of student engagement during high school restructuring . Baltimore: PublishAmerica. MacCallum, R. C., & Browne, M. W. (1993). The use of causal indi cators in covariance structure models: Some practical issues. Psychological Bulletin, 114(3), 533-541. MacCallum, R. C., Browne, M. W., & Suga wara, H. M. (1996). Power analysis and determination of sample size for covariance modeling. Psychological Methods, 1 , 130-149. MacCallum, R. C., Roznowski, M., & Necow itz, L. B. (1992). Model modifications in covariance structure analysis: The pr oblem of capitalization on chance. Psychological Bulletin, 111 , 490-504. MacIver, D. J., Stipek, D. J., & Daniels, D. H. (1991). Explaining within-semester changes in student effort in junior high school and senior high school courses. Journal of Educational Psychology, 83 (2), 201-211. MacKenzie, S. B., Podsakoff, P. M., & Jarvis, C. B. (2005). The problem of measurement model misspecification in beha vioral and organizational research and some recommended solutions. Journal of Applied Psychology, 90 (4), 710-730. Madon, S., Smith, A., Jussim, L., Russell, D. W ., Eccles, J., Palumbo, P., et al. (2001). Am I as you see me or do you see me as I am? Self-fulfilling prophecies and selfverification. Personality and Social Psychology Bulletin, 27 (9), 1214-1224. Maehr, M. L. (1976). Continuing motivation: An analysis of a seldom considered educational outcome. Review of Educational Research, 46 , 443-462 Manzo, K. K. (2003, July 10). NAEP writing scores improve, but not for seniors. Education Week [Electronic version]. Re trieved July 7, 2005 from www.edweek.org/ew/articles /2003/07/10/42naep_web.html . Mardia, K. V. (1970). Measures of multivaria te skewness and kurtosis with applications. Biometrika, 57 , 519-530. Marks, H. M. (1995). Student engagement in the classrooms of restructuring schools . Madison, WI: Center on Reorganization and Restructuring of Schools. (ERIC Document Reproduction Se rvice No. ED381884) Marks, H. M. (2000). Student engagement in instructiona l activity: Patterns in the elementary, middle, and high school years. American Educational Research Journal, 37 (1), 153-184.
264 Marsh, D. D., & Codding, J. B. (1998). The new American high school . Thousand Oaks, CA: Corwin. Marsh, H. W. (1996). Positive and nega tive global self-esteem: A substantively meaningful distincti on or artifactors? Journal of Personality and Social Psychology, 70 , 810-819. Marsh, H. W., Balla, J. R., & MacDonald, R. P. (1988). Goodness-of-fit indexes in confirmatory factor analysis: The effect of sample size. Psychological Bulletin, 103 , 391-410. Martin, A. J. (2005). Exploring the eff ects of a youth enrichment program on academic motivation and engagement. Social Psychology of Education, 8 , 179-206. Martin, J. A. (1987). Structural equati on modeling: A guide for the perplexed. Child Development, 58 , 33-37. Mather, M., Shafir, E., & Johnson, M. K. (2000). Misremembrance of options past: Source monitoring and choice. Psychological Science, 11 (2), 132-138. Mathews, J. (1998, May 27). Schools treat â€œsenior slump.â€ The Washington Post , p. B1. Mayer, R. E. (1998). Cognitive, metacogni tive, and motivational aspects of problem solving. Instructional Science, 26 , 49-63. Mayher, B. (1998). The college admissions mystique . New York: Noonday. McCarthy, M., & Kuh, G. D. (2005, Septem ber 9). Student engagement: A missing link in improving high schools. Teachers College Record [Electronic version]. Retrieved November 2, 2005, from http://www.tcrecord.org/Content.asp?ContentID=12162 . McCombs, B. L., & Whisler, J. S. (1989). Th e role of affective variables in autonomous learning. Educational Psychologist, 24 (3), 277-306. McDermott, P. A., Mordell, M., & Stolzfus, J. C. (2001). The organization of student performance in American schools: Discip line, motivation, ve rbal and non-verbal learning. Journal of Educational Psychology , 93 (1), 65. McDonald, R. P., & Ho, M. R. (2002). Prin ciples and practice in structural equation analyses. Psychological Methods, 7 (1), 64-82. McLain, B. (2002, April). Strategies to engage students in learning: Results of a statewide survey of public high schools in Washington State . Paper presented at the meeting of the American Educational Re search Association, New Orleans, LA. McLaughlin, M. W. (1994). Somebody knows my name. Issues in Restructuring Schools (Report No. 7), 9-11.
265 McNeil, L. (1984). Lowering expectations: The impac t of student employment on classroom knowledge . Madison, WI: Wisconsin Cent er for Educational Research. McQuillan, P. J. (1997). Humanizing the comprehensive high school: A proposal for reform. Educational Administration Quarterly, 33 , 644-682. McWhirter, E. H., Bandalos, D. L., & H ackett, G. (1998). A causal model of the educational plans and career e xpectations of Mexican American high school girls. Journal of Counseling Psychology, 45 (2), 166-181. Meier, D. (1995). The power of their ideas: Lessons for America from a small school in Harlem . Boston: Beacon Press. Messick, S. (1962). Response st yle and content measures from personality inventories. Educational and Psychological Measurement, 22 , 1-17. Michelau, D. K. (2002). Whatâ€™s the cure for senioritis? State Legislatures, 28 (6), 29-32. Midgley, C., & Feldlaufer, H. (1987). Student sâ€™ and teachersâ€™ deci sion-making fit before and after the transition to junior high school. Journal of Early Adolescence, 7(2), 225-241. Midgley, C., Feldlaufer, H., & Eccles, J. S. (1989). Change in teacher efficacy and student selfand task-related beliefs in mathematics during the transition to junior high school. Journal of Educational Psychology 81 (2), 247-258. Miller, B. S. (1972). The humanities approach to the modern secondary school curriculum . New York: The Center for A pplied Research in Education. Miller, R. (2001). Lesser and greater expectations: The wasted senior year and collegelevel study in high school (Report No. HE 035 275). Washington, DC: Association of American Colleges and Universities . (ERIC Document Reproduction Service No. ED469019) Miller, R. B., Greene, B. A., Montalvo, G. P ., Ravindran, B., & Nichols, J. D. (1996). Engagement in academic work: The role of learning goals, future consequences, pleasing others, and perceived ability. Contemporary Educational Psychology, 21 , 388-422. Mitchell, J. (2005, February 20). Hartford schools may toughen graduation requirements; Officials consider adding courses, senior project. The Baltimore Sun , p. 20B. Mitchell, T. R. (1985). An evaluation of the validity of correlational research conducted in organizations. Academy of Management Review, 10 , 192-205. Morgan, M. (1985). Self-monitoring of attained subgoals in private study. Journal of Educational Psychology, 77 (6), 623-630.
266 Mounts, N. S., & Steinberg, L. D. (1995). An ecological analysis of peer influence on adolescent grade-point-average and drug use. Developmental Psychology, 31 , 915922. Mulaik, S. (2003, April 23). Replications and cross-validations [Thread 104]. Message posted to http://www.aime.ua.edu/cgi-bin/wa?A1=ind0304&L=semnet . Mulaik, S. (2004, January 8). Cross-validati on procedures [Thread 39]. Message posted to http://www.aime.ua.edu/cgi-bin/wa?A1=ind0401&L=semnet . Muller, C. (1998). Gender differences in parental involvem ent and adolescentsâ€™ mathematics achievement. Sociology of Education, 71 , 336-356. Mullis, I. (1998). Mathematics and science achievement in the final year of secondary school: IEAâ€™s Third International Mathematics an d Science Study . Chestnut Hill, MA: Boston College. Murdock, T. B., Anderman, L. H., & Hodge, S. A. (2000). Middle-grade predictors of studentsâ€™ motivation and be havior in high school. Journal of Adolescent Research, 15 (3), 327-351. Murphy, J. (1991). Restructuring schools: Captur ing and assessing the phenomenon . New York: Teachers College Press. Murphy, J., Beck, L. G., Crawford, M., Hodges, A., & McGaughy, C. L. (2001). The productive high school: Creating personalized academic communities . Thousand Oaks, CA: Corwin. Muthn, L. K. (2006, March 29). Large and varied design effects [Msg. 2]. Message posted to http://www.statmodel.com/cgi-bin/discus/discus.cgi . Muthn, L. K., & Muthn, B. O. (1998-2006). Mplus userâ€™s guide (4th ed.). Los Angeles, CA: Muthn & Muthn. National Assessment Governing Board (2004). Governing board awards contract to StandardsWork, Inc., to improve sch ool and student participation in 12th grade NAEP . Retrieved July 8, 2005 from http://www.nagb.org/release/award_1004.html . National Center for Edu cation Statistics (2002). The condition of education 2002 (Report NCES 2002). Washington, DC: U.S. G overnment Printing Office. Retrieved April 22, 2006, from http://nces.ed.gov/pubs2002/2002025.pdf . National Commission on Excellence in Education (1983). A nation at risk: The imperative for educational reform . Washington, DC: U.S. Department of Education.
267 National Commission on NAEP 12th Grade Assessment and Reporting (2004). 12th grade student achievement in America: A new vision for NAEP (A report to the National Assessment Governing Board). Retrieved March 1, 2006, from http://www.nagb.org/releas e/12_gr_commission_rpt.pdf . National Commission on the High Sc hool Senior Year (2001a). The lost opportunity of senior year: Finding a better way . Princeton, NJ: The Woodrow Wilson National Fellowship Foundation. National Commission on the High Sc hool Senior Year (2001b). Raising our sights: No high school senior left behind . Princeton, NJ: The Woodrow Wilson National Fellowship Foundation. National Governors Association (2005, July). Summary of RateYourFuture.org survey findings . Paper presented at the annual meeting of the National Governors Association, Des Moines, Iowa. Retrieved April 25, 2006, from http://www.nga.org/Files/ppt/RATE YOURFUTURESURVEY.PPT . National Research Council and In stitute of Medicine. (1998). Protecting youth at work . Washington, DC: The Nati onal Academies Press. National Research Council and In stitute of Medicine. (2004). Engaging schools: Fostering high school student sâ€™ motivation to learn . Washington, DC: The National Academies Press. Natriello, G. (1984). Problems in the eval uation of students and student disengagement from secondary schools. Journal of Research and Development in Education, 17 (4), 14-24. Natriello, G. (1987). The impact of evaluation processes on students. Educational Psychologist, 22 (2), 155-175. Natriello, G., & McDill, E. L. (1986). Pe rformance standards, student effort on homework, and academic achievement. Sociology of Education, 59 , 18-31. Nevin, J. A. (1988). Behavioral momentum and the partial reinforcement effect. Psychological Bulletin, 103 , 44-56. Nevitt, J., & Hancock, G. R. (2000). Im proving the root mean square error of approximation for nonnormal conditions in structural equati on modeling. Journal of Experimental Education, 68 , 251-268. Nevitt, J., & Hancock, G. R. (2001). Perf ormance of bootstrapping approaches to model test statistics and parameter standard error estimation in structural equation modeling. Structural Equation Modeling , 8 , 353-377. Newmann, F. M. (1981). Reduc ing student alienation in hi gh school: Implications of theory. Harvard Educational Review, 51 (4), 546-564.
268 Newmann, F. M. (1989). Student en gagement and high school reform. Educational Leadership, 46 (5), 34-36. Newmann, F. M. (1998). How secondary schools contribute to academic success. In K. Borman and B. Schneider (Eds.), The adolescent years: Social influences and educational challenges (pp. 88-108). Chicago: Univ ersity of Chicago Press. Newmann, F. M., Marks, H. M., & Gamoran, A. (1995, April). Authentic pedagogy and student performance . Paper presented at the meeting of the American Educational Research Association, San Francisco, CA. Newmann, F. M., & Wehlage, G. G. (1995). Successful school restru cturing: A report to the public and educators by the Center on Organization and Restructuring of Schools. Washington, DC: American Federation of Teachers. (ERIC Document Reproduction Service No. ED387925) Newmann, F. M., Wehlage, G. G., & Lam born, S. D. (1992). The significance and sources of student engagement. In F. M. Newmann (Ed.), Student engagement and achievement in American secondary schools (pp. 11-39). New York: Teachers College Press. Nicolini, M. B. (1999). Pictures of an exhibi tion: Senior graduation projects as authentic research. English Journal, 89 (1), 91-98. Nunnally, J. C. (1967). Psychometric theory . New York: McGraw-Hill. Nunnally, J. C. (1978). Psychometric theory (2nd ed.). New York: McGraw-Hill. Nunnally, J. C., & Bernstein, I . H. (1994). Psychometric theory (3rd ed.). New York: McGraw-Hill. Oâ€™Grady, A. (1999). Information literacy skills and the senior project. Educational Leadership, 57 (2), 61-62. Page, R. N. (1999). The uncertain value of school knowledge: Biology at Westridge High. Teachers College Record, 100 (3), 554-601. Pajares, F. (1996). Self-efficacy beliefs in academic settings. Review of Educational Research, 66 (4), 543-578. Pajares, F. (2002). Gender and perceived se lf-efficacy in self-regulated learning. Theory Into Practice , 41 (2), 116-125. Pajares, F., & Kranzler, J. (1995). Self-effi cacy beliefs and general mental ability in mathematical problem-solving. Contemporary Educational Psychology, 30 , 426443.
269 Pajares, F., & Miller, M. D. (1994). Role of self-efficacy and self-concept beliefs in mathematical problem solving: A path analysis. Journal of Educational Psychology, 86 (2), 193-203. Palmer, R. F., Graham, J. W., Taylor, B., & Tatterson, J. (2002). Construct validity in health behavior research: Interpreting la tent variable models involving self-report and objective measures. Journal of Behavioral Medicine, 25 (6), 525-550. Pannwitt, B. (1979). A changing, more challenging twelfth grade: Antidotes for senioritis (Report No. EA012161). Reston, VA: National Association of Secondary School Principals. (ERI C Document Reproduction No. ED176421) Paris, S. G., & Turner, J. C. (1995). Situat ed motivation. In P. R. Pintrich (Ed.), Student motivation and cognition (pp. 213-237). Mahwah, NJ: Lawrence Erlbaum. Parizek, D., & Kevasan, S. (2000). From classrooms to careers: The Senior Mastery Process at Henry Ford Academy . (ERIC Document Re production Service No. ED465913) Parker, L. O. (1992). Easing the graduation tr ansition: Career seminars for seniors. The School Counselor, 39 (5), 394-398. Parsand, B., & Lewis, L. (2003). Remedial education at de gree-granting postsecondary institutions in fall 2000. U.S. Department of Educa tion, National Center for Education Statistics. Washington, DC: U.S. Government Printing Office. Paulhus, D. L. (1984). Two-component mode ls of socially desirable responding. Journal of Personality and Social Psychology, 46 , 598-609. Pedersen, S. (2003). Motivationa l orientation in a problem-bas ed learning environment. Journal of Interactive Learning Research, 14 (1), 51-77. Perie, M., & Moran, R. (2005). NAEP 2004 trends in academic progress: Three decades of student performance [NCES 2005]. U.S. Department of Education, National Center for Education Statistic s. Washington, DC: U.S. Government Printing Office. Peter D. Hart Research Associates (2005). Rising to the challenge: Are high school graduates prepared for college and work ? A study of recent high school graduates, college instructors, and employers for Achieve, Inc . Retrieved May 12, 2006, from http://www.achieve.org /files/pollreport.pdf . Peterson, K. (2003). Overcoming senior slump: The community college role . (Report No. JC030348). Washington, DC: Office of Edu cational Research and Improvement. (ERIC Document Reproduction Service No. ED477830) Peterson, K. (2005, July 14). Push to reform high schools gaining . Retrieved July 15, 2005, from http://www.stateline.org .
270 Pintrich, P. R. (1999). The role of motiva tion in promoting and sustaining self-regulated learning. International Journal of Educational Research, 31 , 459-470. Pintrich, P. R. (2003). Motiva tion and classroom learning. In W. M. Reynolds & G. E. Miller (Eds.), Handbook of psychology (Vol. 7): Educational psychology (pp. 103122). New York: Riley. Pintrich, P. R., & De Groot, E. V. (1990) . Motivational and self-regulated learning components of classroom academic performance. Journal of Educational Psychology, 82 (1), 33-40. Pintrich, P. R., & Schrauben, B. (1992). Studentsâ€™ motivational beliefs and their cognitive engagement in classroom tasks. In D. H. Schunk & J. L. Meece (Eds.), Student perceptions in the classroom (pp. 149-184). Hillsdale, NJ: Lawrence Erlbaum. Pope, J. (2005, December 18). Early d ecision on college is popular again. Associated Press Online . Retrieved January 4, 2006, from Lexis-Nexis Academic. Porter, L., & Lawler, E. (1968). Managerial attitudes and performance . Homewood, IL: Irwin-Dorsey. Powell, A. G. (1996). Motivating students to learn: An American dilemma. In S. H. Fuhrman & J. A. Oâ€™Day (Eds.), Rewards and reform (pp. 19-59). San Francisco: Jossey-Bass. Powell, A. G., Farrar, E., & Cohen, D. K. (1985). The shopping mall high school . Boston: Houghton Mifflin. Ramsden, P., & Entwistle, N. J. (1981). E ffects of academic departments on studentsâ€™ approaches to studying. British Journal of Educational Psychology, 51 , 368-383. Ramsden, P., Martin, E., & Bowden, J. ( 1989). School environment and sixth form pupilsâ€™ approaches to learning. British Journal of Educational Psychology, 59 , 129-142. Randhawa, B. S., Beamer, J. E., & Lundberg, I. (1993). Role of mathematics selfefficacy in the structural model of mathematics achievement. Journal of Educational Psychology, 85 (1), 41 -48 Reeve, J., Jang, H., Carrell, D., Jeon, S., & Barch, J. (2004). Enhancing studentsâ€™ engagement by increasing t eachersâ€™ autonomy support. Motivation and Emotion, 28 (2), 147-169. Renninger, K. A., Hidi, S., & Krapp, A. (Eds.). (1992). The role of interest in learning and development . Hillsdale, NJ: Lawrence Erlbaum. Resnick, L. B. (1987). L earning in school and out. Educational Researcher, 16 , 13-20.
271 Riedel, J. A. (2001). Academic engage ment in two Re:learning high schools. Dissertation Abstracts International, 63 (01), 46. (University Microforms No. 3038330) Rigdon, E. (2006a, May 3). Re: Question abou t Mplus and categorical variables [Thread 119]. Message posted to http://bama.ua.edu/cgi-bin/wa?A1=ind0605&L=semnet . Rigdon, E. (2006b, June 2). Re: Cross Valida tion . . . play it again Sam [Thread 33]. Message posted to http://www.aime.ua.edu/cg i-bin/wa?A1=ind0606&L=semnet . Riley, R. W. (2000, Feb. 22). Seventh a nnual state of American education address: Setting new expectations. Speech delivere d February 22, 2000, at Southern High School in Durham, NC. Retrieved April 22, 2006, from http://www.ed.gov/Speeches/02-2000/000222.html . Rinne, C. H. (1998). Motivating st udents is a percentage game. Phi Delta Kappan, 79 (8), 620-626. Rocklein, M. W. (1994). Making connectio ns: A high school-college partnership. NASSP Bulletin, 78 (565), 110-116. Rosenbaum, J. E. (1998). College-for-al l: Do students understand what college demands? Social Psychology of Education, 2 , 55-80. Rosenberg, M. (1997, June 22). Comm unity work eases senior slump. The New York Times , p. 13. Rosenthal, R., & Jacobson, L. (1968). Teacher expectation a nd pupilsâ€™ intellectual development . New York: Irvington. Roth, J., & Damico, S. B. (1994, April). Broadening the concep t of engagement: Inclusion of perspectives on adolescence . Paper presented at the meeting of the American Educational Research Association, New Orleans, LA. Rotter, J. B. (1966). Generalized expectancies for internal versus external control of reinforcement. Psychological Monographs, 80(1). (Whole No. 609). Ryan, A. M. (2000). Peer groups as a cont ext for the socializa tion of adolescentsâ€™ motivation, engagement, and achievement in school. Educational Psychologist, 35, 101-111. Ryan, A. M., & Patrick, H. (2001). The cl assroom social environment and changes in adolescentsâ€™ motivation and e ngagement during middle school. American Educational Research Journal, 38 (2), 437-460.
272 Ryan, R. M., Connell, J. P., & Deci, E. L. (1985). A motivational analysis of selfdetermination and self-regul ation in education. In C. Ames & R. Ames (Eds.), Research on motivation in educati on, Vol. 2: The classroom milieu (pp. 13-51). New York: Academic Press. Ryan, R. M., & Deci, E. L. (2000). Self-d etermination theory a nd the facilitation of intrinsic motivation, social development, and well-being. American Psychologist, 55 (1), 68-78. Ryan, R. M., & Powelson, C. L. (1991). Autonomy and relatedness as fundamental to motivation and education. Journal of Experimental Education, 60 (1), 49-66. Sadler, D. R. (1989). Formative assessment and the design of instructional systems. Instructional Science, 18 , 119-144. Sandberg, V. C. (1981). Administeri ng curriculum change: A case study. Dissertation Abstracts International, 42 (06), 2424. (University Microforms No. 8123289) Sandel, L. (1990). Teach us (then) what we need to know (now) . (Report No. CS010236). Hampstead, NY: Hofstra Univer sity. (ERIC Document Reproduction Service No. ED 323 521) Sarason, S. B. (1990). The predictable failure of educational reform. San Francisco: Jossey Bass. Sax, L. J., Astin, A. W., Korn, W. S., & Mahoney, K. M. (1997). The American freshman: National norms for fall 1997 . Los Angeles: University of California-Los Angeles. Schafer, J. L. (2000). NORM: Multiple imputat ion of incomplete multivariate data under a normal model (Version 2.03) [Comput er software]. University Park: Pennsylvania State University, Department of Statistics. Schafer, J. L., & Graham, J. W. (2002). Missi ng data: Our view of th e state of the art. Psychological Methods, 7 , 147-177. Schaffer, S. P., & Richardson, J. C. (2004) . Supporting technology integration within a teacher education system. Journal of Educational Computing Research, 31 (4), 423-435. Schaufeli, W. B., Martnez, I. M., Pinto, A. M., Salanova, M., & Bakker, A. B. (2002). Burnout and engagement in university students: A cross-national study. Journal of Cross-Cultural Psychology, 33 (5), 464-481. Schiefele, U. (1991). Interest , learning, and motivation. Educational Psychologist, 26 (34), 299-323.
273 Schlenker, B. R. (1980). Impression management: The self -concept, social identity, and interpersonal relations . Monterey, CA: Brooks Cole. Schmitt, N. (1994). Method bias: The impor tance of theory and measurement. Journal of Organizational Behavior, 15 , 393-398. Schumacker, R. E., & Lomax, R. G. (2004). A beginnerâ€™s guide to structural equation modeling (2nd ed.). Mahwah, NJ: Lawrence Erlbaum. Schunk, D. H. (1981). Modeling and attributio nal effects on childre nâ€™s achievement: A self-efficacy analysis. Journal of Educational Psychology, 73 , 93-105. Schunk, D. H. (1983). Developing childrenâ€™s se lf-efficacy and skills: The roles of social comparative information and goal setting. Contemporary Educational Psychology, 8 , 76-86. Schunk, D. H. (1984). Enhancing self-eff icacy and achievement through rewards and goals: Motivational and informational effects. Journal of Educational Research, 78 , 29-34. Schunk, D. H. (1985). Self-effi cacy and classroom learning. Psychology in the Schools, 22 , 208-222. Schunk, D. H. (1989). Self-efficacy and achievement behaviors. Educational Psychology Review, 1 , 173-208. Schunk, D. H. (1991). Self-efficacy and academic motivation. Educational Psychologist, 26 (3&4), 207-231. Schunk, D. H. (1995). Self-efficacy and edu cation and instruction. In J. E. Maddux (Ed.), Self-efficacy, adaptation, and adjustme nt: Theory, research, and application (pp. 281-303). New York: Plenum Press. Schunk, D. H., & Miller, S. D. (2002). Self-e fficacy and adolescentâ€™s motivation. In F. Pajares & T. Urdan (Eds.), Academic motivation of adolescents (pp. 29-52). Greenwich, CT: Information Age Publishing. Schunk, D. H., & Pajares, F. (2002). The deve lopment of academic self-efficacy. In A. Wigfield & J. S. Eccles (Eds.), Development of ac hievement motivation (pp. 1631). San Diego: Academic Press. Schunk, D. H., & Swartz, C. W. (1993). Go als and progress feedb ack: Effects on selfefficacy and writing achievement. Contemporary Educational Psychology, 18 , 337-354. Sedlak, M. W., Wheeler, C. W., Pullin, D. C., & Cusick, P. A. (1986). Selling students short: Classroom bargains and academic reform in the American high school . New York: Teachers College Press.
274 Seeley, D. S. (1989). A new para digm for parent involvement. Educational Leadership, 47 (2), 46-48. Shar, D. (2005, January 17). Senioritis: Need for good grades keeps ailment in check. The Winston-Salem Journal , p. D1. Shaunessy, E. (2004). The senior project and gifted education. Gifted Child Today, 27 (3), 38-51. Shell, D. F., Murphy, C. C., & Bruning, R. H. (1989). Self-efficacy and outcome expectancy mechanisms in read ing and writing achievement. Journal of Educational Psychology, 81 , 91-100. Shernoff, D. J., Csikszentmihalyi, M., Schneid er, B., & Shernoff, E. S. (2003). Student engagement in high school classrooms fr om the perspective of flow theory. School Psychology Quarterly, 18 (2), 158-176. Shernoff, D. J., & Hoogstra, L. (2001). Continuing motivation beyond the high school classroom. New Directions for Child and Adolescent Development, 93 , 73-87. Shulman, L. S. (1986). Paradigms and research programs in the study of teaching. In M. C. Wittrock (Ed.), Handbook of research on teaching (3rd ed.) (pp. 3-36). New York: MacMillan. Sills-Briegel, T., & Fisk, C ., & Dunlop, V. (1996/1997). Graduation by exhibition. Educational Leadership, 54 (4), 66-71. Simon, B. S. (2001). Family involvement in high school: Predictors and effects. NASSP Bulletin, 85 (627), 8-19. Sizer, N. F. (2002). Crossing the stage: Re designing senior year . Portsmouth, NH: Heinemann. Sizer, N. F. (2003). Reclaiming senior year. Independent School, 62 (3), 24-31. Sizer, T. R. (1984). Horaceâ€™s compromise: The dilemma of the American high school . Boston: Houghton Mifflin. Sizer, T. R. (1992). Horaceâ€™s school: Redesignin g the American high school . Boston: Houghton Mifflin. Skinner, E. A. (1995). Perceived control, motivation, and coping . Newbury Park, CA: Sage. Skinner, E. A., & Belmont, M. J. (1993). Mo tivation in the classroom : Reciprocal effects of teacher behavior and student e ngagement across the school year. Journal of Educational Psychology, 85 (4), 571-581.
275 Skinner, E. A., Wellborn, J. G., & Connell, J. P. (1990). What it takes to do well in school and whether Iâ€™ve got it: A pr ocess model of perceived control and childrenâ€™s engagement and achievement in school. Journal of Educational Psychology, 82 (1), 22-32. Skinner, E. A., Zimmer-Gembeck, M. J., & C onnell, J. P. (1998). Individual differences and the development of perceived control. Monographs of the Society for Research in Child Development, 63 (2-3). Small, R. V., & Arnone, M. P. (2000). Turning kids on to research: The power of motivation . Englewood, CO: Teacher Ideas Press. Smith, D., & Langfield-Smith, K. (2004). St ructural equation modeling in management accounting research: Critical an alysis and opportunities. Journal of Accounting Literature, 23 , 49-86. Snee, R. D. (1977). Validation of regr ession models: Methods and examples. Technometrics, 19 (4), 415-428. Spector, P. E. (1987). Method va riance as an artifact in self -report affect and perceptions at work: Myth or significant problem? Journal of Applied Psychology, 72 , 438443. Spooner, M. (2002). Creative teenage students: What are they telling us about their experiences in (and around) our high schools. The Alberta Journal of Educational Research, 48 (4), 314-326. Steiger, J. H. (1990). Structural model evaluation and modification: An interval estimation approach. Multivariate Behavioral Research, 25 , 173-180. Steiger, J. H. (1999). EzPATH: Causal modeling . Evanston, IL: SYSTAT, Inc. Steinberg, L., Brown, B. B., & Dornbusch, S. M. (1996). Beyond the classroom: Why school reform has failed and what parents need to do . New York: Touchstone. Steinberg, L., Dornbusch, S. M., & Brown, B. B. (1992). Ethnic differences in adolescent achievement: An ecological perspective. American Psychologist, 47 (6), 723-729. Steinberg, L., Lamborn, S. D., Dornbusch, S. M., & Darling, N. (1992). Impact of parenting practices on adol escent achievement: Authorit ative parenting, school involvement, and encouragement to succeed. Child Development, 63 , 1266-1281. Stevenson, R. B. (1990). Engagement a nd cognitive challenge in thoughtful social studies classes: A study of student perspectives. Journal of Curriculum Studies, 22 (4), 329-341.
276 Stiggins, R. J. (2001). Student-involved classroom assessment (3rd ed.). Upper Saddle River, NJ: Prentice Hall. Stinchcomb, K. (2005, March 16). Insidiou s â€œvirusâ€ strikes se niors countywide. The Maryland Gazette , p. C2. Stinchcombe, A. (1964). Rebellion in a high school . Chicago: Quadrangle Books. Stipek, D. (2002). Motivation to learn: Integrating theory and practice (4th ed.). Boston: Allyn and Bacon. Streiner, D. L. (2003). Starting at the begi nning: An introduction to coefficient alpha and internal consistency. Journal of Personality Assessment, 80 (1), 99-103. Student Affairs Task Force (1975). What should we do with our se nior year? A report of the 1974-75 Commissionerâ€™s Student Advisory Committee . Albany, NY: New York State Education Department. (ERI C Document Reproduc tion Service No. ED119367) Summers, J. (1989). The senior pr oject: A walkabout to excellence. English Journal, 78 (4), 62-64. Tabachnik, B. G., & Fidell, L. S. (2001). Using multivariate statistics (4th ed.). Boston: Allyn and Bacon. Tay, L. P. (2005, November). Does the survey response scale format matter? Paper presented at the meeting of the Intern ational Military Te sting Association, Singapore. Retrieved June 8, 2006, from http://www.internationalmta .org/Documents/2005/2005133P.pdf . Tepper, B. J., & Tepper, K. (1993). The e ffects of method varian ce within measures. The Journal of Psychology, 127 (3), 293-302. The Secretaryâ€™s Commission on Achi eving Necessary Skills. (1991). What work requires of schools: A SCANS report for America 2000 . Washington, DC: U.S. Department of Labor. Retrieved July 27, 2005, from http://wdr.doleta. gov/SCANS/whatwork/ . Thomas, J. W., Bol, L., Warkentin, R. W., Wilson, M., Strage, A., & Rohwer, W. D. (1993). Interrelationships among, studen tsâ€™ study activities, self-concept of academic ability, and achievement as a f unction of characteristics of high school biology courses. Applied Cognitive Psychology, 7, 499-532. Thompson, M. (1996). Fenced in by delu sions: Parents and the college admissions process. Independent School, 56 (1), 1-12. Thompson, S. (1998). Kafkaâ€™s Metamor phosis: Transforming the pre-graduation malaise. Gifted Child Today, 21 (1), 36-38, 48.
277 Toch, T. (2003, December 3). Small schools, big ideas. Education Week, 23 (14), 32, 44. Tomsho, R. (2005, February 8). Senior bl ues: When high schools try getting tough, parents fight back. The Wall Street Journal , p. A1. Trusty, J. (1996). Relationship of parental involvement in teensâ€™ career development to teensâ€™ attitudes, perceptions, and behavior. Journal of Research and Development in Education, 30 (1), 63-69. Trusty, J., & Dooley-Dickey, K. (1993). Alie nation from school: An exploratory analysis of elementary and middle school studentsâ€™ perceptions. Journal of Research and Development in Education, 26 (4), 232-242. Trusty, J., Watts, R. E., & Erdman, P. (1997). Predictors of parentsâ€™ involvement in their teensâ€™ career development. Journal of Career Development, 23 (3), 189-201. Tsuzuki, M. (1995). Seni or projects: Flexible opportunitie s for integration. In W. N. Grubb (Ed.), Education through occupations in American high schools (Vol. 1): Approaches to integrating ac ademic and vocational education (pp. 134-147). New York: Teachers College Press. Tucker, C. M., Zayco, R. A., Herman, K. C ., Reinke, W. M., Trujillo, M., Carraway, K., et al. (2002). Teacher and child variables as predictors of academic engagement among low-income African American children. Psychology in the Schools, 39 (4), 477-488. Tuckman, B. W. & Sexton, T. L. (1991). Th e effect of teacher encouragement on student self-efficacy and motivation for self-regulated performance. Journal of Social Behavior and Personality, 6 , 137-146. Tunstall, P., & Gipps, C. (1996). Teacher feedback to young children in formative assessment: A typology. British Educational Research Journal, 22 (4), 389-404. Turner, S., & Lapan, R. T. (2002). Career self-efficacy and perceptions of parent support in adolescent career development. The Career Development Quarterly, 51 , 44-55. U.S. Department of Labor (2005). Work activity of high school students: Data from the National Longitudinal Survey of Youth 1997 (USDL 05-732). Retrieved June 1, 2006, from http://www.bls.gov/nls/nlsy97r6.pdf . Vallerand, R. J., Fortier, M. S., & Guay, F. (1997). Self-determination and persistence in a real-life setting: Toward a motivational model of high school dropout. Journal of Personality and Social Psychology, 72 (5), 1161-1176. Vallerand, R. J., & Reid, G. (1984). On th e causal effects of perceived competence on intrinsic motivation: A test of cognitive evaluation theory. Journal of Sport Psychology, 6 , 94-102.
278 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. Vaznis, J. (2005, June 26). High schools tackle â€œsenioritis.â€ The Boston Globe , p. B7. Viadero, D. (2001, April 11). Ge tting serious about high school. Education Week, 20 (30), 1, 18, 20, 22. Voelkl, K. E. (1995). School warmth, st udent participation, and achievement. Journal of Experimental Education, 63 (2), 127-138. Wade, T. (1999). Fighting high school senior sl ump: The spread of an alternative senior program. Phi Delta Kappan, 80 (10), 763-765. Walberg, H. J. (1984). Families as partners in educational productivity. Phi Delta Kappan, 65 (6), 397-400. Walker, J. M. T., & Hoover-Dempsey, K. V. (2001, April). Age-related patterns in student invitations to parental involvement in homework . Paper presented at the symposium, Parental involvement in ho mework: What do we know and how do we really know it? at the meeting of the Amer ican Educational Research Association, Seattle, WA. Warr, M. (1993). Parents, peers, and delinquency. Social Forces, 72 , 247-264. Watkins, D., & Hattie, J. (1990). Individual a nd contextual differences in the approaches to learning of Australian s econdary school students. Educational Psychology, 10 (4), 333-341. Wedman, J., & Diggs, L. (2001). Identifying barriers to technologyenhanced learning environments in teacher education. Computers in Human Behavior, 17 , 421-430. Wehlage, G. G., Rutter, R. A., Smith, G. A ., Lesko, N., & Fernandez, R. R. (1989). Reducing the risk: Schools as communities of support . New York: Falmer Press. Weiner, B. (1974). Achievement motivation and attribution theory . Morristown, NJ: General Learning Press. Weldy, G. R. (1984). Coping with the twelft h-grade letdown: A solution to senioritis. NASSP Bulletin, 68 (470), 89-93. Wentzel, K. R. (1994). Relations of social goal pursuit to social acceptance, classroom behavior, and perceive d social support. Journal of Educational Psychology, 86 , 173-182. Wentzel, K. R. (1997). Student motivation in middle school: The role of perceived pedagogical caring. Journal of Educational Psychology, 89 , 411-419.
279 West, S. G., Finch, J. F., & Curran, P. J. (1995). Structural equation models with nonnormal variables: Problems and reme dies. In R. H. Hoyle (Ed.), Structural equation modeling: Concep ts, issues, and applications (pp. 56-75). Thousand Oaks, CA: Sage. West Virginia Department of Education (2002, October 4). State conference to evaluate reinvention of senior year for We st Virginia high school students . Retrieved April 21, 2006, from http://wvde .state.wv.us/news/509/. Wettersten, K. B., Guilmino, A., Herrick, C. G ., Hunter, P. J., Kim, G. Y., Jagow, D., et al. (2005). Predicting educational and vo cational attitudes among rural high school students. Journal of Counseling Psychology, 52 (4), 658-663. What Kids Can Do (2002). Learning outside the lines: Si x innovative programs that reach youth. (ERIC Document Reproduction Service No. ED481279) Whitacre, G. H. (1984). A new twelfth-gr ade curriculum: Planning for a transitional year. NASSP Bulletin, 68 (475), 114-116. Whitehead, A. N. (1929). The aims of education . New York: The Free Press. Whittaker, T. A. (2003). The performance of cross-validation indices used to select among competing covariance structure models (Doctoral dissertation, University of Texas-Austin, 2003). Dissertation Abstracts International, 64 , 4362. Wigfield, A. (1994). The role of childrenâ€™s achievement values in the self-regulation of their learning outcomes. In D. H. Schunk & B. J. Zimmerman (Eds.), Selfregulation of learning and performan ce: Issues and educational applications (pp. 101-124). Mahwah, NJ: Lawrence Erlbaum. Wigfield, A., & Eccles, J. S. (1989, March). Relations of expectanc ies and values to studentsâ€™ math grades and intentions . Paper presented at the meeting of the American Educational Research Association, San Francisco. Wigfield, A., & Eccles, J. S. (2000). E xpectancy-value theo ry of achievement motivation. Contemporary Educational Psychology, 25, 68-81. Wigfield, A., & Tonks, S. (2002). Adol escentsâ€™ expectancies for success and achievement task values during the middle and high school years. In F. Pajares & T. Urdan (Eds.), Academic motivation of adolescents (pp. 53-82). Greenwich, CT: Information Age Publishing. Wiggins, G. (1989). Teaching to the (authentic) test. Educational Leadership, 46 (7), 141-147. Williams, G. C., & Deci, E. L. (1996). In ternalization of biop sychosocial values by medical students: A test of self-determination theory. Journal of Personality and Social Psychology, 70 (4), 767-779.
280 Williams, S. B. (1987). A comparative st udy of black dropouts and black high school graduates in an urban public school system. Education and Urban Society, 19 (3), 311-318. Wills, R. (2005, April 14). A glimpse of reality. Pittsburgh Tribune-Review , p. D1. Wimberly, G. L., & Bernstein, D. J. (2005). Senior exit survey: Class of 2004 (A report from the Department of Shared Ac countability, Montgomery County Public Schools). Retrieved March 15, 2006, from http://www.mcps.k12.md.us/departments/ sharedaccountability/reports/2005/senior %20exit%20survey%20report%20FINAL.pdf . Winters, M. A. (2000). Senior proj ect: A paradox in critical pedagogy. Dissertation Abstracts International, 61 (07), 2652. (Universit y Microforms No. 9980087) Wise, S. L., Cameron, L., Yang, S., & Davis, S. (2005). Information literacy test: Test development and administration manual. Harrisonburg, VA: Center for Assessment and Research Studies, James Madison University. Wolters, C. A. (2003). Understanding procra stination from a self -regulated learning perspective. Journal of Educational Psychology, 95 (1), 179-187. Wolters, C. A., & Pintrich, P. R. (1998). C ontextual differences in student motivation and self-regulated learning in mathematics, English, and social studies classrooms. Instructional Science, 26 , 27-47. Yair, G. (2000). Educational battlefields in America: The tug-of-war over studentsâ€™ engagement with instruction. Sociology of Education, 73 , 247-269. York-Barr, J., & Paulsen, T. (1997). Stude nt perspectives on hi gh school experiences and desired life outcomes. High School Journal, 80 (2), 81-94. Zimmerman, B. J. (2000). Self-efficacy: An essential motive to learn. Contemporary Educational Psychology, 25 , 82-91. Zimmerman, B. J., Bandura, A., & Martinez -Pons, M. (1992). Self-motivation for academic attainment: The role of self-effi cacy beliefs and personal goal setting. American Educational Research Journal, 29 , 663-676. Zimmerman, B. J., & Martinez-Pons, M. (1990) . Student differences in self-regulated learning: Relating grade, sex, and gifte dness to self-efficacy and strategy use. Journal of Educational Psychology, 82 (1), 51-59. Zuckerman, M., Porac, J. F., Lathin, D., Smith, R., & Deci, E. L. (1978). On the importance of self-determination for intrinsically motivated behavior. Personality and Social Psychology Bulletin, 4 , 443-446.
281 Zuker, R. F. (1997). Stress points in the coll ege transition: What to expect/how to help students cope. The College Board Review, 182 , 14-18.
282 BIOGRAPHICAL SKETCH Bryan Duff graduated from Princeton Univer sity in 1996 with a bachelorâ€™s degree in psychology and a certificate of mastery in linguistics. After college Bryan taught mathematics and psychology at independent hi gh schools in Florida and New Hampshire. He earned his M.Ed. from the University of New Hampshire in 2004. His culminating masterâ€™s project was a mixed-methods evalua tion of the Senior Project program at his school and a proposal for a new model. Main taining this focus as a doctoral student, Bryan served a two-year assistantship as a Senior Project coordinator at P.K. Yonge Developmental Research School. Bryan is married to Kerin Duff, a speech-l anguage pathologist. Their son, Liam Patrick, was born in August 2006. Bryan curr ently is on faculty at Berkshire School in Sheffield, Massachusetts, where he coor dinates the independent study program and teaches mathematics.