o J u ,n no. to -. . I ‘0‘! I. .II “.IL. lb“. .' \.‘ roubntfil '0 “ID\JO- IO: 0.: .0 .. .3. I; . . P. i. OIL-III... '.. N ‘ _ Jul. .3 -jnc.a td“ ll 0.. “infiiflo II. 9. . O. I u 'I ”"1 rctwllv ‘ ". at 0". .N n‘ IIL " vii. I. l. 0"” Ch. .00 Cult |I - t I. n o. ”.1 “4 $53.... i..m...du.,. duh-‘u-h vl “IIL 0- O.‘ I I . ‘l“n.l t ' llo. “| .. II h ‘ne g'v's I O “.01!“ llllf'i- In la 0 -‘u '\ -II‘ (vii... 00f . ‘ hum. .‘xol‘l‘ 0......013!‘ . Iflll . I. ‘I ' ‘ I I . l! I. I... L I h .li‘fi‘ ‘ 0‘ .C‘ II «.nfll. l: ..u ..‘ I- Illl‘.t...i 0 .II i.::' I I . l l... 0 O ‘1.“ 11;, ‘II'|‘ I. "O ‘aflf’ $ ‘%r .I' lfl’lII - :I‘“ .Id‘wlr Jill." t i. o p Q o 1. o 15'. - . "O 0 . - l‘b ii'I Iv .fufixtfl.r|vr§ II... . 3 ... I; .i‘Ll. :. ’1. .. I I ll . .II. 5 O f'rllo‘ I a flail" Ti .' c‘ll‘llll d D l OOIQIIJufiohi:|-i“;c" ll . IAOIII. . ‘ IIIII In... 1.08%.. flu. ‘I..' '- flan: 4 ‘ o ‘0 i.‘ ‘é’.’ it . ‘b‘doofi ““11 .II. 3“: -0... l 1. ‘u I .04 \ ‘3‘ V O ‘. .. ‘I... 0..“ I ‘01... 1...“- I‘lCUI -r ‘ y a . D 1‘ O.| ’4‘... I.“ at‘t‘llr’ Ill. flit: . ! ‘01:...1‘J‘1J19f.‘ . 3! ‘o ' nlo'_."r.{\ 1 II-‘ 1 \ . \ PM. I‘ll lllllllllllllllllllllllllllllll“Hillllllllllalllll mass 3 1293 10492 L! is it. A at L University This is to certify that the dissertation entitled A STUDY OF INSTITUTIONAL ATTRITION AMONG FIRST-TIME COLLEGE FRESHMEN: FALL 1981 presented by William James Latta ' has been accepted towards fulfillment of the requirements for . . r Ph. D. degree in Administration and Curriculum mine I Ma ajor professor Date May 1983 MSU is an Affirmative Action/Equal Opportunity Institution 0- 12771 [M .’ ————- ._~. -—.—.—— MSU RETURNING MATERIALS: Place in book drop to ngaAmgs remove this checkout from .—:—. your record. FINES will be charged if book is returned after the date stamped below. affiguiflfifiiflfifirrfibgF' l;( 200 A302 2:: 7 ’ JU6J6117929127 75mm we; gill 205100 (”W99 Jumzm A STUDY OF INSTITUTIONAL ATTRITION AMONG FIRST-TIME COLLEGE FRESHMEN: FALL 198l By william James Latta A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Administration and Curriculum 1983 ma — 3787 ABSTRACT A STUDY OF INSTITUTIONAL ATTRITION AMONG FIRST-TIME COLLEGE FRESHMEN: FALL 1981 By William James Latta The purpose of this research was to conduct an institutional study of why freshmen who entered college during Fall Term l98l did not return to resume their degree work in Fall l982. The investiga- tion was conducted within the framework of Tinto's (l975) conceptual model and focused on the problem of the voluntary dropout. Students chosen for this study were college freshmen who enrolled in a four- year program during Fall Term l98l, were enrolled during Winter Term l982, and lived in university residence halls. The general design of this research was longitudinal, with data collections occurring before college enrollment and at the mid- point of the first academic year. Selected items from the Student Profile Section of the American College Test provided the pre-enrollment information. A questionnaire developed from two previously researched instruments was used to obtain data about the students' college experi- ences. Both data collections were held until Fall Term 1982, or the beginning of the second year. At that time, the sample was divided into two groups, returning and nonreturning, with academically dis- missed students deleted. William James Latta The procedures of discriminant analysis and cross-tabulation with chi-square were used to test the eight hypotheses of this investi- gation. The data analysis revealed that substantive discriminant functions could not be found distinguishing returning from nonreturn- ing freshmen. However, the chi-square tests did reveal moderate differences. In considering the specific variables that affected student attrition in this study, high school GPA, class rank, and the degree of certainty that one has toward his/her college major were found among the precollege characteristics. The number of collegiate extra- curricular activities, family interactions, perceived intellectual development, amount of perceived faculty concern for student develop- ment and teaching, and college GPA were significant among the college- experience variables. Additional differences predicted between male and female students were found. This study also demonstrated the usefulness of Tinto's model in explaining the dropout process for nonreturning as well as retention of returning students. However, it became clear that the model is far from complete as there are some obvious shortcomings in its ability and need to account for attrition influences that exist outside the insti- tution. The results of this study indicated a need to further research and understand the specific nature of college experiences that contrib- ute to attrition. Additionally, influences external to the institution warrant investigation to gain a more complete understanding of the problem. ACKNOWLEDGMENTS AND DEDICATION It has often been said that thesis research is a solitary project. Images of a dimly lit room, pieces of crumpled-up paper scattered about, and a lonely figure poring over numerous notes and computer printouts appear to fit this notion. However, my purpose here is to demonstrate that the exact opposite was my experience. In fact, there were so many who gave their active support and ongoing encouragement that there was never a question of "If I finish," but "When I finish"! I would like to begin by thanking my doctoral committee, who gave me ongoing and sound guidance throughout the years of my program. Before his departure for Illinois, Dr. Gary North provided me with solid direction and leadership as my Chairman; he will always be an important model in my career. Dr. Richard Featherstone, my current Chairman, was always glad.to see me and could always be counted on for his warmth and reassuring smile. I have been fortunate to have learned some of his insights and to have grown from his knowledge shaped through wisdom and experience. Dr. James Rainey was instru- mental in the formation of my cognate and never hesitated to give me careful and candid advice; I quickly learned to implicitly trust his judgment. Dr. Eldon Nonnamaker willingly joined my committee at Dr. North's departure; his assistance and years of administrative ii experience were always gratefully respected. Dr. Fred Ignatovich has had a greater influence on me than he will know or than can be measured. He has always represented a strength of competence and rigor that is the most necessary element of any doctoral program. I will always value and continue to integrate the knowledge from his courses as well as carry a deep admiration for his critical observa- tions. During the course of my research, three pe0ple were key ele- ments in smoothing out the rough edges and helped to keep me focused on progress. Dr. Jim Studer quickly lent his support and enabled me to cut through the initial red tape of collecting a sample. I always felt his enthusiasm and positive support. Jim (soon to be Dr.) Wall has provided me with more tangible support and encouragement than could be imagined. Jim always had an understanding ear, a foot squarely planted in my back end, and arms strong enough to give me the room to finish this task; when the going got tough, Jim kept me going. Finally, there is one man who deserves more credit than any computer program could produce. Dr. Bill Rosenthal was nothing short of a miraculous discovery. There is no question that Bill's help and patience made this project possible. His computer expertise and technical knowledge were invaluable as tapes were merged, samples were coded, variables were recorded, tables were produced, and statistical tests were performed. 1 only hope to be able to help some other doc- toral candidate as Bill has helped me. He is truly representative of my experience at Michigan State. The completion of this dissertation and doctoral program rep— resents a dream fulfilled. Four people intimately shared in this dream, and it is to them that my degree is dedicated. My wife has given more than could be measured by any yardstick known to history. She gave me more time to study than could be asked, listened to more term paper rewrites than could be expected of an editor, allowed me more time in the library than would be healthy for a bibliographer, and tolerated my outbursts of frustration beyond the point justifiable in a marriage. She has been my most important continuous and unconditional source of strength and reassurance. My parents, though they may not realize it, gave me the personal foundation to reach this moment. Mom has given me a sense of optimism and hope throughout my life that has equipped me with the vision to see possibilities, not barriers. I have also been fortunate to have gained her sense of humor. This gift has always helped me thoroughly to enjoy a humorous story and to keep the details of life in perspective. Dad has taught me the importance of setting high standards without ever compromising my personal integrity. From his guidance I've come to know thoughtfulness, understanding and patience, critical analysis, and the respected gift of persistence. Dad has not always received the credit he deserves in my growth, but he should know that the positive influence of his life on mine has been instrumental not only in reaching this goal but many more to come. My son, who is too young to know what all of this means, deserves the final recognition. It is hoped that this dissertation will be a sym- bol of the importance of dreaming dreams and then setting out to iv accomplish them. It now becomes the task of his father to expose him to the fullness of life's possibilities and to instill in him a sense of purpose that he feel compelled to follow no one's footsteps but his own. TABLE OF CONTENTS Page LIST OF TABLES ........................ viii LIST OF FIGURES ........................ x Chapter I. INTRODUCTION ..................... Purpose of the Study ................ 4 Need for and Importance of This Study ........ 5 Definition of Terms ................. 6 Background of the Conceptual Model ......... 7 Delimitations of the Study ............. l3 Hypotheses Tested .................. l4 Design of the Study ................. l5 Limitations of the Study .............. l8 Overview of Subsequent Chapters ........... 19 11. REVIEW OF THE LITERATURE ............... 21 Introduction .................... Zl The Rate of Attrition ................ 23 Precollege Characteristics and Their Importance for the Dropout .................. 29 Attrition: Toward a Formulation of Theory ...... 44 Summary of Chapter II ................ 66 III. DESIGN AND METHODOLOGY ................ 7O Hypotheses ..................... 7l Population and Sample ................ 72 Instrumentation ................... 75 Data Collection and Statistical Analysis ...... 8] Summary ....................... 84 IV. PRESENTATION AND ANALYSIS OF THE DATA ......... 86 Introduction .................... 86 Analysis of the Hypotheses ............. 89 Summary ....................... l33 vi Page V. SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS ........ l37 Summary: Development of the Study ........... T38 Findings and Interpretation Within the Framework of Tinto's Model .................. l4l Verification of Data Coding and Analysis ....... 155 Conclusions ...................... l58 Recommendations for Further Research ......... l6O Reflections ...................... T62 APPENDICES ........................... 164 A. ITEMS SELECTED FROM THE AMERICAN COLLEGE TEST: STUDENT PROFILE SECTION ................ T65 B. QUESTIONNAIRE COVER LETTER AND COLLEGE EXPERIENCES SCALE II ....................... l7O BIBLIOGRAPHY .......................... T75 vii .h bhhbhw .10 .11 .12 .13 LIST OF TABLES MSU Student-Enrollment Projections, l980 to 1985 ..... National Rates of Attrition by Research Source ...... Persistence/Attrition of Domestic MSU Students (January 15, 1982) ................... Comparison of Subscale Reliability Coefficients--CES II. . Sample Data ....................... Chi-Square of Significant SPS Variables: Total Group . . . Chi-Square of Significant SPS Variables: Females ..... Chi-Square of Significant SPS Variables: Males ...... Stepwise Selection Summary of SPS Variables for the Total Group ...................... Stepwise Selection Summary of SPS Variables for Females Stepwise Selection Summary of SPS Variables for Males Chi-Square Results on the Variable, Participation in School Activities, for the Total Group ......... Chi-Square Results on the Variable, Number of Family Interactions, for the Total Group ........... Chi-Square Results on the Variable, Intellectual DevelOpment, for the Total Group ............ Chi-Square Results on the Variable, Current Grade Point Average, for the Total Group .............. Chi-Square of Significant CES II Variables for Female Students ........................ Chi-Square of Significant CES II Variables for Male Students ........................ Page 26 27 80 88 92 93 95 97 97 lOZ 105 106 Table Page 4.14 Stepwise Selection Summary of CES II Variables for the Total Group ...................... 115 4.15 Discriminant Analysis Classification Summary: CES II Variables: Total Group ................. 117 4.16 Stepwise Discriminant Summary of CES II Variables for Females ...................... 118 4.17 Stepwise Discriminant Summary of CES II Variables for Males ....................... 118 4.18 Comparison of Discriminant Functions of SPS and CES II Variables for the Total Group ............. 121 4.19 Comparisons of Discriminant Functions of SPS and CES II Variables for Females and Males ............ 122 4.20 Stepwise Selection Summary of CES II and SPS Variables for the Total Group .................. 127 4.21 Stepwise Selection Summary of CES II and SPS Variables for Female Students .................. 128 4.22 Stepwise Selection Summary of CES II and SPS Variables for Male Students ................... 130 ix LIST OF FIGURES Figure Page 2.1 A Conceptual Model for Dropout From College ....... 59 I CHAPTER I INTRODUCTION During the late 19505 and into the 1960s, higher education was one of the most rapidly expanding institutions in the United States. Under the presidential administrations of John Kennedy and Lyndon Johnson, higher education enjoyed a wealth of public and pri- vate support. At that time, a favorable growth climate brought increasing enrollments, many requests for research and public service, and expansion of facilities and faculties (Garbarino & Aussieker, 1975). However, as higher education encounters the demands of a new decade, the 19605 can be viewed as the "golden years" and a period that brings only pleasant memories in the wake of uncomforting com- plexities in the 19805. Many issues are confronting institutions of higher education during the current period. Some of the more pertinent challenges requiring improved problem-solving abilities are declining resources, increased planning with prioritization of programs, use of management systems, and decreased real financial support (Gardner, 1977; Pifer, 1976). Not the least problem, and perhaps one of the most important ones for institutions in the 19805, is the maintenance of enrollment levels. A recent report from the Carnegie Commission on Higher Edu- cation indicated that enrollment declines during the 19805 will present a problem with such severity that it will create a "new revolution" in academia. Although small liberal-arts colleges appear to be most vulnerable, large public universities, particularly one that was the focus of this study, will also be affected (Scully, 1980). Enrollment levels are influenced by new enrollees each year (entering freshman classes), students who return or transfer into the school from a previous enrollment, and the rate at which students drop out before completing their degree. When one of the first two groups of students declines, institutional officials are hopeful that the other groups will offset any noticeable differences. How- ever, when new-student enrollment is down along with that of return- ing and transfer students and the rate of attrition is at an unaccept- able level, the educational quality of the institution, its financial future, and even its survival can be affected (Scully, 1980). To illustrate the seriousness of enrollment declines at one institution, Michigan State University, a large midwestern land-grant school, projects a 2 % decrease in the undergraduate population by 1985 (Table 1.1). In Fall 1980, the institution enrolled 36,620 undergraduate students; however, projected enrollment in 1985 is expected to be 29,400 undergraduates. Although somewhat compensated by readmits, transfers, and returning students, the overall enrollment decline is only marginally offset.1 Given this rate of decline, the 1Enrollment figures were obtained from a 10-13-81 memorandum distributed by the MSU Office of Planning and Budget, Michigan State University. Figures were compiled by Dr. William Sperber. .mco_pummoea mew memo» omen» gee mmcam_dm Aeom-v om~.e Aem-v oo¢.mm AN¢-V ooe.om Axe-v oce._m “Ne-v oom.mm A5 -V Nmm.¢m omo.om _muoe Aeom-v _mo.m Rem-v mev.om A5Q-V me_._m flee-v meo._m “an-“ omm.mm A55 -V mmm.em oom.¢~ o=_=esuaa AN__-V _m_ Ase oo_.e Aew-v oo_._ Aav oom._ Aem+v ooN._ A55 -V mo_.p _MN._ cageeseaax Aem_-v «me Hem-v ooa._ Aem-v ooo.~ Aav oo_.N Aem+v oo_.m Aeme-v mec.m emm.m ememcoce A5N~-v 5N8._ em-v mmm.m A5N-V mm_.o Aym-v mmm.o A5e+v ome.o Aee_-v ww~.o mcm.e ma_5 Sweee :52 0mm. sect coma. memo, cmmmp mmwo_ _wo_ owm_ «Smacaemeaucs «mango m>_um_:E:u .mmm_ OB omm_ .m=o_uumwccq Seas__occm-ucmuapm =ms--._._ 5.55e attrition rate of undergraduates and their reasons for leaving the institution become even more critical and alarming concerns. Students who do not return to an institution tend to do so for one of three reasons. Either the student has graduated, was academically recessed, or voluntarily elects not to return. As an example, it was found that approximately 17 to 20% of new freshman enrollees at MSU are not expected to return for their 50phomore year. Further, a majority of these students make a voluntary decision not to return, as Opposed to being academically recessed, dismissed, or suspended for disciplinary reasons (Tinto, 1975). During a time of increasing enrollments, this rate of drop out from the freshman class may not be of major concern. However, given the current circum- stances compounded by other problems mentioned at the beginning of this chapter, attrition of almost one—fifth of each new entering class becomes a major concern. Purpose of the Study The purpose of this study was to examine the reasons why freshmen who entered Michigan State University during Fall Term 1981 did not return to resume their degree work in the Fall of 1982. This investigation was conducted within the framework of Tinto's (1975) model of why students drop out. The model was particularly approp- riate in that although it has implications for students who drop out from higher education in general, it also focuses on institutional attrition. Further, initial research (Baumgart & Johnson, 1977; Terenzini & Pascarella, 1980; Munro, 1980) has begun to validate various components of the model, supporting its usefulness in explain- ing rather than simply describing students who do not persist. There- fore, following from this framework, variables were examined that discriminate between students who leave MSU and those who return for the following year. Precollege characteristics and experiences on the campus were the categories under which more specific variables were studied. Need for and Importance of This Study The intention of this research was to investigate student attrition at one midwestern institution, Michigan State University. There has not been a study on this subject in the past ten years at MSU, and studies before that time were limited because they sought only to describe characteristics associated with dropout behavior and did not try to explain the findings within the framework of a theo- retical model. At a time when enrollments are declining, it becomes even more necessary to understand why students who initially enroll do not return. The results obtained from an individual-institution study may have many uses. Initially, the results can be compared with national findings to determine the degree of similarity between the two settings. Also, institutional programs, services, and college personnel can benefit from the identification of those aspects of campus life most closely related to attrition. Further, the findings may illuminate those components of the institution that are positively related to persistence and eventual success of the student. In his recent book, Four Critical Years, Astin (1978) wrote that unfortunately, more decisions are made in higher education on the basis of financial considerations than on sound educational research. He went on to say that it is therefore incumbent on researchers to demonstrate that studies also have a cost/savings- benefit capability. Toward this end, it is hoped that the findings of this study will help provide additional information for improving an individual student's success. Astin (1975) wrote that it may, in fact, be more cost effective to take additional steps to retain those enrolled than to incur costs of recruiting and selecting students to replace those who have not returned. Definition of Terms First—time college freshman: An MSU student who entered college for the first time at MSU, as indicated by the appropriate code on the Student Master Data Tape (SMDTL.duringFa11 Term 1981. Institutional attrition: The act of leaving Michigan State University before completing the degree requirements for a bachelor's degree. Nonreturning student: A first-time college freshman who did not return to Michigan State University for his/her second year after three consecutive terms of enrollment (Fall 1981, Winter 1982, and Spring 1982). Persister or returning student: A first-time college fresh- man who returned to Michigan State University for his/her second year after three consecutive terms of enrollment or Fall Term 1982. Goal commitment: Refers to the degree of decidedness toward the actions of obtaining a college degree (Tinto, 1975). Institutional commitment: Refers to the degree of decided- ness that a person has toward attending one particular institution over another (Tinto, 1975). Academic integration: The degree to which the student becomes successfully involved in the academic process in the college or university setting. Indications of this type of integration would be grade point average and the individual's amount of involvement in academic-related activities, as well as the student's attitude toward his/her own level of intellectual development (Tinto, 1975). Social integration: The degree to which the student becomes successfully involved in nonclassroom aspects of university life. Indicants of this type of integration would be the amount of involve- ment in peer-group interactions, nonclassroom faculty contacts, extra- curricular activities, work activities, and other interactions on the campus. Background of the Conceptual Model Research related to the college-student dropout has been extensive, varied, somewhat contradictory, and limited in its ability to explain why students leave particular institutions and/or higher education in general. A majority of studies from approximately 1920 until early 1964 have focused on the rate of attrition and precollege characteristics associated with the undergraduate dropout. Not until Spady's study in 1970 was a model or comprehensive theoretical framework proposed that attempted to synthesize the characteristics associated with dropouts and that sought to explain attrition as an interactive process between the individual's attributes and the col- lege environment. Spady's study stimulated three subsequent attempts to develop explanatory models, with the most recent and comprehensive framework attributed to Vincent Tinto in 1975. The first study considered to be a hallmark in attrition research was conducted by Iffert in 1957. Although his study was a follow-up to McNeely's (1938) work, it drew national recognition because of the scope of his work and provided a stimulus that led to many more intensive future efforts. Essentially, Iffert studied characteristics (background, residential, and academic factors) that were thought to discriminate persisters from dropouts. Also inter- ested in the national rate of attrition, he found that 40% of enter— ing students graduate within four years, an additional 20% of all entering students complete a baccalaureate beyond the four-year period, and the remaining 40% never finish their degrees. Many precollege factors associated with persistence and attrition have been documented in the literature. Two of the better- researched characteristics have been the socioeconomic status of the family and the academic preparation and high school success of the student. Astin (1964) and Panos and Astin (1968) found that students who came from families that were well educated, tended to be more affluent, and were academically successful as high school students had higher persistence rates than those students who came from other backgrounds. However, Trent and Medsker (1968) disputed these findings and presented the argument that precollege characteristics associated with family background probably have more to do with the initial choice of going to college as opposed to significantly influ- encing completion. Other studies, despite contradictions in the literature, have continued to assert the importance of precollege characteris— tics. Gurin, Newcomb, and Cope (1968) studied parents' religious affiliation, level of education, rural-urban background, size of high school, high school class rank, and scores from the Scholastic Aptitude Test. Their major finding was that precollege character- istics and background have significant effects on student attitudes and attrition rates. Another variable found to be associated with attrition rates in the category of background characteristics is the student's own evaluation of the high school facilities, curriculum, and instruction. Summerskill (1962) found that students who experi- enced academic difficulty in college and subsequently dropped out placed a majority of the responsibility on the poor quality of their high school. Also, the importance of formulating goals and the expected level of performance have consistently been shown to be precollege discriminators of persistence and attrition. Thistlethwaite (1963), Panos and Astin (1967), Wessel et a1. (1978), and Hutchison and Johnson (1980) all found that students who had chosen a field of study, expected to perform well, had planned to graduate, and/or intended to go to graduate school were more likely to persist in college than were other students who did not form these goals before 10 college enrollment. These few studies, citing the above-mentioned precollege variables, are only a part of the large volume of research conducted in this area. In an attempt to bring coherence to the attrition studies associated with precollege background and experiences, as well as other findings concerning why students do not continue their educa- tion, three major comprehensive literature reviews were conducted by Spady (1970), Cope and Hannah (1975), and Pantages and Creedon (1978). Although a more detailed review of their findings is presented in Chapter II, one of their primary conclusions was that dropping out is the result of a complex process over time. This process includes the student's entering characteristics as well as influences of the col- legiate environment operating in an interactive manner. Viewed in this way, the acts of dropping out or staying in college are the result of a much more complex interaction of variables than originally thought. Building from previous attrition research and Durkheim's study of suicide, Spady (1970) constructed the first process model explaining attrition/persistence behavior. His model theorized that individual student attributes interact with various sources of demand in the college environment, such as academic courses, faculty, administrators, and peers. Sources of demand, primarily from the social and academic systems of the college, present opportunities for successful or unsuccessful integration into the collegiate environ- ment. Spady concluded that the student either finds rewards in these 11 two areas or experiences frustration and lack of integration, result- ing in the eventual dropout behavior. Kamens (1971) and Rootman (1972) also hypothesized that the college environment is a critical determinant in the student's per- sistence toward a college degree. Kamens studied college size and its effect on the "status allocating" ability of the institution through college prestige and quality. Rootman, in line with Spady's work, proposed a "person-fit" model of college-student interaction. Similar to the work of Getzels and Guba (1954, 1963), he sought to explain dropout behavior as the result of dissatisfaction with the college environment. The dissatisfaction is the result of an incon- gruence that exists between individual background characteristics and attributes of the student (Getzels and Guba labeled these "needs disposition) and the role expected of the students by the college (Getzels and Guba labeled this in the same manner--"role expectations"). Essentially, the student discovers that he/she does not fit the envi- ronmental expectations, begins to experience conflict and dissatis- faction, and eventually copes with this experience by dropping out. Finally, Tinto (1975), after a comprehensive study of pre- vious frameworks, developed a "conceptual schema for drop out from college" (p. 95). He proposed that students' precollege character— istics lead to the initial formulation of two types of commitment. One type of commitment is related to the idea of completing a college degree (goal commitment), whereas the other is to the school the student attends (institutional commitment). Once the student begins the degree program, interactions occur in the social and academic 12 subsystems of the college environment. The experiences on the campus then alter initial commitments as the student finds either integration into or isolation from the academic and social arenas. These altera- tions, in turn, cause a new formulation of goal and institutional commitments, leading to eventual completion or drop out from either the specific institution or higher education altogether. Terenzini and Pascarella (1980) conducted a series of studies at Syracuse University that tested Tinto's model. Their research found support for the model as a viable framework within which attrition/persistence can be Explained. They found that stu- dents who persist or withdraw are not "mere reflections" of their precollege characteristics but are subject to the influences they find in the institution they attend, both academically and socially. However, their work, by their own admission, is limited in that the model was tested at only one institution, and the students may not have been representative of students at other schools. Second, as Tinto's model focused primarily on a particular institution as opposed to higher education in general, findings about the environ- ment at one school may not accurately represent the environment at other schools. Third, their studies, although using a variety of precollege variables and particular methodology to measure social and academic integration, "probably only begin to reflect the complexity of the model's major constructs" (p. 349), pointing to the need for additional research. Finally, their research did not take into account such variables as precollege commitment toward obtaining a degree, expected level of academic performance and success, degree of 13 decidedness about a particular program of study, or expected level of academic assistance that would be desired for successful comple- tion from the prospective student's point of view. Delimitations of the Study This research study was delimited to one institution within higher education: Michigan State University. Further, the subjects of this thesis were first-time college freshmen who entered during Fall Term 1981, lived in university residence halls, were enrolled during Winter Term 1982, and were not enrolled in any of the two- year degree programs. Previous literature supports the recommendation that insti- tutions must conduct their own research to identify accurately the reasons why students leave school. ITIthlS way, more appropriate corrective action can be taken to reduce the rate of institutional attrition (Hutchinson 8 Johnson, 1980). As Terenzini (1980) wrote, "Institutional contributions to attrition can be more clearly deline- ated and assessed. . . . Results will have greater potential for valid interpretation and wider administrative planning utility." The entering-freshman class was chosen for this study becuase of its projected high rate of attrition, approximately 17 to 20% between the freshman and sophomore years. A study conducted by the institution's Office of Planning and Budget and records from the Office of the Registrar showed that the freshman year has the highest 14 dropout rate of the four classes.1 On a national level, high rates of attrition were also reported by Cape and Hannah (1975) at the end of the first year. Their research cited findings from McNeely (1938), Iffert (1957), Summerskill (1962), Skaling (1971), Panos and Astin (1968), and Astin (1972). These investigations showed consistently across all the studies except McNeely's, which reported 60%, that approximately 40% of the entering freshmen do not complete the bac- calaureate degree. Eventual attrition of MSU students has reached this level (45% from the 1975 class), but because of the period under study in this investigation, it was expected that the rate would be in the range of 17 to 20%. Hypotheses Tested In view of the purpose of this study, the abbreviated back- ground literature that has been presented, and the review of litera- ture presented in the second chapter, the following hypotheses will be tested: 1. The Student Profile Section of the American College Test, which will be used to measure entering-freshman charac- teristics, will successfully discriminate between freshman- year persisters and nonreturning students. 2. The precollege characteristics of family income, high school academic performance, and educational aspirations (measured by decidedness of college major, expected level of college performance, and college degree) will be the best discriminating variables between freshman persisters and nonreturning students among all measured prematricu- 1ation variables. 1Summary statement from interviews held with Dr. William Rosenthal, MSU Office of Budget and Planning, and Phyllis Wilke, Assistant Registrar, November 4, 1981. 15 3. The College Experiences Scale II, which will be used to measure social and academic integration, will success- fully discriminate between freshman-year persisters and nonreturning students. 4. Variables related to social and academic integration will be better discriminators between returning and nonreturn- ing students than precollege characteristics. 5. The School Activities, Peer, and Faculty Interactions Scales measured by the CES II will be the most powerful discriminating scales between returning and nonreturning freshmen of any variables found to discriminate between the two groups along the social-integration dimension. 6. Factors related to academic integration will be better discriminators between returning and nonreturning stu— dents than those factors related to social integration. 7. The CES II scales that measure the amount and quality of faculty contact will be the single most powerful discriminating variable between freshman-year persisters and nonreturning students of any factor measured in the study. 8. Those variables that successfully discriminate between returning and nonreturning students will be different when controlled for sex. Design of the Study The design of this study was longitudinal rather than post- hoc or cross-sectional. Specific to this type of research, the lon— gitudinal design has been found to be more advantageous and superior to the other two approaches. The post-hoc or autopsy design calls for the collection of information from students who have already dropped out. Typically, this method encounters low response rates, does not have the benefit of comparison groups, and the data only provide descriptions of stu- dents who left. Also, it has been found that students who are l6 surveyed after they drop out may tend to rationalize their responses, as opposed to giving the "true reasons." The cross-sectional design makes some improvements on the above approach in that it provides for a comparison group. Using this method, data are collected from students who are currently enrolled (usually late into their first year) and held until the beginning of the next academic year. At that time, the students who were surveyed are separated into two groups, depending on whether or not they returned to the institution. Data for the two groups are then com- pared. This design allows for a higher response rate but is still lim- ited in that it does not take into account prematriculation data (Terenzini, 1980). The longitudinal design further improves on the cross- sectional approach. Using this method of gathering information while the student is still enrolled, it also requires the researcher to col- lect pre-enrollment data. According to Terenzini (1980), "institutions that participate in national precollege student information programs, such as . . . the American College Testing Program's Assessment pro- gram, have an advantage over institutions that do not" (p. 262). With these two sets of data, multivariate analysis can be performed taking precollege differences into account. This design is not only supported by this more thorough type of analysis, but it also affords a fairly high response rate, a comparison group as mentioned, and specific to this study, provides information that is required for Tinto's (1975) model (Lea, Sedlacek, & Stewart, 1979). 17 To facilitate the use of the longitudinal design for this study, a random sample of first-time college freshmen who entered during Fall Term 1981 and were enrolled Winter Term 1982 was drawn, selecting every "nth" student. After selecting the sample, two sources of information were collected. The American College Test Student Profile Section (SPS) provided precollege characteristics and demographic data on the sampled students, anda questionnaire, the College Experiences Scale II (CES II), was administered the seventh week of Winter Term 1982. Both sources of information were held until after Fall Term registration in September 1982. At that time, data for those students who returned for Fall Term were separated from the data for those students who did not return. With the data separated into two groups, statistical analysis was begun. To test the hypotheses, the subprograms Crosstabs, Dis- criminant, and Stepwise Discriminant from the software package Statistical Analysis Systems were used. The Crosstabs program pro- gram provided a chi-square analysis testing for independence between the nonreturning and returning students on the variables chosen for this study. The results from the chi-square analysis can be suggestive of variables that will discriminate between the groups. However, the procedures Discriminant and Stepwise Discriminant not only tested which variables actually discriminate between the groups but how well they discriminated. '“Hwamathematical objective of discriminant analysis is to weight and linearly combine the discriminating variables in some fashion so that the groups are forced to be as statistically distinct 18 as possible" (Nie et a1., 1975, p. 435). Once discriminant functions have been computed, they can be analyzed by a series of statistical tests to measure the success with which the variables actually discrimi- nate between the groups. Limitations of the Study; Delimitations of this research study—-that is, the focus and scope of the research--have already been examined earlier in this chapter. The purpose of this section is to present the reader with the inherent weaknesses of this thesis that are not within the researcher's ability to control. However, these weaknesses are not so great as to preclude the value of this study, but rather to assist in placing the results and subsequent conclusions in perspective. One limitation is that the researcher must rely on the authen- ticity of the student's answers found within the SPS and CES II. A risk is run in that the responses may be more reflective of subjective feelings about facts as Opposed to the facts themselves. An additional weakness lies in the instruments used to measure the students' precollege characteristics as well as their social and academic integration on the campus. Tinto's model suggests that factors leading to a student's feeling of social and academic integration or isolation are very complex. The researcher has relied on the instruments designed with this model in mind; however, they may not contain a complete representation of the variables required to uncover the complexity of the model. 19 It is also recognized that the results of this study were specific to Michigan State University and a defined population of students within that setting. To the extent that other institutions resemble MSU and its students, the findings are generalizable. Yet, as pointed out earlier, and serving as a part of the rationale for this study, other institutions should develop their own data to assess more accurately the specific circumstances that affect their attrition and persistence rates. Finally, this investigation was intended to provide a back- drop for further study and follow-up on the MSU campus. Restricted by various resource limitations, this research was not able to exhaust all the possible reasons that are associated with returning and non- returning freshmen. Additionally, those variables found to be related to freshman-year attrition may not be representative of other classes of students. Additional research and study is warranted for sophomore-, junior-, and senior—year attrition. Overview of Subsequent Chapters The introduction and background to the research problem was presented in this first chapter. Specifically, the purpose and value of the study, definition of terms, hypotheses, design, delimitations, and limitations of the study were all covered in Chapter 1. Chapter II provides a more detailed literature review per- tinent to this thesis. The review is presented in three parts. Part I reviews studies of national attrition rates and their compari— son to MSU figures. The second part of the chapter presents research 20 information relative to precollege characteristics, while the third section presents a development of studies that have led up to Tinto's model and research that has begun to test it. It will also be found that the last part of the chapter more clearly describes the research previously cited to support the research hypotheses. Chapter III reviews the research design, population sample, instrumentation, and data-analysis techniques. Chapter IV follows from the third chapter by presenting the findings of the data analysis. The fifth and final chapter draws conclusions from the find- ings, poses recommendations for further investigation, and summarizes the entire research project. CHAPTER II REVIEW OF THE LITERATURE Introduction It has already been pointed out that the literature pertain- ing to college attrition is extensive. Studies that date back to 1924 have principally covered individual characteristics associated with dropping out and rates of attrition (Summerskill, 1962). The scope of individual attributes that have been studied includes per- sonality characteristics, socioeconomic status, high school academic achievement, high school extracurricular accomplishments, age, sex, financial assistance, hometown location and size, predicted scholas- tic aptitude, personal motivation, clarity of and commitment to goals, and parental and peer-group influence. More recently, environmental characteristics of the collegiate setting relative to their influence on attrition have been researched (Pantages & Creedon, 1978). Yet, despite the range and breadth of these many studies, The failure of past research to delineate more clearly the multiple characteristics of dropouts can be traced to . . . inadequate attention given to . . . the development of theoretical models that seek to explain, not simply describe, the‘processes that bring individuals to leave institutions of higher education. (Tinto, 1975, p. 1) Recognizing this limitation, research by Pervin and Rubin (1967), Feldman and Newcomb (1969), Nasatir (1969), Spady (1970), Kamens (1971), Rootman (1972), and Tinto (1975) has begun to synthesize 21 22 and integrate previous work into a more meaningful and understandable whole. With respect to studies that have documented attrition rates, although many, they are equally plagued by problems of interpretation. Difficulties are encountered in three areas: definitions of dropout, the time allowed for follow-up, and the "representativeness of samples used" (Kolstad, 1977, p. 4). However, a general review of findings from four-year college studies in comparison with rates at Michigan State University provides a reasonably interesting standard of refer- ence. Such a review is forthcoming in this chapter. Because of the vast amount of literature that exists on the college dropout and the excellent comprehensive reviews of Spady (1970), Tinto (1975), Cope and Hannah (1975), and Pantages and Creedon (1970), an additional review in this thesis would be supererogatory (Baumgart & Johnstone, 1977). However, with the intention of provid- ing focus to the current research and a backdrop to the theoretical framework of Tinto's (1975) model, major findings are reviewed. Spe- cifically, findings from the comprehensive reviews that correspond to the selected individual characteristics studied in this thesis are presented. In addition to this research, studies that preceded and had a strong bearing on the development of Tinto's model (relat- ing more closely to the interaction of individual and environmental characteristics) are also important to discuss. Therefore, this chapter is organized into three sections. Part I briefly reviews national studies of attrition rates in compari- son to those found at Michigan State University. The second part of 23 this chapter reviews research findings of selected precollege char- acteristics that are associated with collegiate attrition. The final section of this chapter reviews studies that provided the theoreti- cal underpinnings for Tinto's model. A review of the model is pre- sented, as well as the findings of follow-up studies to Tinto's concepts. Parts II and III are intended not only to familiarize the reader with the subject area, but also to focus and support the research hypotheses stated in the first chapter. The Rate of Attrition Nationally, Summerskill (1962) has been one of the most often quoted researchers on attrition rates. He reviewed more than 35 studies conducted during the period from 1913 to 1953, covering both national and institutional samples. As a result of his work, like Kolstad (1977), he also cautioned readers about the definitional prob- lems with the phrase "attrition rate." Some studies only reference students who drop out at the end of one year, two years, etc., and return or transfer at a later time. Other investigations examine the permanent dropout from higher education. Regardless, Summerskill cited the more well-designed national studies of McNeeley (1937) and Iffert (1957) as presenting results from more carefully researched efforts. McNeeley (1937) found that 62% of the entering students to a particular institution left within four years. Of those students who left, 17% transferred, leaving approximately 45% who permanently dr0pped out. Iffert (1957) found that approximately 40% of the 24 entering students permanently drop out from higher education. Both authors also found a slightly higher rate in the less selective public colleges and universities as opposed to the more exclusive private institutions. Further, one-half of the total rate who dropped out did so before the s0phomore year. It is important to consider that not only do freshmen have the highest rate of not returning for their second year (based on these and other studies), and given the theory of the dropout process postulated by Tinto (1975), it may reasonably be concluded that they also have the highest percentage of permanent dropout of any class. Following the work of McNelly and Iffert, Eckland (1964a, 1964b) published two studies that challenged Summerskill's analysis of pre- vious work. He found that the number of students who re-enrolled at another institution or eventually returned to their first school of matriculation was higher. His work demonstrated that just over 70% will graduate from some institution, and this may run as high as 74%. Therefore, the permanent dropout rate is reduced to an amount of approximately 30%. However, while Eckland showed a variation from McNeeley and Iffert, Summerskill did note that he found broad varia- tions of 12 to 82% in the 35 studies he examined. Yet, despite these differences, the authors unanimously concluded that, where possible, the term "dropout" must be unambiguously defined and the institutional type specified. A major limitation of the above studies, in addition to those mentioned, is that they did not present a true representation of all higher-education institutions. Fetters (1977) pointed out that these 25 previous studies did not use a "representative probability-based sample" and ran high risks of bias. He cited the National Longi- tudinal Study (NLS) for the High School Class of 1972 as correcting for this bias as approximately 50% "of the sample entered over 1,800 institutions of higher education" (p. 19). As a result of this work, it was reported that after two years, 23.5% withdrew from all four- year colleges (5.28% due to academic dismissal and 18.26% were vol- untary). When compared to type of institutional control, public four-year colleges lost 28.5% of all entering students after two and one-half years as compared to 22.3% for private institutions, which was found to be significant at p = .01 (pp. 22-24). However, in contrast to Fetter's point about sample repre- sentativeness, it is important to note that the NLS follow-up occurred after only two years from the point of initial matriculation. Yet, as Kolstad (1977) noted, the number of students who drop out has been declining over the past 22 years. (See Table 2.1.) How does MSU compare with these national data? To what extent are they similar or dissimilar to the above findings? A preliminary report produced by the Office of Planning and Budget at MSU yielded the figures found in Table 2.2. This report is preliminary in that it does not take into account those students who graduated at the end of a practicum or internship experience and were not enrolled for academic credit during the term. Dr. William Rosenthal (1982), analyst, stated that this missing group of students may cause an error factor of approximately 3 to 4%. The data base is currently being updated to accommodate this finding. 26 Table 2.l.--National rates of attrition by research source. Attrition Rate Fall of . Cg;;gge F311?!5”“ 12a§eE§ilig§; & S°urce Universities 1950 4 years 46 Iffert, 1957 1959 4 years 39 Trent & Medsker, 1968 1960 4 years 22 Bayer, 1968 1961 4 years 35 Astin 8 Panos, 1969 1961 10 years 21 El-Khawas & Bisconti, 1974 1966 2 years 25 Adams, 1969; Jaffe & Adams, 1969 1966 4 years 28 Astin, 1972 1966 5 years 25 El-Khawas & Bisconti, 1974 1972 2 years 23.5 Peng, Asburn, & Dunteman, 1976; NLS Study Source: Kolstad, 1977, p. 4. Given an adjustment of -4% to build in the missing graduates, it is still readily seen that after a six-year follow-up for the class of 1973, five-year follow-up for the class of 1974, and so forth, the rate of attrition from MSU is somewhat higher than Kolstad's (1977) observation. However, the studies presented in Table 2.1 reflect dropouts from higher education, whereas the MSU data reflect only those students who left the institution and did not necessarily drop out of higher education altogether. Recalling Eckland's (1964) find- ing that as many as 74% graduate from some institution, it cannot be concluded from the MSU data that 35.9% in 1973 or 38.6% in 1974 left higher education entirely. Yet it would be unlikely that a significant 27 cop N~¢.A ~.m mme m.o~ owe.m ~.AP mpm._ Po. _ mem_ oo_ “No.5 A.~ ea, N._N Ako.m m.m~ Nom._ _. 5 meme co. oeo.~ 5.? _N. F.me 5mm.¢ ¢.~m emN.~ A. _m “Amp co, Poc.u o.P we _.- 08¢.F m.mm mom.~ ¢.~m 88¢.N chap oo_ mmo.k m. mm 0.8 «we ~.m¢ meo.m m.om aem.m mum, oo_ .om.m p. op o.m com o.mm Noo.~ 5.5m moo.¢ «km. cop ~m~.o P. m 5., mm, m.mm o~¢.N o.No omp.e mum. 5 .oz 5 .oz 5 .oz 5 .oz 5 .oz _aaoe .eeeo oesaemo< oceucmpm coca um__oecm 502 Name _Pma eaa> twp—05cm ppwum meowmm umpmzcmso .Amwmp .mp xemzcmwv mpcmuaum 3m: uwpmmeou mo coppweupm\mucmumwmemauu.N.N «Fame 28 number of those groups transferred and graduated from another school, especially in view of national findings. Despite the judgments and rationalizations one may choose to explain the MSU figures or those from another institution, they must be compared not only in and of themselves, but also with projected enrollment patterns. Further, with both sets of figures in mind (see Chapter 1, page 3), values need to be placed on the acceptability of those data. As a context for future readers of this thesis, it is important to note that the president of MSU has chosen not to find the current figures appropriate and acceptable. As this research is being conducted, a retention task force, in addition to specialized recruit- ing strategies, has been commissioned by the president's office. In conclusion, there is still a great variance in the attri- tion rates, depending on the size and control of the specific insti- tution under review. Nationally, across all types of four-year colleges and universities, there is an apparent decrease in the number of students who never finish a degree in higher education. The most important data to consider are those collected by the interested institution; this should be done in conjunction with enrollment pro- jections. Finally, it is noteworthy to recognize that leaving higher education may, in fact, be an appropriate and healthy decision for some students: But, recent data (Astin, 1975) suggests that such students represent a small minority and that students usually leave college for negative reasons: boredom, finances, poor grades and so forth. Positive reasons, such as a good job offer, are given by less than ten percent of the dropouts. It is easier to believe that students are taking time to "find them- selves“ than it is to confront the limitations of institutional programs and policies. (Astin, 1977, p. 260) 29 Precollege Characteristics and Their Importance for the Dropout There is little question that the individual background char- acteristics of the entering college student have an effect on the student's overall chance for success. Ample evidence exists to sup- port the finding and can easily be found in the aforementioned com- prehensive literature reviews (Spady, 1970; Cope & Hannah, 1975; Tinto, 1975; Pantages & Creedon, 1978). Most recently, a study involving a series of precollege characteristics encompassing demo- graphic, financial, academic, motivational, and socialization vari- ables of three freshman classes "at a private liberal arts college in Kansas" showed the continuing importance of such factors. Hutchinson and Johnson (1980) used a discriminant-analysis procedure to isolate variables that separated persisters from withdrawers and academic dropouts or those students who did not re-enroll for their second year and had less than a 2.00 GPA out of a maximum 4.00. They found six variables formed a discriminant function that correctly classi- fied 80% of the dropouts, 63.4% of the persisters, and 39.1% of the group who voluntarily withdrew for a combined figure of 63.4% who were correctly reclassified. Among the six variables, four were directly related to the category of precollege characteristics: high school GPA, father's occupational level, orientation to a vocational goal, and the high school counselor's estimate of success. Further, as documented in the literature reviews, the scope of such characteristics is extremely broad, and new variables are added as the understanding of the college dropout becomes more clearly 30 focused. However, Tinto's (1975) analysis appropriately concluded that some characteristics present a greater influence on the eventual attrition/persistence of the student than others. He went on to clarify these factors as those that pertain to the family (such as socioeconomic status and income level), educational experiences before college entry (i.e., size and type of high school and evaluation of this experience), individual attributes (i.e., measures of academic aptitude, academic achievement, and sex), and expectations of college "attainments" (i.e., expected, level of college degree, academic performance, college major, and extracurricular involvement). The purpose of this section is to present the major findings of the com- prehensive literature reviews and those of more recent studies to familiarize the reader with this aspect of attrition research and to provide support and focus for some of the hypotheses (notably Hypothe- ses l, 2, 3, and 8) stated in Chapter I. Individual Attributes Sex, race, individual measured aptitude, and academic achieve— ment are all obvious reflections of the individual. In some instances, their effect on attrition/persistence documents an obvious conclusion, whereas in others more penetrating analysis is required. Cope and Hannah (1975) found that, for a majority of the studies, sex differences did not present significant distinctions in total attrition. This was confirmed by other comprehensive reviews. However, the rate of attrition and reasons given for leaving have differed by sex. Citing studies by Demos (1968), Nelson (1966), 31 Astin (1964), and Tinto (1975), Pantages and Creedon (1978) found that women tend to drop out with greater frequency than men and do so at schools that have a high ratio of men to women. Additionally, once a female student has left, the chances of her re-enrolling to complete the degree are much less than for men. Yet, more women will eventually graduate within the four-year period, while, as Astin (1972) found, more men will eventually complete their degree require- ments beyond the traditional period. But as Spady (1971) found, more men tend to be drapouts for academic reasons, whereas more women tend voluntarily to withdraw. This conclusion is significantly related to the differences found among the reasons given by each sex. In a review of more than 15 studies dating to Iffert (1957), Cope and Hannah (1975) found that the differences were primarily a result of the sex roles filled by men and women. They concluded: Intellectual-aesthetic and social orientations, which are more central to the feminine role, tend to be related to attrition for women but not for men. Feelings of adequacy and competence, more central to the masculine role, are related to attrition for men but not women. (p. 16) Yet, such a conclusion is time bound and changes as men's and women's roles continue to evolve and more women enter the work world. Bean and Creswell (1980), studying reasons given by female drapouts at a religious coeducational, liberal arts college in a midwestern city, found that women who dropped out between their first and second years cited: the educational experience would not be helpful in getting a job they did not feel a sense of self-development their commitment to the institution was very low they had family responsibilities. (pp. 320-27) 32 Although the gap between dropout reasons for men and women will probably remain, it may be narrowing somewhat, especially as more women view education as a step to a successful career. Again, however, as with other attrition studies, sex differences must be considered in the context of the particular institution to provide the most meaningful analysis. The variable of race presents similar conclusions as did sex. That is, results are not consistent over time, racial-ethnic groups differ in the reasons they give for persistence/attrition, and the results tend to reflect the prevailing social climate. For example, in 1973, Astin found no significant difference in the proba- bility of Black, Oriental, or American Indian students' graduation rates, although they were slightly lower than majority students' rates. However, in 1977 his project group found that while the per- sistence rate for Black students was lower than that for Caucasian students, this largely depended on the institution. Predominantly Black institutions graduated a greater percentage of Black versus non-Black students, whereas the reverse was true for predominantly White institutions. Holding preparation and ability constant, Black women had a higher persistence rate than White women (Astin, 1975). He went on to point out that Black students are more likely to drop out for reasons of money and marriage than are Caucasian students. Further, persistence was positively related to minority students' access to nontraditional ways of demonstrating knowledge and credit by examination (Sedlacek & Webster, 1978). Again, institutional 33 context asserts itself as a powerful variable preventing one from making sweeping generalizations across all possible institutions. Turning to measures of academic aptitude and achievement, it would seem logical for one to conclude that these indices are fairly strong predictors of collegiate success. But, as each of the compre- hensive literature reviews is examined, this finding is only half correct. Cope and Hannah (1975) began their review on this tapic by stating that, in general, it has been found that dropouts tend to have lower achievement scores and measured aptitude than nondropouts. Further, these differences have often been found to be significant. They pointed to a study by Ivey, Peterson, and Trebbe (1966) as one of the better studies that represent the general conclusions men- tioned at the beginning of this discussion. Using high school class rank, SAT scores, and some personality measures, they found "high school rank is the most effective predictor of collegiate success and the CEEB-SAT provides a significant addition to [high school rank]" (p. 202). Yet, Cape and Hannah cautioned readers of the term "statis- tical significance." They pointed out that with large samples sig- nificant findings usually occur with almost any difference. Thus it is important for reviewers to separate what is significant from what is substantive and apply appropriate judgment that would lead to practical and meaningful action. _;..__ Tinto (1975) began to clarify the findings of ability, achievement, and aptitude by suggesting that the importance of their association with collegiate success is not meant to imply success 34 with respect to persistence. Referring to the ability to achieve in the collegiate setting, Astin (1972) found that high school GPA is a better predictor of success in the college classroom than are pre- college aptitude tests. Vaughan (1968) stated that the literature has clearly pointed to the need to separate those students who volun- tarily withdraw from those who are academically dismissed. One can surmise from Vaughan's findings that measures of ability, achieve- ment, and aptitude are more strongly related to academically dismissed dropouts than those who voluntarily withdraw. Citing several studies, Pantages and Creedon (1978) clarified this topic by showing that the association between measures of ability, aptitude, and achievement and college persistence is erroneous: "Several studies have pointed out that although high school perform- ance is an accurate predictor of college academic success, it does not predict persistence at the college level" (p. 63). They further supported their conclusion demonstrating that Eckland (1964) found high school class rank to be twice as powerful a predictor of which students would drop out, as opposed to the relative permanence of their dropping out. In a further study by Johansson and Rossmann (1973), who supported Vaughan's (1968) study, he found measures of scholastic aptitude to discriminate more powerfully between "volun- tary dropouts" and academic dismissals. There was relatively little difference between voluntary dropouts and persisters on the variable of scholastic aptitude. Collectively, then, the literature in this area has pointed to the importance of examining the precollege scholastic achievements 35 and aptitude of students with respect to signaling academic success in college and those students who are likely to drop out due to aca- demic failure. However, while these measures have a bearing on per- sistence, the relationship provides minimal guidance in explaining the voluntary dropout. In fact, an interesting study conducted by Gurin, Newcomb, and Cope (1968) demonstrated that background experi- ence, particularly related to the family, had more influence on per- sistence than did the variable of "academic preparation." Family Background Gurin, Newcomb, and Cope's (1968) research studied character- istics of entering freshmen that were associated with attrition. They constructed two scales: Cosmopolitanism and Academic Preparation. The cosmopolitanism scale contained variables such as type of home- town, size of high school, and family background. Academic prepara- tion contained those factors reviewed in the above discussion. Spady (1970) examined this study and found that when both scales were con- currently controlled, cosmopolitanism showed a stronger relationship to persistence than, as noted above, academic-preparation variables. He wrote: The basic implication of their finding is somewhat contrary to what some researchers might expect to emerge from a set of variables such as this, for . . . the advantages thought to accrue to individuals with particular kinds of attitudes do not exist independently of their family background. (pp. 69-70) Many factors can comprise the category of family background, not all of which bear a strong association with attrition/persistence. Items such as socioeconomic status, parental education, family income, 36 racial background (already discussed), family relationships, and parental expectations would seem appropriate for this discussion. Although a number of studies have investigated the effects of socioeconomic status and family income, Summerskill (1962) pointed out that the results have not been consistent. Citing an early study by Pearlman (1952), who controlled for scholastic aptitude and high school performance, he found that there were no significant differ- ences between dropouts and persisters with respect to parental occupa- tion, income, and other related variables. However, investigating another study by Farnsworth (1955), Summerskill pointed out the importance of the family's cultural and educational values as they relate to persistence. That is, in families in which education and higher learning are valued, in which intellectual matters are openly discussed, and family relationships are open and supportive, students tend to be more successful and to have a higher rate of graduation from college. Pantages and Creedon (1978) supported Summerskill's earlier findings examining a number of studies that investigated socioeconomic status, parents' education, family income, and social status. The most important association they found between these variables and persistence was from research by Astin (1973) and Eckland (1965). The results were that these factors did not have a direct relation- ship to the process of dropping out; however, they were significantly related to which students were more likely to return after dropping 37 out. Therefore, the overwhelming conclusion was that socioeconomic- status factors hold minimal value for predicting attrition. Yet, while socioeconomic-status variables are not strongly related to attrition and persistence, the natures of family values, relationships, and educational expectations are. Spady (1970) briefly reviewed five studies that supported these other aspects of family background: - Congdon (1964): more successful college students character- ized their families as open, accepting and had a positive relationship with their parents. - Jones (1955) and Weigand (1957): students who were poor achievers described their families as being overprotective, more demanding, and having increased tension and disturbance as opposed to the families of higher achieving students. - Hood (1957): students who were dropouts reported that they were not certain of their parents' support for their educa- tional pursuits and were less likely to openly confront issues on which they would tend to disagree. - Trent and Riegle (1965): students who were graduates said they turned to their parents for advice, were praised by them and felt a strong interest from them toward their educational degree. This finding was significantly different from those who dropped out. (p. 70) It is easily seen that the quality of relationships between a student and his/her family, including the expectations that parents hold for their students' education, is a very strong influence on persistence. However, it is important to look beyond simply acknowledg- ing this influence and to examine its manifestation-—commitment. In a study involving 1,407 students and 1,331 parents at three midwestern liberal arts colleges, Hackman and Dysinger (1970) investigated commitment to college on the part of the students and the influence of parental expectations on degree of commitment. Their study was very carefully conducted as they examined the differences 38 between persisters, transfers, and returnees, academic dismissals, and voluntary withdrawals. Appropriate to this discussion, parental expectations were found to be significantly higher for persisting students compared to all other groups. This influence was found to be most different with those students who voluntarily withdrew: The results suggest that a student's home and family may be highly important in determining his reaction to the college experience. . . . In addition, the more parents indicate that they believe the student will perform well in college, the stronger the measured commitment. Finally, when students or parents indicate that "it has always been expected" that the student would attend college, commitment tends to be higher. . . The present results leave little doubt that the parents' own commitment to their child's college education is si nificant in understanding who persists and who does not. (p. 3201 To summarize these various studies, socioeconomic factors and social status bore only a weak relationship to persistence. A majority of these articles pointed to persisting students as coming from more affluent homes with better educated parents, yet some other signifi- cant studies failed to replicate this claim. Upon further investiga- tion, it was seen that persistence was not so much related to what the family is, but how it acts in relation to one another. A strongly supportive family may be characterized as one that is genuinely interested in the future of its members and holds high expectations for their success--a finding that may not only positively influence persistence in college but in other settings as well. Educational Experiences Before College Entry Aspects of this subject have already been covered in the discussion of academic preparation as reflected by high school GPA, 39 class rank, and scholastic aptitude. However, it is important to review the effect of high school size and type--that is, public or private. Summerskill (1962) cited studies by Thompson (1953) and Suddarth (1957) and suggested that graduates from larger high schools are more likely to persist in college than are those from smaller high schools. Yet he clearly stated that these findings were not conclu- sive and that he was unable to explain the results with any certainty. Pantages and Creedon (1978) reviewed the work of Sexton (1965), who supported the previous finding but went on to point out research showing that only the very small high schools (under 79 students) had higher attrition rates. Interestingly, Cope (1972) found a con- nection between college size and high school size, suggesting that a "breakeven point" may exist: Those students who came from smaller high schools had better persistence rates if they attended small colleges, and vice versa. He went on to explain these results by say- ing that the "fit" between the student and the school is the crucial linkage. A more recent study by Downey (1978) investigated the differ- ences between entering freshmen from different-sized high schools who attended Kansas State University. Using the American College Testing Program data to equate students and stratify them according to high school size, he investigated first-semester grades, level of extra- curricular involvement, and persistence. The results of his study showed that although students from various-sized high schools differed, the majority of the differences were in nonacademi c school accomplishments. 40 Students from smaller high schools were more frequently involved with nonclassroom activities than were students from larger high schools. There was only minimal information that suggested students from smaller high schools were not performing as well academically and that these same students showed a ''lower rate of persistence after three completed semesters" (p. 355). Therefore, the major conclusion was that there was virtually no relationship between attrition/persistence and high school size, with the exception Of very small high schools. Unfortunately, the literature describing the type of high school (public or private) and its effect on persistence was not any more helpful than the review pertaining to size. For example, con- flicting findings existed between Sexton (1965) and Astin (1973). Sexton, summarizing from earlier work, stated that the results pointed to public rather than private school students having higher persist- ence rates. And, with a limited rationale to support her hypothesis, she suggested that students in public schools had higher levels of achievement needs. An Opposing view by Astin (1973) found the exact opposite in his research and stated that the academic standards of private schools are higher and therefore Offer better preparation for the rigors of college studies. Despite these findings, size and type may not be the important factors that deserve attention. Rather, other elements such as curriculum, facilities, quality Of the faculty, high school student resource services, and community support appear to be the primary influences in the quality of a student's prepara- tion. In fact, Spady (1970) located four studies in which dropouts. 41 cited poor preparation on the part of their secondary schools, attributing this to some of the variables just mentioned. Further, Dyer (1968) studied the effects of high school preparation on per- sistence and concluded that characteristics relating to the quality of facilities, faculties, and so on, affect educational attainment and consequently persistence. It may be that all too Often educators ignore the obvious and dismiss criticisms by former graduates who dropped out of college as rationalizations for their own lack of motivation and ability to learn. It is much easier to discount the comments of "disgruntled students" than it is to confront the limita- tions of one's institutional program and facilities. Eypectations Of the Collegiate Experience It has already been pointed out that expectations, particu- larly from one's family, play an important role in the success and persistence of the student. Hackman and Dysinger (1970), in the same article, went on to show that the expectations that the student has of him/herself are no less important. Mean scores from items such as "Highest Level of Education Planned" and "Plans to Continue at the Present College" were higher for persisters than for any other cate- gory of nonreturning student and most different from the group, "voluntary withdrawals" (p. 316). Other studies have strongly supported this finding. Spady (1970) stated that it is probably one of the most unambiguous find- ings that the literature supports as being associated with persist- ence. He cited the work of Sewell and Shah (1967), Trent and 42 Medsker (1968), and Astin (1964), pointing out that these studies showed that persisters had (1) more definite educational plans, (2) had intentions of graduate or professional school more fre- quently, and (3) generally saw the importance of college more readily than nonreturning students. Additionally, work by Astin (1973) and Rossman and Kirk (1970) added further support to these results by showing that those students who do not expect to graduate have lower graduation rates as compared to those who expect to graduate. Rossman and Kirk (1970) suggested that these students have a tendency to create situations in which they can't succeed and, in fact, promote their own self-defeating behavior to the extent of it becoming a self- fulfilling prophecy. Finally, as Fetters (1977) demonstrated, even after controlling for socioeconomic background and high school GPA, students with low aspiration had higher withdrawal rates than those with higher levels of aspiration in college. If a student's expectations and their strong relationship to persistence and attrition are one of the more "unambiguous" findings of past research, they are also one of the most persistent over time. Wessell, Engle, and Smidchens (1978) showed marked differences between the persistence and withdrawal rates of 2,438 freshmen at a four-year midwestern college, based on expectations they held for studying a particular major or being undecided. Those students who did not have any expected direction of study tended to withdraw with greater frequency than those students who did. This finding held true even for students who transferred. Hutchinson and Johnson (1980) also found that the expectations one held for a particular vocational 43 goal successfully discriminated between persisters, academic drop- outs, and voluntary withdrawers. Persisters were more likely than the other groups to aspire to a particular goal. These conclusions began to bring to the surface an issue related to expectations—-that is, the degree of decidedness and certainty one holds for an academic major and career goals. In a study of 767 students at an urban, public commuter college, Muskat (1979) found significant differences between returning and nonreturning students where persisters were more likely to have decided on a course of study and felt a high degree of certainty about the decision and decided on career goals feeling equally certain. Tinto (1975) would argue that measuring one's expectations and/or plans as well as the certainty felt toward those plans is actually a measure of commitment. In his model (presented later in this chapter),luamade the point that commitment is placed after family background, individual attributes, and previous educational experi- ences: "It is suggested that one's commitments are themselves a reflection of a multidimensional process Of interactions between the individual, his family, and his prior experiences in schooling" (p. 103). More important not only to his model but also an Observa- tion of the literature, Tinto concluded that commitment is the single most important variable determining college completion. Precollege Characteristics in Retrospect The past several pages presented studies and conclusions of precollege factors associated with the college dropout. While raising 44 many interesting questions and suggesting various notions about why such variables may be significant, they lacked a comprehensive frame- work within which to explain the findings. Summary statements by Spady (1970), Tinto (1975), and Cope and Hannah (1975) agreed that the literature in this aspect of studying attrition/persistence is "conceptually incoherent and methodologically uneven" (Terenzini & Pascarella, 1978, p. 349). As will be seen in the next section of this chapter, the study of the college dropout has begun to move beyond this stage into more advanced and potentially beneficial under- standings of why some students return to college, eventually graduat- ing, and others do not. Attrition: Toward a Formulation of Theoryy Although a majority of the research on attrition consisted of defining individual and institutional characteristics associated with dropping out, the focus began to shift in the early 1960s. At that time, a more interactive way of viewing the dropout phenomenon caused researchers to conclude that both the student and the collegiate envi- ronment played a role in the process. Neither the student nor the institution was seen as the primary contributor to attrition and/or persistence. Some of the studies that explored this approach were conducted by Keniston and Helmreich (1965), Pervin and Rubin (1967), Nasatir (1969), and Rootman (1972). As will be seen, these studies brought the hypothesis Of "student-college fit" into clear view. As reported by Cope and Hannah (1975), Keniston and Helmreich (1965) explored the concept of congruence between the individual and 45 his/her environment. Specifically concerned with the identity development of students, they found that the degree of identity development was positively related to the amount of congruence the individual felt within the environment (Cope & Hannah, 1975). That is, the more compatible the college environment is with the individual student's beliefs, values, and aspirations, the more congruence exists, and this facilitates individual-identity formation. This belief was very similar to a contemporary Of the researchers who developed the theory of cognitive dissonance. Festinger (1957) stated that two cognitions held by the individual are said to be dissonant when they are not in a logical relationship to each other or are incompatible with the person's current experience. Festinger went on to note that the dissonance was "drive-like," and the resulting tension was unpleasant, causing the person to do something to reduce the tension (Jones & Gerard, 1967). In view of the current subject, in order for greater identity development to occur and the tension to be reduced, the student may seek a new environment, obviously resulting in drOp- ping out of the present institution. Pervin and Rubin (1967) undertook a study at Princeton Uni- versity that followed from Keniston and Helmreich's work. They hypothesized that a lack Of "fit" between the students and the char- acteristics of the college created dissatisfaction within the student. As the differences between the way the student viewed him/herself and the image held of the college grew, dissatisfaction eventually increased, leading to the point of dropping out. They also stated that differences between the self and the college were more associated 46 with nonacademic reasons than academic ones but did not support this statement with any rationale. Regardless, the data from their study supported the belief that the degree Of discrepancy between the student's view of self and of the college was positively related to the probability of dropping out. They did find some evidence to support the idea that dissatisfaction was also related to the aca- demic arena, yet, as noted above, a majority of their findings rein- forced the hypothesis that the discrepancies were related to nonacademic reasons: These data indicated that the main source of variance was in the way students perceived the college. . . . Those in the high discrepancy and high dissatisfaction group rated the college as higher on the following items: dull, bored, uncreative, competitive, tense, ritualistic, conventional, conservative, Republican, status-oriented, militaristic, repressive, conforming, coercive, detached, snobbish, cold, callous, unfriendly, and intolerant. (p. 293) However, the researchers cautioned that their study did not take individual-student differences into account. They suggested that future research distinguish between the unique characteristics of the individual and the context of the environments they perceive. Concurrent with the work of Pervin and Rubin, Nasatir (1969) researched individual student differences and how they interacted with living environments on the campus. Apparently following the previous work Of the congruence model, he explored the reconmenda- tions Of Pervin and Rubin (1967) and conducted one of the more syste— matic studies at that time. His research involved the administration of a series Of psychological tests and attitude questionnaires to entering students at the University of California, Berkeley. His 47 intention was to determine which groups of students were primarily "academically oriented" or "socially oriented." The academically oriented students were defined as those students who felt "the most important reason for attending college is to obtain a basic general education and appreciation of ideas," whereas nonacademically oriented students viewed "the development of the ability to get along with people, prepare for marriage" as the most important reason (p. 293). In a similar analysis of following where these students resided, Nasatir classified residence halls as either academic or nonacademic. With these categories of living situations and individuals, the student dropout rate was measured. As might be expected, some of the results tended to confirm common-sense reasoning, yet one raised an interesting issue: - The dropout rate was higher in "non-academic“ residence halls compared to "academic" residence halls. - The dropout rate for non-academic students was higher than for academically oriented students. - Academically oriented students had higher dropout rates in non-academic living environments than in academically classi- fied residence halls. - The dropout rate for students who held a non-academic orien- tation was twice as high as the academically oriented students in non-academic settings. - HOWEVER-- - The dropout rate for non-academically oriented students in non-academic settings was still lower than their rate of dropout in academic settings. (p. 297) This last finding raises a new concept from Nasatir's results and for subsequent research--that is, integration. This finding points to the importance of congruence in terms of its logical out- come. The concept Of integration implies the positive effect of bringing the environment and the student into a harmonious relationship. 48 Nasatir pointed this out by stating that while an academically oriented student may not find support for his/her point of view in the nonacademic living situation, he/she is nonetheless supported by the "intellectual activities from the university culture at large. . . Without an academic orientation, and without a supportive con- text, students manifest a high rate of failure" (p. 296). The same holds true for nonacademically oriented students, who have lower attrition rates in nonacademic environments than they do in academic environments. In 1972, Rootman wrote Voluntary Withdrawal From a Total Adult Socializing Organization. With Nasatir's results in mind, he studied the effect that integration had on the retention of first- year students at the U.S. Coast Guard Academy. He proposed that the environment presents various expectations on the "role" of the cadet. and these interact with the individual's personality, needs, and goals. When a conflict occurs between what is expected and the person's interpersonal domain, stress results, eventually leading tO "strain." Beyond a certain point (depending on the individual), the cadet begins to look for coping mechanisms to increase the "fit" between the person and the role. In some cases, the chosen mechanism is withdrawal, thus exercising control over the environmental expecta- tions by changing the setting. Rootman found support for his hypothesis when such variables as personality, actual and perceived attachment to persons in the academy, interests, and values were all significantly related to withdrawal: "Whether or not a person will survive in a military 49 academy for a year appears to be largely a function of the degree to which his own prOperties 'fit' the role of the cadet at entry and the degree to which he 'fits' the group with which he is socialized" (p. 266). One of the major contributions of this research is found in the preceding statement. That is, the student must not only fit the structural environment as demanded by the institution, but also the social-interpersonal environment as found within the students, faculty, and staff. As seen in Nasatir's work, both environments must be taken into consideration as each has its separate influences that can help support or alienate the student. A clearer definition of the differential effects will become apparent in the subsequent research review. Nevertheless, to become socialized in an institu- tional setting, according to Rootman, appears very similar to what Nasatir (1969) referred to as integration and what Pervin and Rubin (1967) referred to as congruence. The theory of person-role "fit" and integration permeates a great deal of the literature on why students leave a particular institutional setting or higher education altogether. Other studies by Pace (1962), Stern (1960), Astin (1964, 1965), and others have strongly supported the notion that attrition research must approach the phenomenon by observing the interaction of the student and the environment. Their findings have consistently corroborated results that found persistence positively correlated with integration. Even without the explanatory backdrops Of congruence and "student- environment fit," this finding has endured. In 1975, a nationwide attrition study of 41,000 undergraduate students at 358 two- and 50 four-year institutions was conducted by Astin. Although he did not attempt to present a theoretical framework within which to explain his findings, he concluded: After examining the fit between the student and institution, it appears that in general, persistence is enhanced if the student attends an institution in which the social backgrounds of other students resemble his or her own social background. Such interactions are most apparent with the town size, reli- gion, and race of the student. (pp. 144-45) As noted above, although Astin did not presuppose any expla- nation for his findings (bringing them together into a coherent rela- tionship), previous literature did offer some interpretations. That is, student and institutional characteristics interact, leading to varying levels of congruence or dissonance. As will be seen in sub- sequent research, the students develop certain expectations about what is an acceptable or unacceptable level Of integration; if the desired (expected) level of integration is not achieved, the student may choose to withdraw. In 1970, a significant new direction in the study of attri- tion was undertaken by Spady. Noting what he referred to as a lack of theoretical and empirical coherence in the literature, he developed the first interdisciplinary model of the dropout phenomenon. The model was based on the sociological contributions of Durkheim's studies of suicide, as well as the research ("1 attrition that had been conducted up to that time. While Durkheim discussed and described many specific types of suicide, he noted that "integration" was an important causal factor in many of the suicides. He found that as a person began to withdraw from society, the chances for suicide increased. Further, 51 Of those who committed suicide, many could be described as having very little social contact or few social connections. Yet, social affiliation as observed by Durkheim was only one level of integration. He also noted that a lack of "normative congruence" limited the degree to which an individual became integrated into the social environment-- that is, agreement between individual beliefs, values, and attitudes and those held by society. With these Observations and research, he wrote: The individual alone is not a sufficient end for his activity. He is too little. . . . There is in short, in a cohesive and animated society a constant interchange of ideas and feelings from all to each and each to all, something like a mutual moral support, which instead of throwing the individual on his own resources, leads him to share in the collective energy and supports his own when exhausted. (E. Durkheim, 1951 p. 210) Spady (1970) was quick to point out that suicide and dropping out of higher education are far apart in terms of the severity of their consequences. Yet his intention was to show that the condi- tions surrounding each experience are very similar. That is, the degree to which the individual becomes integrated with the "society" or institution in higher education in each experience is a key vari- able. Further, the factor of integration occurs on two levels: socially (affiliation) and cognitively (normative congruence). Therefore, according to Spady, dropping out is to be viewed as a process involving input variables brought to the situation by the student and process variables presented by the existing collegiate environment. He stated that the collegiate setting provides Oppor- tunities for reward. In turn, the rewards gained by the students are 52 indirect measures Of the amount of integration achieved in the experi- ence. Therefore, the individual brings to the environment both aca- demic aptitude and a particular set of attitudes, values, and beliefs (as a function of family background). The collegiate environment presents Opportunities for the student to become integrated in the setting through classroom experiences, academic assignments, support services, and other peers. As a result of the interactions between the student and the environment, rewards may or may not be found (Spady, 1970). Academically, the rewards that can be obtained through classroom interactions are grades and the student's own intellectual development. Socially, the student may experience varying levels of friendship support and consistency in values, attitudes, and beliefs. While the degree Of normative congruence is a reward having a direct bearing on social integration, lack of congruence on an institutional level may be without consequence. That is, while the student may find the general philosophy of the institution to be at variance from his/her own, they may have found a particular faculty or depart- ment that supports their ideas and opinions in spite of the larger system (Spady, 1970). Yet, insufficient grade performance can also limit any amount Of integration achieved if the student is involun- tarily dismissed. "In my view, then, full integration into the . . . life of the college depends on successfully meeting the demands of both its social and academic systems" (Spady, 1971, p. 39). After developing this model, Spady went on to test it using the data collected from the 1965 entering-freshman class at the University of Chicago. He collected precollege information, including 53 family background, high school experiences, and expectations of college. After those students began their spring semester, a second questionnaire was administered to gather perceptions of the environ- mental influences, friendships, satisfaction with courses, and so on. After a four and one-half year period passed, it was determined that 38% of the students had not completed their degree and were labeled "dropout." Using rmittiple-regression analysis, the major findings Of this study were: 1. Significant differences in the interaction patterns between men and women were found to influence eventual persistence/attrition decisions. 2. Collegiate expectations, family background, and high school performance were influences but not primary determinants of interactions that occurred on the campus. 3. Intellectual development (stimulation in course work, expansion of cultural and intellectual outlooks, critical thinking, self-evaluation of academic performance) varies more with nonclassroom faculty contact than peer interactions and is seemingly unrelated to high school performance. 4. Intellectual development was found to be greatest for those who are predisposed to learning and have not had such experi- ences due to poorly equipped high schools or small community settings. 5. Yet, grade performance is strongly related to the academic quality Of the high school, academic performance in high school, and measured math and verbal ability. 54 6. Interpersonal contacts, though strongly related to social integration, tended to reduce high academic achievement. Spady stated that this is probably due to the extremely competitive and demanding academic climate of the school as well as the nature of the student selected. 7. Only indirectly influenced by entering characteristics, satisfaction is reached in the early months of the freshman-year experience as a direct function of academic, intellectual, and social interactions. 8. Institutional commitment is formed much earlier in the educational experience than was originally hypothesized. However, the level of commitment does change over time as a function of the experiences on the campus. 9. Dropout decisions for women are heavily dependent on com- mitment to the college formed by the fulfillment of interpersonal needs. Academic-related variables run second to this during the freshman year. 10. Dropout decisions for men are primarily based on grade performance and are only supported by a lack of social integration and institutional commitment. "What counts most is their ability to meet the formal performance standards of the faculty, irrespective of the more intrinsically based social and intellectual factors" (Spady, 1971, p. 61). 11. Those students who eventually graduate differ from early dropouts most dramatically on initial-year academic performance more than any other factor. Further, in the second, third, and fourth 55 years, the sex differences tend to diminish, with women becoming more academically oriented than their first-year counterparts (pp. 57-62). The fact that Spady conducted his study in the mid-19605 is very evident in his findings. The cultural role of men and women tends to be readily apparent and shows a definite relationship to dropout and/or persistence decisions. Thus the need continually to update research in this area becomes very clear. Regardless, Spady did make a major contribution to the understanding of the dropout process by bringing together much of the literature at that time. Further, he created a significant foundation and stimulus on which to base future research. One of the most notable efforts prompted by Spady's work was that Of Vincent Tinto. As will be recalled, Chapter I highlighted Tinto's 1975 model and presented it as providing the theoretical framework for this current investigation. Before a more complete description Of the model, a number Of factors should be considered. Upon initial review, it becomes apparent that the model ful- fills many of the conditions embodied in a strict definition of theory. That is, the model (a) seeks to describe a phenomenon, (b) attempts to provide an understanding of previously unexplained events, (c) more clearly organizes and interrelates previous knowl- edge, and (d) allows for more efficient research (Reynolds, 1975). The final condition, assisting with the prediction Of future events, is finding support through follow-up research of Baumgart and John- stone (1977) and Pascarella and Terenzini (1980). However, the research is not yet conclusive. A second and major reason for 56 choosing this model is that it allows the examination of institu- tional contributions to the attrition phenomenon. That is, while the model seeks to explain the process of completely dropping out of higher education, it more clearly focuses on why students leave a particular institution (the major intention Of this study). A third reason for selecting Tinto's model is that it is clear, concise, and easy to follow. The framework is designed with a solid basis in research, and the concepts are clearly delineated in its components. And finally, as noted above, initial research has begun to validate the major tenets Of the model and supports its value for application to other institutions in higher education. Although the design Of the model differs from Spady's (1970), there are important similarities. Both systems use Durkheim's theories of suicide as a basis for construction. Also, using the concept of integration, both models state that this outcome will depend not only on the educational environment but also on the back- ground characteristics of the student. Tinto (1975), as noted earlier, categorized these factors into family background, precollege school- ing, individual attributes, and collegiate expectations (manifested in levels of commitment). Further, both models refer to the colle- giate environment as being composed primarily of two subsystems: academic and social. However, the Tinto model explores the differ- ential effects of each dimension more carefully. He stated that as integration into each area has a different effect on the individual and is therefore complimentary to the other, a careful balance must be maintained. TOO much integration and involvement in one area 57 may have deleterious effects on the student. As an example, a student who is primarily committed to academic pursuits may do so at the expense Of developing an ability to form mature social relationships. Viewed from the values and assumptions of the educational-develOpmental theorists, such a pattern prevents the realization of the student's fullest potential (Knefelkamp, Widick, Parker et a1., 1978). While the above similarities are to be noted, important dis- tinctions separate the two models contributing to the superiority of Tinto's framework. First, Spady's model clearly emphasizes the role of social integration, but, as Tinto explained, both social and academic integration must be carefully examined, especially in view of their different effects. Second, a major distinction can be found in the role that individual commitment plays. Spady stated that persistence or attrition is associated with rewards that a student finds in the collegiate environment. Tinto emphasized that alterations of levels Of commitment both to the college and the academic degree lead to dropout decisions. Tinto (1975) stated, similar to Trent and Medsker (1968), that the student's precollege characteristics influence the formation of initial commitments to (a) Obtaining a college degree and (b) attending a particular institution. Commitment and its effect on persistence and attrition were clearly demonstrated by Hackman and Dysinger (1970). As will be recalled, the major findings of their study showed that levels of commitment formed by students before their freshman-year enrollment significantly determined their group (persisters to academic dis- missals) membership the following year. 58 However, Tinto (1975) extended the above findings by stating that the degree and quality of interactions that a student has once he/she has enrolled on the campus and over time (both socially and academically) determine the amount of integration achieved and ulti- mately alter both types of commitment leading to eventual graduation or dropout: Given individual characteristics, prior experiences, and com- mitments, the model argues that it is the individual's integra- tion into the academic and social systems of the college that most directly relates to his continuance in that college . . . [and] . . . given prior levels of goal and institutional com- mitment, it is the person's normative and structural integration into the academic and social system that led to new levels of commitment. . . . The higher the degree of integration Of the individual into the college system, the greater will be his commitment to the specific institution and to the goal of col- lege completion. (p. 96) Tinto also argued that forms of dropout behavior can be explained by this alteration of commitments. For example, students who transfer may have experienced low institutional commitment but higher levels of goal commitment. Students who exhibit relatively low goal commitment but are experiencing high levels of institutional integration may explain why some students stay in a particular col- lege and continually just "get by." Also, despite a certain amount of social integration, low commitment to obtaining the college degree may result in voluntarily dropping out from higher education alto- gether. Finally, Tinto argued that relatively low levels of both types of integration and commitment can lead not only to voluntary withdrawal but, in most cases, to academic dismissal. The longitudinal model proposed by Tinto can be found in Figure 2.1. 59 A.mnmp .oucw» Eoegv mzo_m_uuo PaoQOzo .1 _ _ _ _ . . . _ 1L . . . — _ - . _ . r. pzurh_xzou ah43uuo 4_oz— azsomuxu4~zdu mu_pm_amhumo szuumppmucH .Fp . . Aowmpv wcp~=memh mcwgummp ecu acmqupm>mo pm om ecu mp—mgmummm mm. acousum eoe cgmucoo appzume .op mm. mm. Aommpv ._a pa eacsoo he. etc: .5 Fe. _e. Aommpv .Pm um eaczoo me. weep encamema .m 5: mm 885.355558 3. 8::32»:§ES.~ Ne. Ne. Aommpv .Fa pa amazon me. . peasaopa>ao use spzocm .8 me. me. Aomm_v ._a pa eaczoo ow. m=o_puacapce apesaa .m _m. cm. Aomm_v .Fa um eaczoo km. mmepe>euu< _oogum .4 mm. mm. Aowmpv ._5 pm emczoo me. coepuaaapce eapauaa .m _8. mm. Aommpv ..5 pa emczoa Ne. accepuaaapce 2555 .N mm. mm. Aomm_v .Pm pa eoczoo we. mmeoe>eeu< beamuau< ._ awning. em”? .52 5.28 cowummwumm>cfi ucmcesu Opuumazm gucmwmom maor>mea .HH mmuuumucmwuweemou Auppenmepme Oqumnam eo comreagEoouu.F.m m_am» 81 With one exception, the reliability coefficients computed from the current sample data were consistently lower than those found in previous research. The differences ranged from .01 on the subscale Growth and Development to .32 on the subscale Faculty Concern for Student Development and Teaching. However, the subscale, Work, showed an increase of .18 from previous research to the study currently under investigation. As a result of these values, the CES II demon- strated low to moderate reliability. Although Downey et a1. (1980) and Pascarella and Terenzini (1980) would argue that their subscales are acceptable for measuring group differences, findings from the current study require that the questionnaire results be interpreted carefully. Data Collection and Statistical Analysis A majority Of the sampled students completed the SPS-ACT during their senior year in high school. However, some students may have taken the test as early as March of their junior year. To avoid a problem that would be apparent in two testing periods, "adjustments [were] made in the scores to eliminate any systematic advantages related to educational level and time'of year of testing" (ACT Assess- ment Program, 1980, p. 4). The SPS-ACT information used in this study was Obtained by merging the MSU Office of Admissions Tape for Fall 1981 freshmen with the tape of sampled students in this study. The second data collection Obtained through the use of the CES II occurred during the eighth week Of Winter Term 1982 or the 82 week of February 22, 1982. The questionnaire (coded for follow-up) was accompanied by a cover letter and addressed to each student in the sample. Residence-hall staff (resident assistants) were asked to give the questionnaires to their sampled residents, requesting that the survey be completed and returned in four days. The resident assistants were instructed to give the questionnaires to the students and only request that they read the instruction page and not pressure students toward completing the instrument in any way. Students were also provided with a brown, addressed mailing envelope and instructed to place the completed questionnaire in the envelope (sealing it) and return the material to the resident assistant. The completed ques- tionnaires were returned to the building directors, who, in turn, forwarded them to the researcher. Following these two data collections, SPS-ACT and the CES II, the information was held until after registration Fall Term 1982. Approximately two weeks after registration or on the tenth day of the term, the MSU Registrar's Office compiled an enrollment tape. This tape, with the assistance of the Office of Planning and Budget, was matched against the sample of students who returned usable surveys. As a result, a list of returning students and nonreturning students was compiled. Additionally, as mentioned earlier, the nonreturning students were further subdivided into two groups: those who were academically dismissed and those who voluntarily chose not to return. The information--return, nonreturn, dismissed, or voluntary--was coded into the master data file for statistical analysis. 83 To test the hypotheses, the subprograms Crosstabs, Discrimi- nant, and Stepwise Discriminant from Statistical Analysis Systems were used. The Crosstabs procedure provided a chi-square analysis of the returning and nonreturning students as a preliminary indicator of those variables that may be discriminators between the two groups. The Stepwise Discriminant Analysis isolated those variables that were the "statistically best discriminators," and the Discriminant proce- dure constructed a classification function with the "best discrimi- nators" using this function to predict the group membership from a student's score on the discriminating variables (Nie et al., 1975, p. 435). Analysis of Hypotheses 1 through 7 began with the Crosstabs procedure using a chi-square to test for group independence. The intention of these results, as already mentioned, was to provide an initial indication of the variables most likely identified as dis- criminators through the discriminant procedure. Further, chi-square reveals the direction of group differences, whereas discriminant only isolates the distinguishing variables. For example, given that discriminant isolates Peer Interactions as a variable that distin- guishes between returning and nonreturning freshmen, the information does not show which group had more (or less) frequency of interaction. The chi-square results can be employed as a means to clarify the direction of this difference. Following the chi-square tests, each of the seven hypotheses was tested using the Stepwise Discriminant procedure. This test isolates variables that combine to form a function that mathematically 84 forces the two groups of students to be as distinct as possible. In addition tO producing functions that may be statistically signifi- cant, it also computes the Canonical Correlation, Partial R**2, and Wilks's Lambda statistics, which are used to judge the presence of substantive or meaningful discrimination. If substantive discrimi- nation appears to be present, a classification table is produced through the procedure, Discriminant. The classification table then reveals the percentage Of students correctly classified as either returning or nonreturning, based on their known group membership. More than 50% of each group of students should be correctly classified to show that the discriminant function produces results better than chance. It is on the basis of substantive discrimination that each of the hypotheses was either accepted or rejected (Nie et al., 1975, pp. 445-48). TO test Hypothesis 8, each analysis was computed three times. The first analysis included total returning students compared with total nonreturning students. The second and third analyses compared returning men with nonreturning men and then returning women with nonreturning women. Differences between each group were noted. Summary This chapter described the methods and procedures used in the study, including the hypotheses, sample, instruments, data col- lection, and statistical procedures for analysis. The population included new MSU freshmen enrolled during Winter Term 1982 who lived in undergraduate residence halls and intended to pursue a four-year degree program. 85 The data for this study were collected through the use of the SPS-ACT program on file with the MSU Office of Admissions and by administering the CES II. These two instruments were used to measure precollege entering characteristics of the student sample and their subsequent experiences on the campus following enrollment along the dimensions of social and academic integration. The statistical procedure, Discriminant, was used to identify those variables that successfully distinguished between returning and nonreturning students, and a stepwise procedure, Wilks's Lambda, was used to identify which variables demonstrated the best discrimi- nating power. Chapter IV presents the analysis and findings from the data gathered for this study. CHAPTER IV PRESENTATION OF ANALYSIS OF THE DATA Introduction This chapter presents an analysis of the data collected to test the hypotheses of this study. The purpose of this research was to examine the reasons why freshmen who entered Michigan State Univer- sity during Fall Term 1981 did not return to resume their degree work in Fall 1982. The study was conducted using a longitudinal design across 12 months and collected two categories Of information. Data were obtained before the student's initial enrollment at Michigan State University (precollege characteristics) and again while the student was enrolled during the first year (college experiences). The population for this study was the entering-freshman class for Fall 1981. Five characteristics were common among this group and were used to reduce the chances Of error due to differences among the students. Each student: 1. Entered college for the first time 2. Was not enrolled in a two-year technical program 3. Had completed the American College Test (ACT) before enrollment 4. Lived in an MSU undergraduate residence hall 5. Was enrolled for Winter Term 1982 86 87 A computer program was developed to sample 3,400 students on a random basis from the total freshman population of 6,288. The 3,400 students met the above five conditions and became the sample studied in the research. Following an extensive literature review to study both previous research designs and variables that were associated with attrition, two instruments were chosen to fulfill the data requirements Of the longi- tudinal design. The American College Test, Student Profile Section (SPS) was used to provide information about the student's precollege characteristics. These data had been collected before the student's enrollment and were stored on a data tape maintained by the MSU Registrar's Office. The College Experiences Scale 11, a question- naire administered during the eighth week of Winter Term 1982, was used to collect information about the nature of the student's colle- giate experiences along two dimensions, academic and social. Once the second data collection occurred, both sets of information were held until the tenth day of Fall Term 1982. On that date, students who had originally participated in the study were separated into two groups, those students who re-enrolled for Fall Term 1982 (returning students) and those who did not re-enroll (nonreturning students). Of the 3,400 students sampled, 2,064 returned questionnaires (or 60.7%) that were administered during Winter Term 1982. Approxi- mately 4.9% or 102 of the returned questionnaires were not usable due to incomplete information or because the student did not have SPS information on file. Following Fall Term 1982 enrollment, it was found that 265 students were not registered, and 25 of those 88 students had been academically dismissed. This reduced the number of voluntary nonreturning students to 240, Of whom 134 were women and 106 were men. There were 1,697 students who were classified as return- ing students, with 982 women and 715 men. (See Table 4.1.) Table 4.l.--Sample data. Total sampled: 3,400 freshman students Questionnaires Returned--W 82: 2,064 (60.7%) Unusable: 102 (4.9% of returned) Not enrolled Fall 1982: 269 students Academically dismissed: 25 students Voluntary nonreturns: 240 students (134 women, 106 men) Returning students: 1,697 students (982 women, 715 men) As previously noted, the SPS data were collected before the student's enrollment, and responses were coded on a data tape. The responses to the questionnaires were key-punched onto data-processing cards and then machine read onto another data tape. Both data tapes were merged by computer assistance and then merged a second time with the Student Master Data Tape (SMDT) to determine returning or non- returning classification. The SMDT contains enrollment information, among other items, and is also maintained by the MSU Registrar's Office. Following the merging of the SMDT, a master data file resulted to accommodate computer-assisted statistical analysis. The software package, Statistical Analysis System, was used to compute 89 the data analysis using the subprograms Crosstabs with chi-square, Stepwise Discriminant Analysis, and Discriminant Analysis. Analysis of the Hypotheses Hypothesis 1 This hypothesis, written in the null form, stated that selected variables from the Student Profile Section of the American College Test will not successfully discriminate between freshman- year persisters and nonreturning students. Before computing the stepwise discriminant analysis to isolate those variables that dis- criminated between the two groups, a cross-tabulation on each of the 12 variables with a chi-square test for independence was computed. This analysis was completed for the total group (returning compared with nonreturning) and for female (returning--FRET and nonreturning-- FNON) and male (returning--MRET and nonreturning--MNON) students. Table 4.2 shows that returning and nonreturning students were not independent on two variables: high school Class rank and decidedness Of college major. More returning students than expected and fewer nonreturning students than expected reported being-in the tap quarter of their high school graduating class. The Opposite was true when the two groups were compared on the second quarter of their class. None of the students in this sample reported being in the third or fourth quarter Of their high school graduating class. 90 Table 4.2.--Chi-square of significant SPS variables: total group. ITEM: High School Class Rank Frequency . . Top Second E t d . Rocpggrgent Niss1ng Quarter Quarter 141 97 Nonreturns 2 158 80 240 59.24 40.76 1,123 548 Returns 26 1,106 565 1,697 67.20 32.80 x2 = 5.87 p = .025 df = 1 ITEM: Decidedness of College Major Frequency . Expected Missing 53:: F§;:;y g:;z Row Percent 104 92 41 Nonreturns 3 82 105 50 240 43.88 38.82 17.30 559 757 362 Returns 19 581 744 353 1,697 33.31 45.11 21.58 x2 = 10.39 p = .01 - df = 2 More returning students than expected and fewer nonreturning students than expected indicated that they were "fairly" or "very sure" about their intended major in college. The trend reversed itself with more nonreturning students than expected and fewer 91 returning students than expected stating that they were "not sure" of their intended college major. Differences by sex began to appear when reviewing female students separately from male students. Table 4.3 shows that return- ing females (FRET) were not independent from nonreturning females (FNON) on two variables: high school GPA and high school class rank. Although the results were somewhat mixed on the variable of high school grade point average, it was generally found that more returning women and fewer nonreturning women than expected had grade point averages in the range of 2.33 to 4.00. The trend requires care- ful review as more nonreturning women than expected compared to return- ing women had grade point averages in the 3.67 to 3.75 categories. The variable of high school class rank demonstrated a more clearly consistent result. Similar to the total group, more return- ing women and fewer nonreturning women than expected reported being in the tap quarter of their high school graduating classes. Again, the trend was reversed when looking at the results for the second quarter. None of the students reported membership in the third or fourth quarter of their graduating classes. The male returning and nonreturning students also showed that the groups were not independent, but on only one item. For the men, Table 4.4 shows the results for the variable, decidedness of College major. Interestingly, none Of the male students indicated that they were "very sure" about their chosen major in college. The results did show, however, that more returning students and fewer nonreturning 92 P u we Po. u a mm.m u Nx mw.Nm m_.Nm Nam m.omm m.emo NF Hume N—m New om.¢¢ em.mm «mp m.m¢ m.Nm _ zozm mm cu ucmuema 2oz emuemso smegmao mcpmmp . .z emaumaxu uncomm ace zucmacmem xcwm mme_u poogum saw: "ZNPH . o_ u we mo. u a No.m_ u Nx D 93 mA 3.3 mmN 5.3 mmép oNA: e; mé N. Nmm m.nmp N.Nep o.mp F.wmp N.pN o.om_ m.NmF N.pm ¢.NF m.am ¢.¢ we Fume amp Ne— NP emp NN mmp mmp no mp ow N mp NP .q oN N .v— ON 0 P e N. «mp N.ON w.wp m.N m.ON m.N p.mp N.oN F.N_ o.P N.m c. op zoze mp .vp m mN N NF mN m p m m acmuema 3cm oo.¢ mn.m No.m om.m mm.m mN.m oo.m mN.N mm.N om.N mm.N mcwmmwz cavemaxu xocoscwem <58 Foogum saw: "smee .mwpmsme "mmpne_em> mam pcmuwepcmwm eo meozcmumsu-u.m.¢ opnm» 93 male students reported being “fairly sure" Of their chosen program of study in college. The opposite pattern was demonstrated for the category, "not sure." Table 4.4.--Chi-square Of significant SPS variables: males. ITEM: Decidedness of College Major Frequency . Expected Missing QOt ngrly Row Percent ure ure 74 32 MNON O 60.4 45.5 106 69.81 30.19 389 317 MRET 9 402.6 303.5 715 55.02 44.84 x2 = 8.12 p = .005 df = 1 Following the chi-square analysis, a stepwise discriminant analysis was computed using the 12 variables selected from the Student Profile Section of the American College Test. The stepwise procedure attempts to determine the variables that successfully discriminate among the 12 considered and the degree Of their discriminating power. The amount of discriminating power can be examined through the use of two statistics computed as a part of the stepwise procedure: the Partial R**2 and Wilks' Lambda. The Partial R**2 computes the amount of variance that is explained by a discriminating variable(s) among the scores of the returning and nonreturning students around their group centroid (grand mean for all the variables of a particular 94 group). Wilks' Lambda, on the other hand, is a measure of residual discrimination: By "residual discrimination," we mean the ability Of the vari- ables to discriminate among the groups beyond the information that has been extracted by the previously computed functions. . . Values of Lambda which are near zero denote high dis- crimination (i.e., the group centroids are greatly separated and very distinct relative to the amount of dispersion with groupsL.(Klecka,1980, pp. 38-39) A third statistic, also computed in the stepwise procedure, is used to measure the usefulness of a discriminating variable or set of variables. The Canonical Correlation produces this third method through a coefficient that "summarizes the degree of relatedness between groups and the discriminant function" (or set of related dis- criminating variables) (Klecka, 1980, p. 36). This statistic, simi- lar to the Pearson product-moment correlation coefficient, is a measure Of association, with values near zero representing little or no association to a maximum Of plus or minus one, showing a strong degree of relatedness. Table 4.5 presents the summary of the first stepwise analysis computed for all returning and nonreturning students. Using the criterion to enter the model of an F-test from an analysis of covari- ance with an alpha level of .10, three variables were entered: 1. NCERTAIN: The degree of decidedness a student has about his/her chosen college major 2. ACTC: American College Test composite score 3. N059: Family income 95 moo. mpoo. ooo. oomo. opw.m NNoo. omoz m moo. «Noo. moo. NoNo. omm.m pmoo. oho< N woo. vooo. moo. eooo. pou.o omoo. zH< A need .mxpmz A poem u pmpagmm mFamPem> um .asogm Pogo» as» Low mmpnmwem> mam eo aeosssm :owuumpmm mmwzamumuu.m.e mpnmh 96 The results showed that the three variables did discriminate between the two groups (returning versus nonreturning students) with "statistical significance." However, the values of the Partial R**2, Wilks' Lambda, and the Average Squared Canonical Correlation demon- strated that the function had very little substantive utility. Only .0022 of the variance was accounted for after the third variable, N059, was entered. Further, the Wilks' Lambda showed that the cen- troids of both groups were almost identical; that is, no group differ- ences could be found. Finally, the canonical correlation suggested a very low degree of association among the variables in the function and the two groups. Tables 4.6 and 4.7 present similar results for the stepwise process when the groups are controlled for sex. The variables NCERTAIN, NESTGPA, and EXPEXT were entered into the stepwise model as a function that discriminated between return- ing and nonreturning female students with statistical significance: 1. NCERTAIN: The degree Of decidedness a student has about his/her chosen college major 2. NESTGPA: Estimated grade point average, first year at college 3. EXPEXT: The amount of a student's expected involvement in college extracurricular activities Similarly, the variables N059, ACTC, and EXTRA were isolated by the model as constructing a function that distinguished between returning and nonreturning male students: 97 mpo. oNoo. com. Nemo. oom.N smoo. (Nexm m mFo. omoo. mom. Novo. NFm.m Nmoo. oeo< N oeo. mmoo. mom. mmoo. own.N mopo. mmoz F coeompmceoo pmuwcocmo avasmo muasmo u uepmwoeum Neam umempcu amum umemacm mmmcm>< A noga .mxpwz A poem d Pawnee; mpneeeo> .mmpes com mmpawwem> mam eo secessm :owuompmm mmwzaoumuu.u.¢ «Fame Fpo. oN—o. mom. Nome. mo_.m Nmoo. pxuaxm m woo. Nmpo. Nmm. «moo. onm.m _¢oo. < A seem .mx_e3 A seen u Paragon mpnmeem> .mmFmEme gee mmpnmwgm> mam mo memessm :oepum—mm mmwzamumuu.o.e epoch 98 1. N059: Family income 2. ACTC: American College Test composite score 3. EXTRA: The amount of a student's high school extra- curricular involvement Again, however, the substantive significance of both functions was very low when considering the amounts of explained variance and residual discrimination. Less than four-hundredths of 1% of the variance was explained by the functions for either group, and the measure of residual discrimination showed the group centroids were, again, practically identical. Therefore, despite the suggested differences from the chi- square analysis and the direction of those differences, the null hypothesis was accepted. Selected variables from the Student Profile Section of the American College Test did not successfully discrimi- nate between freshman-year persisters and nonreturning students. Hypothesis 2 The second hypothesis, written in the null form, stated that the precollege characteristics Of family income, high school academic performance, and educational aspirations will not be found to have the best discriminating power between freshman-year persisters and nonreturning students. The category of educational aspirations was measured by three variables: (1) decidedness of college major-- NCERTAIN, (2) expected level Of college performance during the first year of college--NESTGPA, and (3) highest college degree expected-- NDEGASP. 99 Referring to Table 4.5, the variables of family income (N059) and decidedness Of college major (NCERTAIN) were among the variables with the best discriminating power. Although the researcher was expecting to find either high school class rank or high school grade point average as direct measures Of high school academic perform- ance, the_variab1e measuring achievement, or the American College Test composite score (ACTC), was entered into the model. The variable of ACTC could be viewed as an indirect measure of high school academic performance. Table 4.5 displays the variables of NCERTAIN, ACTC, and N059 (the ordering was based on the magnitude of the F-statistic) as having the most power to separate the groups Of returning and non- returning students. However, as previously noted, while the amount of separation was statistically significant (due largely to the number of degrees of freedom--1,l721), the substantive discrimination was very low. The results of the Partial R**2, Wilks' Lambda, and the Average Squared Canonical Correlation withdrew any confidence that could be placed in the predictive ability of these variables. Tables 4.6 and 4.7 demonstrated differences by sex. The summary for females isolated the variables of NCERTAIN, NESTGPA, and EXPEXT, in that order, as having the most power among all 12 SP5 variables considered in the analysis to discriminate between returning and nonreturning female students. According to Hypothesis 2, NCERTAIN and NESTGPA, both measures of educational aspirations, were expected to be among the discriminating variables. This function came the closest to the alternative Of Hypothesis 2. However, EXPEXT, 100 the expected extracurricular involvements in college, was also sig- nificant with an F-value of 3.109. The summary of male students in Table 4.7 isolated three dif- ferent variables when compared to the female students. The items of family income (N059), American College Test composite score (ACTC), and the number of high school extracurricular involvements (EXTRA) were shown to manifest the most discriminating power for this group. Only one variable was expected under the second hypothesis, family income. Interestingly, family income demonstrated the most discrimi- nating power, accounting for more Of the explained variance (Partial R**2) than the other variables of ACTC and EXTRA. Further, the fact that different items were selected for men than for women leveled support for Hypothesis 8, which was premised on the notion that dif- ferences would be found when variables were controlled for sex. When examined for substantive discrimination as opposed to statistical significance, again the functions for men and women did not present any confidence in the results. As noted during the review of Hypothesis 1, less than four-tenths of 1% of the variance (Partial R**2) was explained for either group, and the measure Of residual discrimination (Wilks' Lambda) showed the group centroids as almost identical. Therefore, in view of the results found for the total group, as well as for female and male students, the null hypothesis was accepted. The variables of family income, high school academic per- formance, and educational aspirations, despite some finding of statistical significance for NCERTAIN, NESTGPA, family income, and 101 the indirect measure of high school performance--ACTC, did not pro- duce the best discriminating power among the 12 SPS variables con- sidered. The researcher defined “best discriminating power“ as being supported not only by statistical significance but, more important, substantive significance. Hypothesis 3 This hypothesis, when written in the null form, stated that the variables measured by the College Experiences Scale II (CES II) to demonstrate social and academic integration will not successfully discriminate between freshman returning and nonreturning students. As with Hypothesis 1, before computing the stepwise discriminant analysis, a cross-tabulation of the variables with a chi-square test of independence between the two groups was computed. Again, the analysis was run for the total group and then separately computed for sex differences. Four variables were found to be statistically significant for the total group: the items of school activities, family inter- actions, intellectual development, and current grade point average. Because of the number of these variabies, separate tables are pre- sented for each one. Table 4.8 shows the results of the variable, school activi- ties. With respect to participating in four to six collegiate activities, more returning and fewer nonreturning students than would be expected reported this level of involvement. Additionally, fewer returning and more nonreturning students than expected reported being m 0 mo mNo. u a Noo.op u Nx 102 mm. om.p om.N pm.“ om.¢— mm.mN co.NN mm.o— m~.o em.N noo.~ p.o —N o.N¢ o.m~_ m.omN mom m.mmm m.moN m.mmp ~.o¢ P mcesumm o NN we ¢Np neN oov com NNN mop mo Nc. mo. mo.N mm.m N¢.op mo.pN mm.m~ mN.—N mN.Fp N¢.m . ovN m. o.m ~.o e.op N.mm mm m.¢m N.o¢ N.pN m.o o mcezpmecoz — N m o mN Nm om Fm NN mp m m N o m e m N p o «cmuema 3cm mcpmmpz umuumaxm cH umumgeuepema mmeuw>mpu< eo emaszz . . mocmscmgu .aaoem quou we» so» .mmep_>epoo Poogum cw coepmaeueuemq .mpampem> on» co mu_:mwe mgmzcmuwco--.o.¢ mpnmp 103 involved in zero to three and seven to nine school activities. With 10 degrees of freedom, this result was significant at the .025 level. Table 4.9 presents the results of the two groups on the vari- able of family interactions. This variable was composed of eight types of interactions the student could have with his/her family while in college. The scale measured the number of family interactions from none to as many as eight. The results showed that more non- returning and fewer returning students than expected reported one to two and seven to eight interactions with their families. This trend was reversed when examining the range of three to four interactions as more returning students than expected compared with nonreturning students had this degree of involvement with their families. Table 4.10 reports the results for the two groups on the variable, intellectual development. This scale was composed of seven items corresponding to the student's satisfaction with the influence that his/her academic experience had on his/her (perceived) intellec- tual development. The scale measured the number of items (0-7) that were given a positive response. The figures indicated that return- ing students were generally more satisfied with the effect of their academic experience on their intellectual development than were nonreturning students. More returning students than expected reported a positive response to five to seven of the items, whereas more non- returning students than expected reported a positive response to only zero to three of the items. The results of the group's response on the variable, current grade point average, are reported in Table 4.11. It is important to 104 o u wu mNo. u a moo.op u Nx mm.N v.Np Fn.—N m¢.MN vu.o~ o¢.mp mo.m N¢._ No. noo.p N.mm F.F~N o.mom P.¢mm mom m.NNN o.oo ¢.mN o.m PF mcgzumm ow ooN mom mom o—m NNN mo «N m —w.n om.mp om.NN po.mN mm.NF m.—— mm.“ mo.N Nw. ovm w n m.mN P.Nm m.mm o.m¢ m.~m o.vp ©.m o.~ p mcgzpmgcoz up Nm om mm om mm mp m P m N o m e m N _ o 28.55 as. mcowaomemucn xepsmd eo amass: mcwmmwz xwmwwwmmw .qaoem Papa» on» Lee .mcowpumempcw zFeEme eo Longs: .mpameee> as» so mopsmme menacmuwco--.m.¢ m_nmh 105 n u on Fooo. u a mmm.~¢ u Nx No.o om.oN No.NN oo.o~ po.m_ m~.o Po.o mo.N umo.~ m.omp o.on m.oom m.PuN o.p¢N F.Nop o.mNp p.om m mcezuwm o¢~ «mm mum PNN emN mmp Npp me e¢.m om.o mo.op Nm.op hm.n~ mm.N~ mN.¢F oN.o oeN N.mp ¢.o¢ N.pm m.mm _.¢m m.NN p.mp m.n F magnumecoz mp pN me on Ne on em mp N o m c m N p o ucmuema 3oz mcwmmvz cmuuoaxm mmcoamom m>wppmoa a cm>wo msmuH eo eonszz aocmscwee .azocm Pepe“ ms» gee .ocmsao_m>mu pmzuuwFPmpcw .mpnmegm> on» so mapsmme mcmscmuwzouu.op.e opnmp 106 wt pooo. n a Nwo.wm u Nx om.NF NB.¢N oN.om om.op om.o~ om.N moo.~ ¢.mop n.oom o.No¢ w.pmm N.no~ o.Pm om mcszpwm moN N—e mom mNm mup me o¢.m oo.w— MN.¢N Po.NN om.PN mn.m NNN o.oN m.¢m p.5o N.m¢ o.oN m mp megapmecoz Np Fe mm No me mp o.eum.m c.m:o.m o.Num.N ¢.Nuo.N m.—nm.~ ¢.pno. m ucwugmmazom :Pmmwz umuuw xm xeommumo were; oumeo\mu:mv:um mo Longsz zucmzcmee .gzoem peace one go» .ommem>m “smog mcmem “cmeezu .mpnmwem> any so mp_=mmg memscmuwgouu.PF.¢ mpnmp 107 note that these scores werea student's self-report of what they believed was their current grade point average. Yet, it has been found in related studies, particularly ones in which students reported their high school grades, that approximately 78.0% of the students reported their GPA's accurately, and over 97.0% were within one letter grade of what school officials reported (ACT Assessment Program, 1973, p. 308). The results showed that there was a consistent trend for both returning and nonreturning students. More returning students and fewer nonreturning students than expected had grade point averages in the 2.50-4.00 range. Conversely, more nonreturning and fewer returning students than expected had grade point averages in the .9-2.4O range. The pattern was clear: Nonreturning students tended to have grades in the lower range of the scale, whereas returning students tended to be more likely found in the 2.50-4.00 range. As will be seen later, this same profile was repeated when the samples were controlled for sex. As in Hypothesis 1, the results began to show differences when the groups were controlled for sex. (However, two variables, intellectual development and current grade point average, were sig- nificant for both groups.) The chi-square analysis for female stu- dents demonstrated that the returning and nonreturning women were dependent on the variables of school activities, family interactions, intellectual development, and current grade point average. The male returning and nonreturning groups were dependent on faculty concern for student development and teaching, intellectual development, 108 and current grade point average. Table 4.12 presents these results for female students, and Table 4.13 demonstrates the results for the male students. Similar to the total group, the results for the variable, school activities, showed more returning and fewer nonreturning female students than would be expected in the range of three to eight activi- ties. The pattern changed with respect to zero to two activities, with more nonreturning female students reporting this amount of par- ticipation in collegiate extracurricular activities. These results clearly suggested that returning female students are characteristically more involved in extracurricular functions than are their nonreturning counterparts. The variable, family interactions, also showed results similar to the total group. More nonreturning and fewer returning students than expected reported zero to two and seven to eight interactions with their families. However, more returning students than expected reported a range of three to five interactions. Given this pattern and its similarity to the total group, the results suggested that there may be an optimal range of interaction frequency with one's family and that this is positively associated with persistence. Continuing with the similarity to the total group, female returning students consistently reported more perceived satisfaction with the influence of their academic experience on their intellectual development than nonreturning students. More returning students than expected reported a positive response to five to seven of the items, 109 N u NN NNNN. u N NNN.NN Nx NN.N NN.NN NN.NN NN.NN NN.NN NN.NN NN.N NN. NN. NNN N.NN N.NNN N.NNN N.NNN N.NNN N.NNN NN N. N N. N N NNNN NN NNN NNN NNN NNN NNN NN N N NN. N NN. NN NN.NN NN.NN NN. NN NN.N NN.N NN. N NN. NNN N. N N. NN N.NN N.NN N. NN N.NN N.N N. N N. N zNza NN NN NN NN NN NN N N N N N N N N N N N N NNNNNNN NON NcowpumcmucN ANNENN No Loneaz chNNNz awmwmwmuw NeoNuONNmNNN ANNENN NZNNN N u NN NNNN. u N NNN.NN Nx NN. NN. N NN.N NN.N NN.NN NN.NN NN.NN NN.NN NN.N NN. N NNN N. N N. N N.NN N.NN NNN N.NNN N.NNN NNN N.NN N. NN N NNNN N NN NN NN NNN NNN NNN NNN NN NN NN. NN. NN. N NN. N NN.N NN.NN NN.NN NN.NN NN.NN NN.N NNN N. N. N N. N N. N N.NN N.NN N.NN N.NN N.NN N.N N zoze N N N N NN NN NN NN NN N N N N N N N N N N N NNNNNNN NON NN NNNNNNNNNNNN NNNNN>NNNN No LNNsNz NNNNNNz quwmmmuw mmmuw>wpu< poosum Nimhm .Nucmcaum mNNEmm Noe NNNNNNNN> NH mmo pcmuNeszNN mo memzamuwsouu.NN.N oNNNN 110 NN NNNN. u N NNN.NN u Nx NN.NN NN.NN NN.NN NN.NN NN.NN NN.N NNN N.NNN N.NNN N.NNN NNN N.N—N N.NN NN NNNN NNN mNN NNN NNN NoN Nm NN.N NN.NN NN.NN NN.NN NN.NN NN.N NNN N.NN N.NN N.NN N.NN N.NN N.N N ZNZN N NN mm NN NN w o.¢-m.m N.N-o.m N.N-N.N N.N-o.N N.N-N.N N.N-N. N pcmuemmazom NNNNNz umpum xm NeonNNo choN NNNNNNNNNNNNNN No emaszz Nucmacmcu mNmem>< ucNoN mumeu acmeeao ":mNN N n ma Nooo. u N NNN.NN u Nx NN.N NN.NN NN.NN NN.NN NN.NN NN.NN NN.N NN.N NNN N.NN N.NNN N.NoN N.NNN N.NNN N.NoN N.NN N.NN N NNNN ow NoN oNN NNN mNN NoN mm oN NN.N NN.NN NN.NN NN.NN NN.NN NN.NN NN.NN NN.N NNN N.NN N.NN N.NN N.NN N.NN N.NN N.N N.N o ZOZN oN NN NN NN NN NN NN N N m N N m N N o N NcmocmNNzom :NNNNz umuom xu mNcoNNmm m>NNNNoN N cw>No NEmNN No Loganz Nocmzcoem acmEQoPm>mo FoswumwpwwcH HEP: .NNNNNNNNN--.NN.N NNNNN 111 whereas more nonreturning students than expected reported a positive response to only zero to four of the items. Finally, the variable of current grade point average showed that more returning female students and fewer nonreturning female students than expected had grade point averages in the range of 3.00-4.00. This pattern was reversed as more nonreturning women and fewer returning women than expected had current grades in the range of .90-2.40. Male students were similar to the female students in that the chi-square results were significant for the variables, intellectual development and current grade point average. However, male students were also found to be dependent on a new variable, faculty concern for student development and teaching. Table 4.13 presents the results for male students. The variable of intellectual development did not present as consistent a pattern for male students as it did for female students. The data indicated that while more returning men than would have been expected reported a positive response to six to seven of the items, more nonreturning men than expected gave a positive response to five items, with the pattern again reversihg on four items and making a final change for zero to three items in the scale. Generally, how- ever, a significantly larger percentage of the returning males compared to the nonreturning males perceived more satisfaction with the effect of their academic experience on their intellectual development. A new variable isolated when controlling for sex was faculty concern for student development and teaching. This was a five-item 112 n ma NNNo. u N NNN.NN u Nx NN.NN NN.NN NN.NN NN.NN NN.N NN.N NNN N.NoN N.NNN N.NoN NNN N.NN N.NN N Nmmz NNN NNN NNN NNN No NN NN.N NN.NN NN.NN NN.NN NN.NN NN.N NoN N.NN N.NN N.NN N.NN N.NN N.N N 202: N NN NN om NN N N N m N N o N pcmoemaazom NNNNNz cmpum xm mNcoNNmm m>NNNNoN N cm>NN NENNN No emaeaz Nocmzcmeu NchuNwN use acmsNon>mo ucmuzum Noe Newucoo NNNNNNN mszN N u NN NNNN. u N NNN.NN u Nx NN.N NN.NN NN.NN NN.NN NN.N NN.N NN.N NNN N.NN N.NNN N.NNN N.NoN N.NN N.NN N.NN N Nmmz co NNN NNN NoN NN NN NN NN.N NN.N NN.NN NN.NN NN.NN NN.NN NN.N NoN N.N N.NN N.NN N.NN N.N N.N N.N N zozz m N NN ON NN mN N N o N N N o N pcmogmNNzom :NNNNz umuum xm mmcoNNmm m>NuNNoN N cm>No NENNN No gmaezz Nucmscmcu Newsaon>mo NNNNQNNNNNNN HZNNN .Nucmuzum NNNE NoN NmNNNNNN> NN mmo NNNUNNNNNNN No memzcmano--.mN.N «NNNN 113 u NN NNNN. u N NNN.NN u Nx NN.NN NN NN NN.NN NN.NN NN.N NN.N NNN N.NN N.NNN N.NNN N.NNN N.NN N.NN N NNN: NN NNN NNN NNN NN NN NN.N NN.NN NN.NN NN.NN NN.NN NN.N NNN N.NN N.NN N.NN N.NN N.NN N.N N Zsz N NN NN NN NN N N.N-N.N N.N-N.N N.N-N.N N.N-N.N N.N-N.N NN.N-NN. N NNNNNNNNNNN :NNNNz umuum xu NeonpNo choN muNNNNNucmuapm No gossaz NucmacmeN NNNcm>< NNNoN NNNLN acmeezu NZNNH .NNNNNNNNN--.NN.N NNNNN 114 scale measuring the number of positive responses to the nature and quality of faculty contact. In Table 4.13, the data demonstrated that more nonreturning students and fewer returning students than expected gave positive responses to zero to three items. More returning stu- dents and fewer nonreturning students than expected indicated a posi- tive response to four to five of the items in the scale. These results suggested that a larger percentage of returning male students than would be expected perceived a larger amount of faculty concern for student development and teaching when compared to nonreturning men. Again, the variable of current grade point average showed more returning men and fewer nonreturning men than expected reported grades in the range of 2.5-4.0. This range was .5 broader than the table for female students. However, like the female students, more nonreturning men and fewer returning men than expected indicated current grades in the range of .9-2.40. Thus, the results were con- sistent across both male and female students on this variable, illus- trating that returning students are more likely to have higher grade point averages than nonreturning students. With the completion of the chi-square analysis, a stepwise discriminant analysis was initiated as a more powerful test of dis- criminating ability among the variables. Again, the criterion to enter the model was an F-test from an analysis Of covariance with a significance level of .10. Table 4.14 presents the results of the procedure considering the 13 variables from the CES II. Five variables were entered: 115 oNo. Nooo. NmN. mNoo. meo.m NNoo. uzthmmN m Nmo. Nooo. NoN. Nomo. mNN.e oNoo. NzNomoNzH N .owumuumsemum.s News «may... A NN.; saws...” emu“; .mmwuw as .NaoNN NNuoN on“ NoN NwNNNNNN> NN muu No NNNEENN coNuumNmN mNN3Nm9m11.NN.N oNNNN 116 l. INTDEV: Intellectual development SCHACT: Number of collegiate extracurricular activities NCURGPA: Current grade point average FACINT: Number of faculty interactions 01-wa PERSTIME: Number of activities devoted to personal time The data revealed that the five variables formed a discrimi- nant function that was statistically significant at the .10 level. A more conservative function would be composed of the first four variables statistically significant at the .03 level. Again, however, statistical and substantive significance were widely separated. The amount of variance that was explained by the Partial R**2 was extremely low, the amount of residual discrimination was very high with a .96 lambda, and, finally, the average squared canonical correlation showed an equally low degree of association between the groups and the vari- ables. A more dramatic illustration Of the lack of substantive dis- criminating power of the function can be seen through a discriminant- classification analysis: As a check of the adequacy of our discriminant functions, we can classify the original set of cases to see how many are correctly classified by the variables being used. The procedure for classification involves the use of a separate linear combi- nation Of the discriminating variables for each group. These produce a probability of membership in the respective group, and the case is assigned to the group with the highest probability. (Nie et al., 1975, p. 436) Table 4.15 presents the discriminant-classification table derived from the above discriminant function. As can be seen, the 117 classification table provides conclusive evidence that the discrimi- nant function does not have the ability to separate between returning and nonreturning students. All of the nonreturning students were incorrectly classified as returning. However, given that all of the returning students were correctly classified, the function may be more indicative of a measure that describes returning students. That is, it lacks variables that describe nonreturning students but not students who persist. Table 4.15.--Discriminant analysis classification summary: CES 11 variables: total group. Number of Observations and Percentages From Group Classified Into Group Nonreturn Return Total 0 226 226 "”59“” 0.0 100.0 100.0 0 1,655 1,655 Rem" 0.0 100.0 100.0 Total 0 1,881 1,881 Percent 0.0 100 O 100 O Note: Due to missing data, 56 of the 1,937 observations were not included. As with the chi-square analysis, a stepwise discriminant pro- cedure was computed separately for male and female students. Tables 4.16 and 4.17 demonstrate the results of these calculations. 118 emo. Nooo. NNN. NmNo. NNN.m NNoo. NomoNzN m oNo. Nooo. oNN. oNoo. Noo.N NNNo. ozooo< A NoNN .NNNNz A NoNN N NNNNNNN NNNNNNN> .NoNNE No» NmNNNNNN> NN mmo No NewEENN acchENNONNu mNNznmumau.NN.N mNNNN mNo. Nooo. NNN. NNNo. NNN.N Nmoo. NzNzNNNzN N :oNNNNmNNoo NNuNNoNNo Noose; NuNeNN N ONNNNNNNN NNNN cmemucm No uwNNNNm mNNeo>< A NNNN .NxNNz A NoNN N NNNNNNN mNNNNNN> um .NmNNENN NNN NmNNNNNN> NN mmo No NNNEENN NNNNNENNONNN mmNzaoum--.NN.N mNNNN 119 The variables INTDEV, SCHACT, COMMACT, and FAMINT were entered into the stepwise model, creating a function that statistically sep- arated returning and nonreturning female students: 1. INTDEV: Intellectual development SCHACT: Number of collegiate extracurricular activities COMMACT: Number of community activities in college #MN FAMINT: Number of family interactions While the variables of intellectual development and school activities were also included in the function for male students, three different variables (from the function for female students) were entered into the discriminant. The three new variables were NCURGPA, FACCONC, and PERSTIME: l. NCURGPA: Current grade point average 2. FACCONC: Faculty concern for student development and teaching 3. PERSTIME: Activities devoted to personal time However, similar to Hypothesis 1, both functions were very low with respect to the substantive significance they provided. The Partial R**2 in both Tables 4.16 and 4.17 indicated that an extremely small portion of the variance in each'group was explained. Further, the lambda showed that a very large amount of residual discrimination was left unexplained by these functions, and the average squared canonical correlation demonstrated that a minimal relationship existed between the series of variables and the returning/nonreturning male and female students. 120 Despite the suggested differences illustrated by the chi- square analysis, variables measured by the College Experiences Scale II did not sufficiently discriminate between returning and nonreturning students. The magnitude of the differences found in the chi-square analysis was not great enough to result in a substantively signifi- cant discriminant function for any of the three groups (total group, males, and females). The null hypothesis was accepted. Hypothesis 4 The fourth hypothesis, written in the null form, stated that the variables measured by the College Experiences Scale 11 (CES II) will not be found to have better discriminating power between freshman- year persisters and students who do not return for their second year than the variables selected from the American College Test Student Profile Section (SPS). This hypothesis was written to test the premise that precollege characteristics would not have as much influ- ence on persistence as would the actual experiences of the student once she/he was on the campus. Tables 4.5, 4.6, 4.7, 4.14, 4.16, and 4.17 need to be re-examined to determine the results of this hypothesis. Further, the statistics that were used to distinguish the "best discriminating" functions were the Partial R**2 (to measure the amount of variance explained by the set of variables), the Wilks' Lambda (which indicates the degree of separation between group cen- troids), and the average squared canonical correlation (which illus- trates the degree of association between the function and the groups of returning and nonreturning students). 121 Tables 4.5 and 4.14 demonstrated the comparison for the total group. Table 4.5 summarized the function of precollege characteris- tics for the total group, whereas Table 4.14 provided the same data but for the college experiences or integration variables. (See Table 4.18.) Table 4.18.--Comparison of discriminant functions of SPS and CES II variables for the total group. Partial Wilks' Average Squared R**2 Lambda Canonical Correlation Table 4.5 (SPS variables) .0022 .990 .009 Table 4.14 (CES II variables) .0017 .959 .040 A similar comparison can be made for female students (Tables 4.6 and 4.16) and male students (Tables 4.7 and 4.17). These compari- sons are illustrated in Table 4.19. The data revealed differences so small that they prohibited any meaningful analysis. Even when the functions were controlled for sex, minimal differences resulted. A5 will be remembered with the data presentations for Hypotheses l and 3, although the functions were composed of statistically significant variables, the amount of separa- tion between returning and nonreturning students was negligible. In summary, no difference was found in the discriminating ability of either the precollege characteristics (SPS variables) or the college 122 experience/integration variables (CES II variables), and the null hypothesis was accepted. Table 4.19.--Comparisons of discriminant functions of SPS and CES II variables for females and males. Partial Wilks' Average Squared R**2 Lambda Canonical Correlation Table 4.6 (SPS variables: females) '0032 '989 ‘01] Table 4.16 (CES II variables: females) 003] ’957 043 Table 4.7 (SPS variables: males) 0037 '980 019 Table 4.17 (CES II 0044 .946 054 variables: males) Hypothesis 5 Written in the null form, this hypothesis stated that of the variables measured by the CES II, the scales of school activities (collegiate extracurricular activities), peer interactions, and faculty interactions will not be found as the-best discriminating variables between returning and nonreturning students along the social-integration dimension. The CES II measures 13 variables, of which nine variables cor- respond to the social-integration dimension. These variables are: academic activities, peer interactions, faculty interactions, school extracurricular activities, family interactions, growth and development, 123 community involvements, personal time, and work. While faculty interactions seemed more logically connected to the academic dimen- sion, it measured primarily nonclassroom faculty contacts. Tinto's (1975) model stipulated that while out-of-class faculty interactions tend to influence academic integration, they are more influential in the students' social as opposed to academic adjustment. Tables 4.14 (stepwise selection summary of CES II variables for the total group), 4.16 (for females), and 4.17 (for males) pro- vided the data to test this hypothesis. Table 4.14 showed that two of the hypothesized variables, school activities and faculty interac- tions, entered the discriminant function and provided more discrimi- nating power than any other variable along the social-integration dimension. In fact, only one other variable from this dimension met the criteria to enter the function: activity devoted to personal time. The variable of peer interactions did not meet the selection criteria and consequently was not an element of the discriminant function for the total group. When controlling for sex, Table 4.16 showed that the variable of school activities was the only scale of the three that were hypothe- sized that discriminated with more power than the other variables selected along the social dimension. As can be seen, in order, community involvements and family interactions were elements of the function for female students, whereas faculty and peer interactions did not meet the selection criteria. The data for the male students were found in Table 4.17. The results showed that only one of the variables hypothesized to be among 124 the best discriminating variables along the social dimension was found. The scale, school activities, was chosen ahead of activities devoted to personal time, the only two variables that met the selec- tion criteria for the discriminant function along the social dimen- sion. However, as indicated by the F-statistic, 3.452 and 3.347, respectively, with significance of .0636 and .0677, the relative contributions of both scales were very close. The support for both the null and alternative hypotheses of Hypothesis 5 was mixed. When examining the total group, the most support for the alternative hypothesis was found. That is, two of the three predicted variables from the social dimension provided more discriminating power than those not hypothesized. Of these two vari- ables, the scale of school activities provided the "best discriminat- ing power" and was also found to be the only hypothesized scale chosen in the discriminant function for female and male students. The third variable hypothesized to be among the "best discriminators" from the social dimension, peer interactions, was not an element of any func- tion, whereas the scale, faculty interactions, was found in the func- tion for the total group. Hypothesis 6 This hypothesis, in the null form, stated that variables measured by the CES II that correspond to the academic dimension will not be better discriminators than those that correspond to the social dimension. Variables measured by the CES II that are found in the academic dimension are the scales: faculty concern for student 125 develOpment and teaching, intellectual development, and current grade point average. Again, Tables 4.14 for total group, 4.16 for females, and 4.17 for males provided the data necessary to test this hypothesis. Similar to Hypothesis 5, the data were mixed for two of the three groups. Table 4.14 for the total group illustrated that of the five elements in the function, the scales of intellectual development and current grade point average held positions one and three, respec- tively. Clearly, the most powerful discriminating variable was intel- lectual development, with an F of 43.078 significant at .0001. Table 4.16 for female students isolated only one of the three vari- ables from the academic dimension, intellectual development. However, as the discriminant function revealed, it was the single most power- ful discriminating variable Of the four scales that met the selection criteria with an F of 22.326 significant at .0001. The most support for the alternative hypothesis--that is, variables from the academic dimension will be "better discriminators" than those from the social dimension between returning and nonreturn- ing students--was found with the male students in Table 4.17. Of the five scales that comprise this function, three variables corresponded to the academic dimension and held positions one, two, and three, respectively. The scales, current grade point average (F = 22.409, prob. > F = .0001), faculty concern for student development and teaching (F = 9.609, prob. > F = .0020), and intellectual develOpment (F = 4.887, prob. > F = .0273), were found to be better discriminating variables than those that corresponded to the social dimensions and were elements of this function. 126 Hypothesis 7 The seventh hypothesis, written in the null form, stated that of all the variables measured in the study by both the SPS (precollege characteristics) and CES II (college experience and integration), the two scales of faculty concern for student development and teaching and faculty interactions will not be the best discriminating vari- ables. TO test this hypothesis, both sets of variables were entered simultaneously into a stepwise discriminant procedure. Three analyses were run so that the total group could also be compared with the resulting function for female and male students. Table 4.20 presents the results for the total group. The results showed that the two most powerful discriminating variables were related to academic integration, intellectual development and current GPA, but were not the two hypothe- sized scales. In fact, only one of the scales, faculty interaction, was an element in the discriminant function, and it provided the least discriminating power of any of the variables in the function. Table 4.21 presents the results for female students. The results showed that, again, intellectual development was the most powerful discriminating variable. The second variable, SCHACT (colle- giate extracurricular activities), did not relate to the academic dimen- sion but more appropriately corresponded to the social-integration dimension. It was not until the fourth position that one of the two hypothesized variables, faculty interactions, was chosen. Therefore, no support was found for the alternate form of Hypothesis 7. 127 NNN. NNNN. NNN. NNNN. NNN.N NNNN. NzNNNN N NNN. NNNN. NNN. NNNN. NNN.N NNNN. NNNz N NNN. NNNN. NNN. NNNN. NNN.N NNNN. NXNNXN N NNN. NNNN. NNN. NNNN. NNN.N NNNN. NzNNNNNN N NNN. NNNN. NNN. NNNN. NNN.N NNNNN zNNNNNNz N NNN. NNNN. NNN. NNNN. NNN.NN NNNN. NNNNNN N NNN. NNNN. NNN. NNNN. NNN.NN NNNN. NNNNNNZ N NNN. NNNN. NNN. NNNN. NNN.NN NNNN. >NNNzN N NNNNNNNNNNN NNNNNNNNN NNNNNN NNNENN N NNNNNNNNN NNNN NNLNNNN NNNN NNNNNNN NNNNN>< A noNN .NxNNz A NoNN N NNNNNNN mNNNNNN> .NsoNN NNNou NNN Low NNNNNNNN> mNm New NH mmo No NNNEENN coNpumNmN mNNszNm--.oN.N NNNNN 128 NNo. Nooo. NNN. NNNo. NNN.N Nmoo. zNuoNzN N NNNNNNNNNNN NNNNNNNNN NNNENN NNNsNN N NNNNNNNNN NNNN NNNNNNN NNNN umemscm mNNNm>< A NoNN .NxNNz A NoNN N NNNNNNN mNNNNNN> .Npcmuzpm mNNEmN NoN NoNaNNNN> mNm NNN NN mmo No NNNEENN :oNNuonN NNNNNNNN--.NN.N NNNNN 129 Similar results were found for male students, as shown in Table 4.22. Similar to the previous two tables, the results demon- strated no support for the alternative form of the seventh hypothesis. While four of the variables in this function were also found in the discriminant function for female students (current grade point average, school activities, intellectual development, and degree of decidedness on college major), differences were found between men and women in both the discriminating power of these scales and the selection of the variables. In view of these results across all three groups, the null form of Hypothesis 7 was accepted: The variables that measured the amount and quality of faculty contact were not found to have the most discriminating power among all the variables measured in this study. Hypothesis 8 The eighth and final hypothesis, written in the null form, stated that those variables found to discriminate between freshman- year persisters and nonreturning freshman students will not be differ- ent when the sample is controlled for sex. The reader will recall that the analyses of the previous seven hypotheses provided separate results for the comparison of the sample based on sex. While the data will not be presented in their entirety a second time, a summary is provided to highlight the findings. Further, a caution is raised in that the differences based on the dis- criminant functions are statistical in nature and, as pointed out 130 Noo. Nooo. NNN. NNNo. NNo.m meoo. zNmoNzN N mNo. Nooo. NNN. NNNo. NNN.N Nmoo. No< A NoNN .NxNNz A NNNN N NNNNNNN mNNNNNN> .Nucmczum NNNE Noe NNNNNNLN> NNN new NN muo No NNNEENN :oNpumNmN NNNszNm--.NN.N «NNNN 131 earlier in this review, lack meaningful substance due to the minimal separation they provide between the groups. Differences based on sex were found in Hypothesis 1 not only with respect to the discriminant functions but also the chi-square analysis. The stepwise procedure selected the variables NCERTAIN, NESTGPA, and EXPEXT for females, while N059 (family income), ACT composite, and EXTRA (high school extracurricular activities) were entered in the discriminant for men. The chi-square found the vari- ables of high school GPA, high school class rank, and EXPEXT signifi- cant for the female students; only one variable, NCERTAIN, was found for the male students. These findings supported the premise that dif- ferences would be found when the sample was controlled for sex. Hypothesis 2 tested for the best discriminating variables of those measured under the category of precollege characteristics. While it was hypothesized that particular variables would be found, for the purposes of this section, it is important to note that the discriminant functions, as noted above, contained different variables on the basis of sex. Hypothesis 3, similar to Hypothesis 1, found differences between the variables that composed the discriminant functions for men and women. While two variables, school activities and intellec- tual development, were components of both discriminant functions, their discriminating power in each group was different. The variables of INTDEV, SCHACT, COMMACT, and FAMINT, in order of discriminating power, resulted in the function for women. The variables NCURGPA, FACCONC, INTDEV, PERSTIME, and SCHACT, also in order of discriminating power, 132 formed the discriminant function for male students. The chi-square results for male and female students were found to be more similar than in Hypothesis 1. For female students, the returning and non- returning groups were found dependent on the variables of SCHACT, FAMINT, INTDEV, and NCURGPA, whereas the men were found dependent on the variables of FACCONC, INTDEV, and NCURGPA. Again, these results supported the importance of testing for differences between sexes. Hypothesis 4 attempted to determine which set of variables, the SPS or CES II, provided more discriminating ability between returning and nonreturning students. As will be remembered, the dif- ferences, though favoring the CES 11 (college integration) variables, were so small that meaningful analysis was not possible. The fifth hypothesis was written to determine the best dis- criminating variables from the CES 11 along the social dimension. Again, while particular variables were hypothesized, for the purpose of this eighth hypothesis, different variables were found when the sample was controlled for sex. The female students' function was composed of SCHACT, COMMACT, and FAMINT, while the male students' function comprised PERSTIME and SCHACT. These results demonstrated that differences were found when the sample was controlled for sex. The sixth hypothesis attempted to determine the discriminat- ing power of social-integration variables compared to academic- integration variables. Again, differences were found when the female discriminant function was compared to the male discriminant function. The function for female students was more mixed with respect to the discriminating power of the academic-integration variables over the 133 social-integration variables. However, the function for male students demonstrated that the academic-integration variables were consistently the more powerful discriminating variables when compared to those that corresponded to the social dimensions. The seventh hypothesis was constructed to determine the most powerful discriminating variables when both the precollege character- istics and college experience/integration variables were entered into a stepwise discriminant procedure. Tables 4.21 and 4.22 showed that the variables selected for the female and male students were different with the exception of four variables that appeared in both functions. Yet, even the four variables that overlapped showed dif- ferent placements in each function with respect to discriminating abilities. Therefore, as the data demonstrated, six out of the previous seven hypotheses revealed different results when the sample was con- trolled for sex. Sufficient support did seem to exist to reject the eighth null hypothesis, although the evidence only provided statis- tical and not substantive significance. Summary The procedure, stepwise discriminant analysis, was employed to test the eight hypotheses in this study. As a preliminary indi- cator of the variables that may be selected by the discriminant pro- cedure, a chi-square analysis was used for two Of the eight hypotheses. In analyzing the results of the hypotheses tested, the variables found in each discriminant function did distinguish between returning 134 and nonreturning students with statistical significance. However, the degree of separation that resulted between the group centroids was so small that the predictive power of the function withdrew any confidence that could be placed in the statistical finding. To illustrate this, a discriminant classification analysis was computed for Hypothesis 3. The classification procedure correctly classified all returning students but incorrectly classified all nonreturning students. Only the chi-square results demonstrated any significant differences that may warrant further pursuit for Hypotheses l and 3. It was found that: 1. Null Hypothesis 1, which was written to determine if a discriminant function could be constructed with precollege character- istics measured by the Student Profile Section of the American College Test, was accepted. However, significant chi-square results were obtained and differences were found in the results based on the inde- pendent variable, sex. 2. Null Hypothesis 2, which was written to determine if family income, high school academic performance, and educational aspirations (degree of decidedness on college major, expected level of college performance, and highest degree expected) would be selected as the most powerful discriminating variables among all the precollege char- acteristics, was only partially rejected. One of the variables, degree Of decidedness on college major, from the category of educa- tional aspirations was the most powerful discriminating variable selected; ACT composite was chosen second. The researcher has decided to accept the ACT composite as an indirect measure of high 135 school academic performance. The other expected variable, family income (N059), was selected as the third most powerfully discriminat- ing variable of all the precollege characteristics. 3. Null Hypothesis 3, which was written to determine if a discriminant function could be constructed with college experience/ integration variables measured by the College Experiences Scale 11, was accepted. However, as with Hypothesis 1, significant chi-square results were obtained and differences were found in the results based on the independent variable, sex. 4. Null Hypothesis 4 attempted to determine which cate- gory of variables, precollege characteristics or college experi- ences, provided the most discrimination between returning and nonreturning students. While the data tended to favor college experi- ences, the differences were so small that the null hypothesis was accepted. 5. Null Hypothesis 5 was written to test for the best dis- criminating power among the college-experience variables and stated that the scales of school activities, peer and faculty interactions would not be selected. However, the variables school activities and faculty interactions were selected, while peer interactions was not chosen. Therefore, this null hypothesis was only partially rejected. 6. Null Hypothesis 6, which was tested to determine if academic-integration variables would be better discriminators than social-integration variables, showed mixed results for the total group and when the sample was controlled for female students. Only 136 the male students showed a consistent pattern isolating academic- integration variables over social-integration variables. 7. Null Hypothesis 7, which was tested to determine if the variables, faculty concern for student development and teaching and faculty interactions, would be the most powerful discriminating variables of all those measured in the study, was accepted. 8. Null Hypothesis 8, which stated that differences would not be found hithe discriminant functions for each hypothesis, was rejected for six of the previous seven hypotheses. Only Hypothesis 4 did not reveal any differences based on the independent variable, sex. Chapter V reports the summary, conclusions, and recommenda- tions for further research from the study. CHAPTER V SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS As higher education enters the decade of the 19805, many significant challenges will demand ongoing attention and effective response. Among some of the more critical issues will be the careful management of a broad range of resources that enable the continued operation of programs and activities. Specific to public higher education, this not only means a careful analysis of general-fund (state tax) revenues, but more important, the size of its student population that provides substantial monetary and nonmonetary support. As a result, attention has turned to how students are recruited with an increased emphasis on retaining them once they are enrolled. Understanding how to enable student persistence to graduation requires knowledge of why students leave a particular institution. Student-attrition studies date back to the 19205 and reflect a number of findings, recommendations, and ways in which to examine and understand the problem. Generally, however, the research has been methodologically uneven, contradictory, and limited in its ability to explain, not merely describe, dropouts from persisters. In part, this has been due to different definitions of the term "dropout," different views of what causes the phenomenon, and an equal number of methods and designs to study the topic. Many of the 137 138 investigations have been conducted at single institutions, presenting an obvious limitation in the ability to generalize the findings. How- ever, it is well known that research that attempts to "cut across" many institutions and provide universal findings tends to seriously mask individual differences that are critically important in respond- ing to the problem. Acting on the fact that institutions are not similar due to specific practices, programs, and procedures, and recognizing the importance that such elements play with respect to understanding attrition, the need to develop "local" information has taken prece- dence over nationally sampled studies. Additionally, it has only been in the last few years that investigators have developed models that seek to explain institutional- and student-interaction effects on dropping out. The purpose of this study was to examine reasons why freshmen who entered Michigan State University during Fall Term 1981 did not return to resume their degree work in Fall 1982. The study was con- ducted within the framework of Tinto's (1975) conceptual model and focused on the problem of the voluntary dropout. This chapter provides a summary of the development of the study, a review of the findings, its conclusions, and recommendations for further research. Summary: Development of the Study The first chapter presented the problem of student attrition and how it related to one institution, Michigan State University. 139 Components of the first chapter included the purpose and importance of the study, definition of terms, the hypotheses, an abbreviated review of the literature that supported the theoretical model, the design of the study, delimitations of the research, and limitations that were not in the researcher's control. Eight hypotheses were generated for this study and are covered in greater detail later in this chapter. A review of the background literature that supported the hypotheses for this investigation was presented in the second chapter. Rates of attrition were examined, which revealed that Michigan State University experiences a slightly higher nonreturn rate than national findings. The contribution of precollege characteristics to the understanding of attrition/persistence was then examined with specific discussion of (1) individual attributes, (2) family background, (3) precollege educational experiences, and (4) collegiate expectations. The chapter then turned its attention to the study of theoretical models that sought to explain the interaction between the student and the college environment. The summary pointed not only to the need for develOping "local understanding" of attrition but also the importance of doing so within the confines of a theoretical reference. A description of how the study was operationalized was pre- sented in the third chapter. Following from previous study, it was decided to continue the examination of precollege characteristics and determine the strength of their influence on freshman drOpout. To obtain this information, the American College Testing Program results for the entering 1981 freshman class were accessed. The Student Profile 140 Section of the ACT provided data in four main categories: individual attributes, family background, educational experiences before college entry, and expectations of the college experience. As will be recalled, this information fulfills the data requirements for the first part of Tinto's (1975) model. To obtain information for the second part of the model, col- lege experiences, a questionnaire was constructed. Containing 11 sub- scales, the instrument gathered information on extracurricular activities, peer interactions, faculty interactions, academic activi- ties, family interactions, growth and development activities, commu- nity activities, personal time, work activities, faculty concern for student development and teaching, an assessment of the student's intellectual development, and the current grade point average. The questionnaire was given to 3,400 students in the freshman class at the approximate midpoint of their first year. Each of the students participating in the sample met five conditions to reasonably insure for homogeneity within the sample. Approximately 60.7% of the questionnaires were returned, with only 4.9% unusable. Both sets of data were held until the tenth day of Fall Term 1982. At that time, the sample was subdivided into two groups; returning and nonreturning students, with those who were academically dismissed deleted from the sample. The data were then prepared for analysis using a cross-tabulation with chi-square and discriminant procedures as outlined in Chapter III. Chapter IV presented the results of the study using both the statistical techniques of discriminant and chi-square. The statistical 141 package employed for this analysis was the Statistical Analysis System, a software package maintained by the MSU data-processing department. Supporting tables for the results were detailed in this chapter and corresponded to the eight hypotheses. The findings for this study are presented in the following section. Findings and Interpretation Within the Framework of Tinto‘s Model The first two hypotheses of this study were concerned with the effect of precollege characteristics on returning and nonreturn- ing students in the study. Hypotheses 3, 4, 5, and 6 focused on the college experiences of the student and how these influences affected the returning and nonreturning students. Hypothesis 7 sought to delineate the importance of precollege variables in comparison with the measures of the student's collegiate experiences. The final Hypothesis (Hypothesis 8) was concerned with differences due to sex and was examined throughout a review of the previous seven hypotheses. Given the parameters of the study and the limitations inherent in its design, the following findings resulted. Null Hypothesis 1 Variables selected from the Student Profile Section (SPS) of the American College Test (used to measure precollege character- istics) will not result in a discriminant function distinguishing freshman-year persisters and students who do not return for their second years. 142 This hypothesis was accepted. Despite the suggested influence of precollege variables (from the chi-square results) and the fact that a statistically significant function could be computed, its prac- tical ability to successfully distinguish between returning and non- returning students was not found. This finding became readily apparent when the function was examined by the small amount of vari- ance that it explained, the high amount of residual discrimination that remained, and the extremely low degree of association that existed between the function and the two groups of students. Yet the data did reveal significant chi-square results: 1. Returning and nonreturning students were found to be dependent on two variables: high school class rank and decidedness of college major. The probability that these results occurred by chance was found to be 2.5 times in 100 and 1 time in 100, respectively. As supported by previous research, proportionately more returning than nonreturning students ranked in the top quarter of their high school class. Further, and showing even greater significance, a larger per- centage of returning students reported being "very" or “fairly sure". of their chosen college major. Tinto's model would suggest that the returning students in this study were predisposed to academic success during their first year as high school class rank is a direct reflection of high school academic performance. Such academic standing facilitates academic integration in the college environment. Further, decidedness of college major is indicative of the student's initial goal commitment to a college degree. As research in Chapter II pointed out, students who are more 143 certain of their goals in college have a greater likelihood of per- sistence than those who are less certain. 2. When the sample was controlled for sex, differences between male and female students were found. For the female student, it was found that a larger percentage of returning than nonreturning women were in the top quarter of their high school class, and the proba- bility that this occurred by chance was 1 in 100. The variable of high school grade point average was also significant but more mixed and did not reveal any consistent pattern. That is, a larger propor- tion of returning than nonreturning female students had GPA's in the categories of 2.50, 2.75, 2.76, 3.25, 3.33, 3.75, and 4.00. Non- returning women demonstrated a larger proportion than their returning counterparts in the categories of 3.00, 3.50, and 3.67. The male students were only dependent on one variable, decidedness of college major. This finding was the most statistically significant (p = .005), with a larger percentage of returning men who were "fairly sure" of their chosen major in college and the opposite with more nonreturning male students who were "not sure." Interest- ingly, none of the men in either group were "very sure" of their intended major. Regardless, this variable directly relates to the notion of goal commitment in Tinto's model and warrants careful con- sideration due to its significance. The fact that this was not a significant issue for women may support Spady (1970) and Cope and Hannah's (1975) finding that the fulfillment of interpersonal needs is more important for women than men. As will be seen, such a finding did occur. 144 Null Hypothesis 2 The precollege variables of family income, high school aca- demic performance, and educational expectations (measured by the variables of decidedness of college major, highest degree expected, and expected grade point average in college) will not be found to have the best discriminating power between freshman-year persisters and students who do not return for their second year. This hypothesis was accepted. Again, despite the suggested influence of precollege variables by the chi-square analysis, a dis- criminant function with sufficient substantive significance could not be obtained. However, statistically significant functions resulted and surprisingly supported the alternative of Hypothesis 2. 1. Three variables were found to discriminate with more power than the other precollege variables between returning and non- returning students. In order, these were: decidedness of college major (a measure of educational expectations), ACT composite score (an indirect measure of high school academic performance), and family income. 2. For female students the variables of decidedness of college major, estimated first-year college grade point average (both indices of educational expectations), and expected extracurricular involve- ments in college formed a statistically significant function. Tinto's model would suggest that the women were predisposed to achieving a balance between goal and institutional commitment. The first two variables more clearly related to initial goal commitment, while the 145 expected involvement in extracurricular activities related to insti- tutional commitment. 3. For the male students, a slightly different configuration resulted. The three variables of family income, ACT composite score, and high school extracurricular involvements formed a statistically significant function between returning and nonreturning men. Unfor- tunately, the chi-square results did not reveal the directionality of the differences between the groups; therefore, it is more difficult to explain the nature of this finding. However, inasmuch as this function was not substantively significant, such an explanation, even if possible, would prove superficial. Null Hypothesis 3 Variables measured by the College Experiences Scale 11 when entered into a discriminant analysis will not form a discriminant function that distinguishes freshman-year persisters from students who do not return for their second year. Similar to the first hypothesis, a statistically significant discrimination resulted, but the degree of actual separation between returning and nonreturning students was minimal. This hypothesis was also accepted. Yet significant chi-square results occurred and should be examined. 1. The total group was found to be dependent on four vari- ables: extracurricular activities in college, family interactions, intellectual development, and current grade point average. The results of the first variable demonstrated that a "balance" existed in the 146 number of activities in which the student should engage. It was found that a larger percentage of returning students participated in four to six activities. On the other hand, nonreturning students were found to be either underinvolved (zero to three activities, sug- gesting a lack of social integration) or overinvolved (seven to nine activities, suggesting social integration occurred to the exclusion of academic integration). A similar balance could also be found on the variable of family interactions. Larger percentages of return- ing students reported three to four interactions, whereas the non- returning students reported only one to two interactions (possibly suggesting a lack of family support while in college) or seven to eight interactions (possibly suggesting either too much dependence on the family or too much interference). The variable of intellectual development showed a clear pattern in that a larger proportion of returning students reported being satisfied with their intellectual development than nonreturning students. The same trend existed for the variable of current grade point average. That is, a larger per- centage of returning versus nonreturning students reported GPA's in the range of 2.50-4.00. The profile of the returning student, according to these results, is indicative of a student who has become both socially and academically integrated into the university. That is, the student is more likely to have found an optimal level of extracurricular involve- ment, to have performed well academically and feels equally good about his/her intellectual growth, and has achieved a satisfactory degree of contact with his/her family. 147 2. When the sample was controlled for sex, the same four variables that were previously described were significant. In fact, an examination of the tables shows that similar patterns existed between the female students and the total sample, showing the influ- ence of the female students. 3. However, an examination of the findings for men showed their concentration on purely academic variables. A larger propor- tion of returning male students reported (1) more satisfaction with their intellectual development, (2) greater concern from the faculty for teaching and student develOpment, and (3) grade point averages in the range of 2.50-4.00. The probability that these results occurred by chance was 1 out of 100 for the first two variables and 1 out of 10,000 for the last variable. Clearly, these results portray the male student as academically integrated, with a strong goal commitment to obtaining the degree. Tinto's model would suggest that this student will persist to completion but may transfer to a school where he can achieve greater social integration as well as academic stimulation. Null Hypothesis 4 Variables measured by the College Experiences Scale 11 will not be better discriminators between freshman-year persisters and students who do not return for their second year than precollege characteristics. The stepwise selection summary tables for the College Experi- ences Scale 11 and the ACT Student Profile Section were compared. Three statistics that indicate which function is more powerful in its 148 discriminating ability were examined. The amount of variance accounted for (Partial R**2), the amount of residual discrimination (Wilks' Lambda), and the degree of association between the function and the groups (Canonical Correlation) were used to determine the "best function." However, the differences between the functions were extremely small. As a result, meaningful analysis was not possible and the null hypothesis was accepted. The data did, however, show an interesting trend: Although no significant results were found that could defi- nitely support college-experience measures or precollege character- istics as the more powerful discriminator of the two functions, there was some suggestion in favor of college-experience measures. It was found that residual discrimination was lower while the canonical correlation was higher for the function composed of measures from the CES II. This finding may point to Tinto's theory, which states the influence of one's experience on the campus is a more powerful determinant of persistence than background characteristics before enrollment. Null Hypothesis 5 Of the nine variables that correspond to the social-integration dimension of the College Experiences Scale 11, extracurricular activi- ties, peer and faculty interactions will not be found to have the best discriminating power between freshman-year persisters and students who do not return for their second year. 149 The statistical findings of this hypothesis are diluted when the degree of separation produced by the CES II is examined. A sta- tistically significant function did result, however, and the variables that entered the function will be reviewed. 1. It was hypothesized, based on previous research, that the most powerful variables that led to social integration would be extra- curricular involvements, contact with peers, and nonclassroom faculty contacts. The results showed that of the nine social-integration variables, extracurricular activities and faculty interactions did produce the best discriminating power over the other seven. Inter- estingly, peer interactions did not even enter the discriminant func- tion but activity devoted to personal time did. These findings could be explained by the fact that all of the students in this sample lived in campus residence halls where extracurricular involvement is encour- aged and, moreover, made convenient. Thus, a majority of peer contact occurs through structured activity which, indicated by the results, may be fulfilling not only a high involvement need but also the social contact. Faculty interaction is also very convenient in the residence- hall setting. Previous research (Terenzini & Pascarella, 1980) has shown the importance of this variable facilitating not only institu- tional commitment (and therefore persistence) but also academic inte- gration as the faculty member models scholarly behavior. The fact that activities devoted to personal time entered the function may suggest the limitation that residence-hall living imposes on one's opportunity for privacy. 150 2. Differences were found when the functions were controlled for sex. For female students, the variable of extracurricular activi- ties was accompanied by community involvements and family interactions. Male students, on the other hand, demonstrated a discriminant only with the variables of school activities and activity devoted to personal time. It seems that while both men and women found extra- curricular involvements as important, women were clearly more socially oriented through community involvements and family contact. As will be recalled, even the chi-square results demonstrated that social- integration variables were more significant than those indicating academic integration for women. Further, the chi-square findings emphasized the opposite for male students as the most significant variables were academically related. Null Hypothesis 6 Variables that correspond to the dimension of academic inte- gration will not be better discriminators between freshman-year persisters and students who do not return for their second year than social-integration variables. As with the other hypotheses, the findings need to be viewed with great caution due to the lack of substantive discrimination pro- duced by the function. However, five variables entered the function through the stepwise procedure and were suggestive of support for the alternative of this hypothesis. 1. For the total group, the findings revealed that the vari- ables of intellectual development and current grade point average were 151 the first and third items to enter the function. The first scale, intellectual development, was found to be the best discriminator, which indicated that above all the variables measured by the CES II, a student's perception of his/her own development has more influence on persistence than does grade point average. Therefore, it may not merely be the objective number of the grade, but how the student interprets the grade, that is important. Interestingly, the variable of extracurricular activities, indicating social integration, was found between intellectual development and current grade point average. Given the chi-square results in combination with the discriminant func- tion, it could be speculated that returning students have successfully negotiated a balance between social and academic involvement. Tinto (1975) would state that these students are not only the most likely to persist in college but will do so at Michigan State University. 2. The most support for the influence of academic-integration variables over social-integration variables was found for male stu- dents. The male returning student was characterized as placing more importance on academic integration due to the clustering of three variables in the discriminant function holding positions one, two, and three out of five. Current grade point average, faculty concern for student develOpment and teaching, and intellectual development provided more discriminating power than social-integration variables. The opposite was found for female students, who showed a greater bal- ance of both social and academic integration. While the academic variable of intellectual develOpment was found to be the most powerful discriminator in the function, the remaining three variables 152 corresponded to the social-integration dimension for women. Again, previous research by Cope and Hannah (1975), Pantages and Creedon (1978), and Ramist (1981) showed that persistence for women indicated a fulfillment of interpersonal and social needs. However, in this study, the importance of academic needs was also prevalent. Null Hypothesis 7 The variables of faculty concern for student development and teaching and faculty interactions will not be the most powerful dis- criminating variables between freshman-year persisters and students who do not return for their second year than any variable measured in the study. The stepwise discriminant analysis procedure was computed using all the variables measured in the study. Although a statis- tically significant function resulted with eight variables, the Partial R**2 and Wilks' Lambda disclosed no substantive discrimination between the groups. Further, the statistical finding did not support the alternative, and the null hypothesis was accepted. The statisti- cal significance of the function revealed: 1. The two most powerful variables that entered the discrimi- nant function were intellectual development and current grade point average. Similar to other findings in this study (and support by the chi-square results), both the student's actual academic performance and assessment of his/her intellectual development contributed most to persistence. 153 2. Differences were again noted by sex with intellectual develOpment and extracurricular activities acting as the two most powerful variables in the discriminant function for women. The vari- ables