Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
00001.tif
(USC Thesis Other)
00001.tif
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
INFORMATION TO USERS
This manuscript has been reproduced from the microfilm master. UMI
films the text directly from the original or copy submitted. Thus, some
thesis and dissertation copies are in typewriter face, while others may be
from any type of computer printer.
The quality of this reproduction is dependent upon the quality of the
copy submitted. Broken or indistinct print, colored or poor quality
illustrations and photographs, print bleedthrough, substandard margins,
and improper alignment can adversely affect reproduction.
In the unlikely event that the author did not send UMI a complete
manuscript and there are missing pages, these will be noted. Also, if
unauthorized copyright material had to be removed, a note will indicate
the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by
sectioning the original, beginning at the upper left-hand comer and
continuing from left to right in equal sections with small overlaps. Each
original is also photographed in one exposure and is included in reduced
form at the back of the book.
Photographs included in the original manuscript have been reproduced
xerographically in this copy. Higher quality 6” x 9” black and white
photographic prints are available for any photographs or illustrations
appearing in this copy for an additional charge. Contact UMI directly to
order.
UMI
A Bell & Howell Information Company
300 North Zeeb Road, Ann Arbor M3 48106-1346 USA
313/761-4700 800/521-0600
Retention Management: Predicting Minority Students
at Risk for Academic Difficulty
By
Mary Louise Maresh
A DISSERTATION
Submitted to the
Faculty of the Graduate School
University of Southern California
In Partial Fulfillment of the Requirements
for the Degree
DOCTOR OF PHILOSOPHY
Department of Counseling Psychology
School of Education
May 1995
Copyright 1995 Mary Louise Maresh
UMI Number: 9617117
UMI Microform 9617117
Copyright 1996, by UMI Company. All rights reserved.
This microform edition is protected against unauthorized
copying under Title 17, United States Code.
UMI
300 North Zeeb Road
Ann Arbor, MI 48103
UNIVERSITY OF SOUTHERN CALIFORNIA
THE GRADUATE SCHOOL
UNIVERSITY PARK
LOS ANGELES, CALIFORNIA 90007
This dissertation, written by
Mary Louise Maresh
under the direction of h..?T... Dissertation
Committee, and approved by all its members,
has been presented to and accepted by The
Graduate School, in partial fulfillm ent of re
quirements for the degree of
D O CTO R OF PHILOSOPHY
Dean of Graduate Studies
Date .J.f
DISi /IMITTEE
Chairperson
21Cju. ?■ mirtu
ii
Acknowledgments
I would like to thank the people who without
their help this dissertation would not have been
completed. First, I would like to thank my husband,
Gregory Horstman, and our children Elizabeth and
Alexander for the time they gave up with me so that I
could work on this project. Second, I would like to
thank Patricia Tobey for all the support and late
night conversations regarding databases. And third, I
would like to thank Barry Gribbons for his insightful
approaches to data analysis. And last but not least,
I would like to thank Nancy Cox for her continuous
moral support and proofing of this document.
iii
Table of Contents
Acknowledgments..................................il
Table of Contents...............................iii
List of Tables....................................v
Abstract....................................... vii
Chapter I: Introduction and Review of the
Literature........................................1
Review of the Literature..................... 6
Early Research in Retention of Students..... 6
Theoretical Background of Attrition.........7
The Spady Model.......................... 7
The Tinto Student Integration Model........8
The Bean Student Attrition Model.........12
Conceptual Model for Non-tradition
Students.............................. 14
Retention Questionnaires.................17
Research Studies Concerned with Student
Retention.................................18
Prediction Models for Retention............18
Minorities in Higher Education.............23
Non-cognitive Variables in Relation to
Academic Success.......................... 26
Self-concept Theory....................... 2 6
Soares and Soares Theory.................27
Marx and Winne Theory................... 27
Shavelson Theory........................ 28
Hattie and Song 3 0
Michael Theory ......................... 31
Research on Academic Self-concept and
Achievement 3 3
Recent Studies about Self-concept, Locus of
Control, on Academic Achievement 3 5
Locus of Control Theory................39
Achievement Motivation Theory..........41
Statement of the Problem.................... 46
Research Questions........................ 48
Chapter II: Method.............................. 51
Sample......................................51
Instrumentation............................. 52
Dimensions of Self-Concept Form H.........53
Intellectual Achievement Responsibility 55
Self Description Questionnaire III.........55
Cognitive Variables....................... 56
Procedure...................................57
Data Analysis...............................58
Limitations to the Study.................... 59
Chapter III: Results............................ 61
Findings of the Study....................... 61
Research Question #1..................... 61
iv
Research Question #2..................... 63
Research Question #3..................... 68
Research Question #4..................... 75
Research Question #5..................... 83
Summary.....................................93
Chapter IV: Discussion.......................... 96
Analysis of the Findings.....................96
Internal Consistency of Questionaires...... 96
Cognitive Variables....................... 98
Affective Variables...................... 100
Cognitive & Affective Variables...........102
SDQ III vs. DOSC-H....................... 106
Practical Implications..................... 108
Implications for Theory.................... 109
Implications for Future Research............113
Conclusions and Recommendations.............114
References 118
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
v
List of Tables
Page
Descriptive statistics for the scales
on the DOSC-H, SDQ III, and IAR.........62
Group descriptive statistics for the
three group discriminant analysis using
cognitive variables.................... 64
Wilks' Lambda and univariate F ratios
for cognitive variables.................65
Canonical discriminant functions of
cognitive variables.....................66
Classification results for cognitive
variables.............................. 67
Group descriptive statistics for the
three group discriminant analysis using
affective variables.................... 70
Wilks' Lambda and univariate F ratios
for affective variables.................71
Summary table for affective variables
in discriminant analysis................72
Canonical discriminant functions of
affective variables.................... 73
Cannonical discriminant function
evaluated at group means (group
centroids)............................. 74
Classification results for affective
variables.............................. 75
Group descriptive statistics for the
three discriminant analysis using
affective and cognitive variables.......77
Wilks' Lambda and univariate F ratios
for affective and cognitive variables.... 78
Summary table for affective and cognitive
variables in discriminant analysis...... 79
Canonical discriminant functions of
affective and cognitive variables.......81
vi
16 Canonical discriminant function
evaluated at the group means
(group centroids)...................... 82
17 Classification results for affective
and cognitive variables.................82
18 Group descriptive statistics for the
three group discriminant analysis using
the DOSC-H.............................84
19 Wilks' Lambda and univariate F ratio for
DOSC-H.................................85
20 Summary Table for variables in discriminant
analysis for DOSC-H.................... 86
21 Canonical discriminant function evaluated
at group means (group centroids).........87
22 Classification results for DOSC-H.......87
23 Grouped descriptive statistics for the
three group discriminant analysis using
SDQ III................................88
24 Wilks' Lambda and univariate F ratio
for SDQ III............................ 89
25 Summary table for variables in discriminant
analysis for SDQ III................... 90
26 Canonical discriminant function evaluated
at group means (group centroids).........92
27 Classification results for SDQ III.......92
vii
Abstract
The purpose of the present study was to examine
the predictive power of cognitive and affective
variables in relation to academic achievement for
college minority students at a predominately white,
urban university. The study utilized three measures of
cognitive ability (High school GPA, SAT Math, and SAT
verbal), five scales of academic self-concept, and four
measures of social self-concept. The measurement
instruments used to obtain the affective data were the
Dimensions of Self-Concept (Form H), the Intellectual
Achievement Responsibility Questionnaire (Short Form
B), and the Self-Descriptive Questionnaire III.
Eighty-seven minority students completed the
instruments during the summer and early fall of 1993.
These students were enrolled at the University of
Southern California which is a predominately white,
urban university. Academic achievement was defined by
the cumulative grade point average after the freshman
year.
The results of the five discriminant analysis
indicated that High School GPA and relationships with
the opposite sex were the best predictors of academic
success or difficulty. SAT Mathematics and SAT Verbal
were not good predictors for academic success. The
viii
group of students who had the highest mean score on the
SAT Mathematics and SAT Verbal had the lowest
cumulative grade point average.
Research in the area of retention of minority
students in higher education is lacking. Future
research needs to promote the development of reliable
measurement instruments specifically with minority
populations in mind. These instruments need to address
the particular social needs and adjustment problems of
minority students at a majority university. One area
of specific concern should be the area of serious
dating and overinvolvement in school sponsored
organizations.
1
Chapter I
Introduction and Review of the Literature
This study is concerned with pre-college
perceptions regarding a student's academic self-
concept and social self-concept which could contribute
to a student's ability to integrate academically and
socially into college life. Rather than look at
attrition long term with a population of minority
students, this study will try to develop an early
alert system which could help identify which minority
students might be at risk for academic difficulty so
that educational intervention programs could be
instituted.
Research in the area of retention of students in
higher education has centered mostly on majority
populations of students. More research has been
needed using minority populations of students. One
reason why minority populations have not been studied
is that students of color have not had a long history
in the American educational system because of
prejudice, poverty, and social injustice.
Administrators and policy makers in higher
education have been concerned about retention, but in
these times of fiscal cutbacks and decreased
enrollment, increased attention is being focused on
retaining students. Most major colleges and
2
universities have been researching their retention
rates and have been trying to implement programs
designed to increase these rates. The number of
students who leave college prior to completion of a
degree now exceeds the number that remain on campus.
As Vincent Tinto (1987) noted, "Of the 2.8 million
students who in 1986 will be entering higher education
for the first time over 1.6 million will leave their
first institution without receiving a degree. Of
those, 1.2 million will leave higher education without
ever completing a degree program, two or four year"
(p.l). The retention rates for minority students in
higher education are even more dismal. Approximately
50% of all white college students will graduate in six
years after starting college as compared to 25% of all
minority students (Stewart, 1991).
The retention rates of major universities and
colleges have also received increased attention
because of the Student-Right-To-Know-Act, a new law
which mandates that major universities and colleges
publish their retention rates for perspective and
continuing students. Therefore, it is in the best
interests of colleges and universities to work toward
high retention rates, as it is likely that many
students will choose to attend those universities and
colleges with the highest rates of retention relative
3
to their individual academic needs. Research in the
area of student retention has been on the rise with
more concern being paid toward predicting which
students are in jeopardy of either experiencing
academic difficulty or risking dropping out of
college.
Although research in the area of student
retention has been on the increase, there have been a
number of problems that have hampered the advancement
of research in the area. Three major problems have
been (a) the lack of a common definition for the term
"dropout", (b) the fact that most research tends to be
atheoretical, and (c) the lack of valid measures or
scales available to predict which students are at risk
for academic difficulty or are prone to drop out of
college.
The problem of how to define those students who
will not continue their education and will not finish
their degree has confounded researchers in this area.
Usually the term "dropout" has been defined to fit the
type of research being completed. For instance, if
the researcher is concerned about the number of
students who leave college and who never earn a
degree, the definition of "dropout" will include those
students who transfer to another university and
complete their degree at the student's second
4
university. If a university's faculty and
administrators are concerned about their own retention
rate, transfer students will be considered as
"dropouts.” All retention studies do not have the
same purpose.
The present study highlights another definitional
problem with research in this area: It is concerned
with predicting which incoming minority freshmen will
experience academic difficulty during their freshmen
year. Academic difficulty and drop-out are not
synonymous terms, although there is a great deal of
overlay. It is hoped that once identified these
students can be targeted for some type of intervention
to keep them in school without experiencing academic
difficulty. These students are not the same ones who
are usually studied in most retention studies. Most
retention investigations target students who leave
college voluntarily — an emphasis which means that
these individuals may or may not have been
experiencing academic difficulty. Students who have
been academically disqualified from the university are
usually not included in retention research. The
present study is directed toward those students who
may be placed on academic probation after their first
year of college.
5
The second difficulty with retention research is
that it is not driven by theory. Research which is
not driven by theory is problematic for two reasons.
First, without a theoretical background research
studies cannot be used to explain the general laws of
the area being researched, and secondly, the research
can be used only in limited ways to predict future
events. The following statement supports this second
concern.
The function of science is to establish
general laws (theory) covering the behavior
of the empirical events of objects with which
the science in question is concerned, and
thereby to enable us to connect together our
knowledge of the separately known events, and
to make reliable predictions of events yet
unknown (Braithwaite, 1955 p.l).
In order to increase the knowledge base in the
retention of students in higher education the research
needs to be based on theory.
The third problem with retention research is that
there are no highly valid instruments which can
accurately predict which students will experience
academic difficulty during their freshman year in
college. The few instruments which are available have
not exhibited promising validity especially with
minority populations or single-item responses on a
questionnaire are used to represent a specific
6
prediction. According to Marsh (1986), the use of
single-item responses on questionnaires has less
reliable and valid psychometric properties than do the
multi-item scales.
Review Of The Literature
Early Research in Retention of Students
Alexander Astin (1970, 1975, 1977, 1982, 1986)
has been researching student attrition in higher
education since the early 1970's. He developed a
model of student departure, "The College Impact
Model," which looked at the impact of the college
environment on students. Basically, the model has
three components: student input, the college
environment, and student output. Astin conducted his
first national survey of entering freshmen in 1961.
This survey was used to develop the Cooperative
Institutional Research Program (CIRP) which began in
1966. In 1973 the University of California, Los
Angeles joined the CIRP and continues currently to
collect data from entering freshmen nationwide to
learn more about the reasons students leave college.
The research completed by Astin (1975, 1977) on
student retention suggested that students are affected
by institutional characteristics, student-institution
fit, and institutional involvement. Some of the
institutional characteristics which affect student
7
attrition are: (a) type of control, (b) geographic
region, (c) being coeducational, and (d) institutional
size. The student-institution fit refers to the
meeting of the institution of the needs of the
student. This fit is achieved by having the major the
student wants and by having the type of other students
with whom this student wants to interact in college.
The institutional involvement pertains to how closely
the student becomes assimilated with activities,
clubs, and other social events on campus. All three
of these major factors influence a student's
likelihood of remaining at a particular university.
Although Astin has completed extensive research
to understand the factors which influence student
attrition, his model does not attempt to explain why
these factors influence students. His model of
attrition is more descriptive of what influences
contribute to attrition rather than why these factors
influence students. To understand the later it is
necessary to look at theoretical models which attempt
to answer the question of "why” students drop out of
college.
Theoretical Background of Attrition: Alternative
Models
The Soadv model. One of the major theoretical model of
retention of students in higher education began with
8
Spady (1970). His theory was based on Durkheim's
(1961) research on suicide. Spady stated that
dropouts from higher education were similar to suicide
victims. Both were dropping out of a social system in
which they were not invested. Spady declared that
students need to share the same values which were held
by the academic community regarding the importance of
academic work and the need for positive reinforcement
from others who can give emotional and financial
support. Without these needs being met the student is
in danger of dropping out of college.
The Tinto student integration model. Tinto (1975,
1987) expanded on Spady's theory and clearly defined
academic and social supports. Tinto's model, The
Student Integration Model, emphasizes the longitudinal
nature of deciding to leave school. Tinto suggested
that leaving school is not a spontaneous decision. He
delineated the background factors which may affect a
student's decision to leave. These background
factors, which include socioeconomic status, high
school experiences, community of residence, ability,
educational plans, aspirations, and motivation, have a
direct influence on the collegiate experience of
students particularly during their freshman year.
In 1987 Tinto expanded his theory to include the
idea of the "rites of passage" which a student needs
9
to accomplish in order to survive at a university or
college. Tinto indicated that the freshman year
transitions were critical to the student's academic
success. These transitions include separation from
family, transition to new values and activities, and
incorporation of a new set of values and behaviors.
These transitions, which include separation from
family and increased independent living, may be
especially difficult for minority students whose
family traditions and support represent a major
component in the students' lives. To separate from
family could mean compromising cultural values and
being disowned by the family (Ogbu, 1989). Often
minority students face a more difficult transition to
college especially at a predominately white
university. One explanation for this has been that
white culture values independence and self-sufficiency
to a great degree, where many minority cultures value
cooperation (Covington, 1992).
The majority of research in the area of retention
in higher education which uses Tinto's (1975) theory
of why students leave college has been completed by
Terenzini and Pascarella (1977, 1978, 1980). In 1977
Terenzini and Pascarella proposed a study to validate
Tinto's model of student attrition which purports lack
of social and academic integration of the students as
10
the main reasons for students' voluntary withdrawal
from college. They found social and academic
integration to be significantly valid predictors for
voluntary freshman attrition. Academic integration
was measured by students' college grade point average
(GPA) after the freshman year, and by a self-report
instrument which measured students' perceptions of the
academic program. Social integration was measured by
a self-report instrument which questioned the students
on their perceptions of their non-academic life and
the number of extracurricular activities with which
they were involved at the university. The results of
a discriminant analysis suggested that social
variables contributed more of the variance between the
leavers and stayers than did cognitive variables such
as scholastic aptitude tests.
In 1978 Terenzini and Pascarella investigated the
predictive ability of precollege characteristics to
ascertain whether they significantly influenced
student attrition. The pre-college characteristics
used by these two researchers were (a) parental
educational level, (b) high school rank and GPA, (c)
Scholastic Aptitude Test (SAT) scores and (d) a
measure of personality needs from Stern's (1970)
Activity Index which measures achievement orientation,
dependency needs, emotional expression, and
11
educability. They found that the precollege
characteristics alone were not significant predictors
of stayers, but when used in conjunction with social
and academic integration data they could help predict
student persistence. In their study, academic
integration variables explained the most variance
between leavers and stayers.
In 1980 Pascarella and Terenzini developed a
multidimensional questionnaire to assess the major
dimensions of Tinto's model. This 34 item
questionnaire measures students' institutional
integration on five independent scales. These authors
suggested that the questionnaire can be used with
freshmen during the second half of the year to
identify which students are at risk for voluntary
departure from college. The questionnaire would have
limited use with freshmen before they began their
college career to ascertain which freshmen might
experience academic difficulty.
The studies conducted by Terrenzini and
Pascarella (1977, 1978, 1980) suggested that Tinto's
model can be useful in understanding the complex
problems associated with college attrition. It has
had limited validity, however, in predicting which
students will have difficulty during their freshman
year. Most of the research completed by Terrezini and
12
Pascarella has not addressed ethnic differences in
student populations and ways in which ethnicity may
affect students' needs in higher education.
The Bean student attrition model. Another attrition
theorist, Bean (1980) has disagreed with Tinto's
analogy of dropout behavior being similar to suicide
behavior. He has suggested instead that the student's
dropout behavior was similar to turnover in work
organizations. Bean was influenced by Price (1977) ,
Fishbein and Ajzen (1975) and Bentler and Speckart
(1979) who explored why people leave their jobs and
how beliefs and attitudes affect behavior. Price
(1977) apparently influenced Bean by researching
institutional work turnover, specifically why people
voluntarily left their jobs to find a different job.
Fishbein and Ajzen (1975) and Bentler and Speckart
(1979) provided Bean with a psychological model which
suggests that an individual's behavior is the result
of a cyclical process by which beliefs affect
attitudes, which in turn lead to intentions and
ultimately to the behavior in question.
In Bean's (1980) theory (the Student Attrition
Model) there are ten organizational variables which
are considered to be indicators of a student's
interaction with the college organization which the
student is attending. These variables (grades,
13
practical value of the education, self-development due
to the education, repetitiveness of school life,
information about academic life and policies,
participation in decision making, peer interaction,
having the courses wanted, fairness of school rules
and regulations, and membership in campus
organizations) have a direct impact on the
satisfaction that a student feels with college. The
more favorable the student feels the less likely the
student is to drop out of college. Opportunity to
transfer and to marry were seen as external variables
which impacted the student's satisfaction and desire
to remain in college.
In a larger version of this conceptual model of
student attrition Bean (1985) suggested three groups
of exogenous variables which affect student dropout
behavior: (a) academic variables (high school gpa),
(b) social-psychological variables (social life), and
(c) environmental variables (outside friends). There
are also endogenous variables which are
socialization/selection factors which influence
attrition. These variables include college grades,
institutional fit, and institutional commitment.
Using this model he studied students at a large
research university and found the endogenous variables
to be more predictive of dropout behavior than the
14
exogenous variables. Specifically, he noted that
socialization had a major impact on a student's
decision to leave college. Peer support was seen as a
necessary but not sufficient element in college
persistence. The subjects in this study were all
white, 23 years old or younger, and not married.
Conceptual model for non-traditional students.
Metzner and Bean (1987) also proposed a conceptual
model of non-traditional student attrition. The group
for whom the model was devised comprises part-time
commuter students at a Midwestern urban university.
Unlike their traditional counterparts, the non-
traditional student dropped out of college for either
academic reasons or non-commitment to college rather
than for social reasons. The results of this study
also suggested that background variables had a greater
effect on college GPA than did other intervening
variables during the school year.
The weakness of Bean's models lies in their
inability adequately to explain why students leave
college. How do these ten organizational factors
impact the student socially, psychologically, and
academically and in turn cause them to decide to drop
out of college? Are all factors equally responsible
for dropout behavior or do they have a weighted or
cumulative effect?
15
Expanding on his model, Bean (1990) developed a
longitudinal model of the type of factors which
influence students to leave college. He incorporated
aspects of Tinto's (1987, 1988) social and academic
integration into the model as factors which influence
students to depart from college. Also included in
this model were organizational variables and
environmental pull variables which were part of his
first model. The attitudes and beliefs about the
academic experience appeared to influence whether or
not they remained at college. These attitudes and
beliefs were not explored, however, as factors which
could influence students during their pre-college
experience on their ability to integrate socially and
academically into college life.
Recent research has attempted to ascertain
whether the two models (Tinto's Student Integration
Model and Bean's Student Attrition Model) can be
merged to learn more about retention theory (Cabrera,
Nora, & Castaneda, 1993). Cabrera et al. utilized
structural equation modeling to study which factors of
both theories would provide the most meaningful
analysis of the variance explained. They found many
similarities between the two theories which were also
noted by Hossler (1984). Both theories regard the
16
retention of students to comprise a complex set of
variables which begin with the pre-college
characteristics of the students and include a match
between the student and his or her institution.
The results of the path analysis indicated that
persistence could be understood better by the
combination of the two theories. A problem with the
Tinto (1987) model alone was the weak contribution
given to environmental factors which are important in
the Bean model and which have been supported by
current research (Cabrera, Nora, and Castaneda, 1993).
Environmental factors such as support from family and
friends and attitudes towards financial matters proved
to be significant in a student's persistence in
college.
A significant problem with much of the current
research being conducted on student attrition is the
lack of operationalized concepts which can be easily
measured by standardized questionnaires. It has
seemed that each study has used different measures to
test for the same construct. For instance, there is
no set way to measure academic and social integration.
Thus researchers have used whatever measures they have
available that seem to fit their own idiosyncratic
definition. Research is underway to rectify the
problems encountered with the specific concepts
17
(social and academic integration) used in retention
research but many gaps and problems remain.
Retention questionnaires. There are two
questionnaires which have been developed to measure
academic and social integration in higher education.
The Student Integration Survey (SIS), was developed
using item response theory to help operationalize the
concepts of academic and social integration (Dowaliby,
Garrison, & Dagel, 1993). The authors of the SIS
questionnaire had struggled for several years using
factor analysis to identify reliable factors but were
not successful. This questionnaire probably relies
more on single-item responses than on multi-item
responses to capture measures of specific constructs.
Krotseng (1992) recently developed a
questionnaire, the Student Adaptation to College
Questionnaire (SACQ) to aid in the study of retention
of students in higher education. This questionnaire
measures overall adjustment to college. Discriminant
analysis correctly classified 79 to 85% of the
students surveyed as either leavers or stayers. There
was no mention of the ethnicity of the students
completing the questionnaire. The results indicated
that this questionnaire was most effective in being
able to predict those students who leave college
during the first semester. Additional research is
18
needed to ascertain whether this questionnaire can
also predict which students will experience academic
difficulty during the remainder of their freshman
year.
Research Studies Concerned with Student Retention
In researching the literature which addresses the
issues of student retention in higher education, this
review has identified two different types of research
studies. Retention literature has been concerned with
what factors (a) influence students after they have
begun their college career and (b) cause them to drop
out of college. The great majority of these studies
have been directed towards persistence rather than
academic difficulty.
Prediction models for retention. Prediction modeling
addresses another type of research in retention
management which is trying to target academically at
risk students, usually freshmen, by using precollege
characteristics as prediction variables. Researchers
in this area have generally used Tinto's (1987) or
Bean's (1990) theory to guide the selection of
variables, or have conducted atheoretical research
because no current theory adequately addressed the
selection of desired variables.
The typical list of pre-college prediction
variables include both cognitive and non-cognitive
19
variables. The most widely used cognitive variables
have been the student's high school GPA and scholastic
aptitude test with verbal and math components.
Typical non-cognitive variables have been academic
self-concept or specific questionnaire items which
have been oriented toward social and academic
integration or perceptions and attitudes regarding
college and college life.
Not all research on non-cognitive variables have
supported their use to help predict students success
in college. Holmes (1992) used two non-cognitive
measures, The Life Style Analysis Test, (Cassel,
1990a) and the Independence Versus Regression Test
(Cassel, 1990b) to help predict student success in
college. He did not find any significant correlation
between the scores on these tests and the student1s
GPA in college. The ethnicity of the subjects was not
revealed in the study.
Some researchers have specifically addressed the
differences found in minority student populations when
looking at specific variables to help predict a
student's academic success in college. These
researchers have found that cognitive variables such
as Scholastic Aptitude Tests (SAT) scores and high
school GPA have not been good predictors of academic
success in minority students (Astin, 1975, Duran,
20
1986) . They have indicated that a number of non-
cognitive variables, such as academic self-concept and
family support, might be more valid predictors of
academic success in these particular populations than
have traditional cognitive measures.
Tracey and Sedlacek (1984) developed the Non-
cognitive Questionnaire (NCQ) which was used to
measure the non-cognitive variables hypothesized by
Sedlacek and Brooks (1976) to be more valid predictors
of academic success for minority students than have
cognitive variables such as SAT scores. The variables
hypothesized by Sedlacek and Brooks were: (a)
positive self-concept, (b) realistic academic
appraisal, (c) understanding racism, (d) delayed
gratification, (e) availability of strong support
persons, (f) leadership experience, and (g) community
service experience. Further research using this
instrument was conducted by Tracey and Sedlacek (1985)
who studied minority students at a predominately white
campus. They suggested that the variables which
influence academic achievement are different for
minority students than are those for white students.
They provided support for the belief that cognitive
variables do not predict success in college for
minority students to any appreciable degree. Success
for minority students is instead linked to non-
21
cognitive variables such as positive self-concept and
positive academic assessment.
White and Sedlacek (1986) found the NCQ to be
predictive of success and persistence of students who
had been admitted to college under special
circumstances (usually they did not meet the
requirements for general admission to the university
but had some special talent that determined their
admission). The sample of students was mixed, with
55% white and 38% black. After the freshman year,
measures reflecting positive self-concept and
successful leadership were shown to be valid
predictors of students' success.
The research literature using the NCQ has
supported the hypothesis that academic self-concept is
an important variable for minority students when
trying to predict which students will have difficulty
during their freshman year. However, the NCQ does
seem to have had some problems with factor stability
(Arbona & Novy, 1990).
Arbona and Novy (1990) found virtually no
predictive power in the NCQ when investigating
persistence and grades at a predominately white
Southern university. The sample consisted of 95
African-American, 96 Hispanic, and 555 Caucasian
students. Arbona and Novy examined the factorial
22
structure of the NCQ. Their analysis yielded eight
factors which were different from the original factors
used by Sedlacek and associates (Sedlacek, 1987;
Sedlacek & Brooks, 1976). The eight factor subscales
found by Arbona and Novy revealed no scale which could
be described as representing positive self-concept or
appraisal of academic ability, although the original
NCQ was purported to measure academic self-concept.
This study raises some doubt as to the validity of the
NCQ.
Poole (1989) studied disadvantaged minority
students, all from low income families, attending
several private and public colleges in New Jersey.
She used two different measures, the Dimensions of
Self-Concept (DOSC-H) (Michael, Smith, & Michael,
1989) and the Intellectual Achievement Responsibility
(IAR) (Crandall, Katkovsky, & Crandall, 1965)
Questionnaires, which measure academic self-concept
and academic responsibility respectively. Poole found
no significant relationship between achievement levels
of those students who experienced academic difficulty
during their first semester and scores on the scales
on the DOSC-H and IAR Questionnaires. This may have
been due to the fact that all the students in the
study were targeted for remedial college courses
rather than the regular college curriculum. The
23
students in this study were unprepared to begin their
college careers.
In summary, the preceding research has lent
support to the hypothesis that minority retention in
higher education is more complicated than majority
retention and that retention may be influenced more by
non-cognitive variables than by cognitive variables.
The non-cognitive variable which appears to have the
most influence on minority persistence is a positive
academic self-concept or belief in academic ability.
Minorities in higher education: retention issues.
Significant opportunities for people of color in
higher education did not become a reality until the
Civil Rights Movement in the late 1950's and early
1960's. Colleges and universities began recruiting
minority students in harmony with the national goal to
provide equal access to quality institutions of higher
education. Despite the legislation generated by the
Civil Rights Movement the goal of equal opportunity
for minority students has yet to be realized. In the
1970's and early 1980's minority enrollment in higher
education increased, but by the late 1980's and early
1990's minority enrollment again decreased. Some
critics blamed this decrease in minority enrollment on
the Reagan Administration and the cutbacks to
education suffered during the years 1980-1990.
24
Retention figures for minority students have been
even more dismal than enrollment figures. College
educators and administrators have recently called for
increased efforts to find ways to help students of
color not only gain admission to college but to
succeed in college once they are there.
At the University of Southern California, which
is a predominately white, private, urban university,
the retention rates for both African American students
and Hispanic students have been the lowest of all
ethnic groups. Asian students on this campus have
similar retention rates as Caucasians (Servis, 1993) .
With such a differential in retention rates, it is
essential to learn as much as possible about how to
predict which students will experience academic
difficulty during their freshman year so that
intervention programs can be initiated to help retain
students of color. Attrition at the University of
Southern California for minority populations seems to
be partially linked to academic difficulties.
Academic difficulty was defined by Servis (1993) as
having a GPA below 2.0 or by receiving an F (Fail) or
W (Withdrawal) in a course.
There has been no theory which looks specifically
at minority attrition in higher education. Some have
25
postulated that there should be no difference in the
reasons minority students leave college as compared to
those for white students (Hossler, 1984). Others have
indicated that minority students experience a
different set of problems from that of majority
students when they reach college (Astin, 1982; Olivas,
1986)
In a book entitled Latino College Students
(Olivas, 1986), Duran wrote a chapter entitled
"Predicting Success in College for Hispanic Students"
which addressed variables that affect Hispanic
retention. He has found that academic variables such
as high school GPA and standardized aptitude test
scores do not predict accurately which Hispanic
students succeed in college. He suggested that non-
cognitive variables within a multivariate statistical
analysis framework be used instead, and that single
university populations be employed as the best way to
learn how to predict college success for Hispanic
students.
Alexander Astin (1982) in Minorities in American
Higher Education reported several factors which he
hypothesized influenced the college success for
minority students. In his studies, high school GPA
was the most valid predictor of college GPA for
African Americans, Hispanics, and Native Americans.
26
He also found that scholastic aptitude tests did not
relate to success in college. Astin observed that
minority students with the best chance of success in
college were those with good high school grades, well
developed study habits, and high but realistic self
esteem in terms of academic ability.
Non-Cocmitive Variables in Relation to Academic
Success
In order to understand why students have not been
successful in college and why they have not been
motivated in order to attain their college degree an
understanding is needed how academic self-concept,
locus of control, and achievement motivation affect
students in their academic endeavors. Specifically,
consideration needs to be given as to whether or not
these concepts can be used with minority populations
in order to understand which students will experience
academic difficulty in their freshmen year.
Self-concept theories. Among psychological constructs
that of self-concept may be one of the most
researched. Yet prior to about 1975, reviews of the
self-concept literature revealed that the research in
this area lacked reliable and valid instruments and
theoretical models. Since 1975 several models or
theories of self-concept have emerged in the
literature. The areas of contention between the
27
models of self-concept lie in part on the beliefs on
whether self-concept is unidimensional or
multidimensional, whether it is hierarchically
ordered, and whether there is a general self-concept
which is a common factor under which all aspects of
self are organized. There have been at least five
major theories of self-concept.
Soares and Soares theory. Soares & Soares (1982)
contended that self-concept is a multidimensional
concept with no higher order factor of a general self-
concept. They maintained that a person's self-
concept develops through experiences, relationships
with significant others, and each person's unique
capabilities which form independent, unrelated facets
of self-concept. There are several problems with the
research using this theory. One problem is the
relatively narrow description of self-concept; they
have been primarily concerned with different aspects
of academic self-concept. Soares and Soares have not
looked at areas such as social self-concept, physical
self-concept or peer self-concept. This emphasis may
be why they have not viewed self-concept to be
hierarchical but multidimensional.
Marx and Winne theory. Marx and Winne (1978)
developed a model of self-concept which contends that
self-concept is a compensatory bi-polar construct.
28
The general facet of self-concept emerges, leaving the
remaining variation to be bipolar among second-level
facets of a hierarchical model. These second-level
facets are inversely related. Lower status on one
facet of self-concept means that the person
compensates with a higher level on some other facet of
self-concept. This compensatory formulation can be
put into general terms. If a student has a high
academic self-concept, he or she will probably have a
low social or physical self-concept. The evidence for
this theory has been sparse (Hattie, 1992) although
there has been some concession that this model may
describe certain populations of people.
Shavelson theory. Shavelson, Hubner, and Stanton
(1976) attempted to address the problems facing
researchers in the area of self-concept. They
proposed a multifaceted, hierarchical theoretical
model of self-concept. They broadly defined self-
concept to be a person's perceptions of him\herself
which are formed through experiences with significant
others (parents, teachers, peers), one's environment,
and attributions of one's behavior.
Shavelson's theory of self-concept is defined by
seven major features. The first feature contends that
self concept is structured or put into categories by
people as they attempt to process the vast amount of
29
information they have about themselves and relate
these categories to one another. Second, self-concept
is multifaceted and the facets are influenced by the
individual and or the group or groups to which the
individual belongs. Third, self-concept is
hierarchical with perceptions of personal behavior at
the core, with inferences about subareas of the self
in the next level, and then with general inferences of
the self. Fourth, the core of the self-concept is
stable and becomes less stable as one moves away from
the core. Fifth, self-concept becomes increasingly
more multifaceted as one transitions from infancy to
adulthood. Sixth, self-concept has a descriptive and
evaluative aspect. Seventh, self-concept can be
differentiated from other constructs such as academic
achievement.
Another perspective of the model is as follows:
the apex of the model (the core element) is divided
into academic self-concept and non-academic self
concept. Academic self-concept is then subdivided
into particular subject areas, and non-academic is
divided into social, emotional, and physical.
Additional research by Marsh (1984, 1986, 1987,
1990, 1992) and Marsh and Hocevar (1985) on the
Shavelson model revealed that the model was indeed
hierarchically ordered, but that academic self-concept
30
was composed of two second-order factors - -
mathematics/academic and verbal/academic - - rather
than the subject areas defined by Shavelson. These
two second-order factors were not highly correlated
with one another but were correlated with mathematics
achievement and verbal achievement, respectively. As
already indicated, Marsh has completed extensive
research on the Shavelson/Marsh model of self-concept
by paying particular attention to within-network
studies which have been concerned with the construct
validity of his self-concept instrument, the Self-
Descriptive Questionnaire (SDQ) (Marsh, 1982).
Particularly noteworthy for this current research
project is the academic self-concept portion of the
SDQ. Because of the strength of the family connection
in the African-American and Hispanic community, the
scale which measures family relationships may provide
added predictability to this researchers prediction
model. Students who perceive that they have more
family support may be academically more successful.
Hattie & Song theory. Hattie and Song (1984) have
made two modifications to Shavelson's self-concept
model. First, they divided academic self-concept into
achievement, ability, and classroom self-concepts.
Ability self-concept is defined by the extent to which
an individual believes he or she is capable of
31
achieving. Achievement self-concept is defined as the
actual product of a person's achievements at any given
point in time. Classroom self-concept relates to the
confidence an individual exhibits in activities
associated with classroom behavior. The second
modification to the Shavelson model occurs in the
classifications under the non-academic portion to the
model. Hattie and Song divided non-academic self-
concept into two second-order factors labeled social
self-concept and self-regard or presentation of self.
Social self-concept is subdivided into family and
peer. Self-regard is subdivided into confidence in
self and physical self-concept.
The differences among the Shavelson, Shavelson &
Marsh, and the Hattie & Song models of self-concept
are relatively slight. All three models postulate
that self-concept is multidimensional, and
hierarchical, and that a general self-concept factor
is present. The differences between these models are
in the second-order factor categories into which
conceptions of self can be grouped.
Michael theory. Michael (Michael & Smith, 1976;
Michael, Smith, & Michael, 1989) has a theory of
academic self-concept in which constructs are
hypothesized regarding perceptions about the self in
terms of level of aspiration, anxiety level, academic
32
interest and satisfaction, leadership ability, and
identification vs. alienation. He devised a
questionnaire to help predict which students may be at
risk for academic failure in school because of a low
degree of self-worth. The theory which guided the
work on the self-report questionnaire entitled
Dimensions of Self-Concept (DOSC) (Michael et al.,
1989) originates from the rationale of affectivity in
school learning. How we feel influences how we learn.
The five subscales in the questionnaire correspond to
the five constructs underlying the theory of academic
self-concept which can hamper or improve school
learning.
The theory begins with academic aspirations which
may be either too high or too low. These aspirations
can lead to anxiety about the fear of failure or loss
of status if these aspirations are not attained.
Students who are relatively free of anxiety and who
have realistic aspirations enjoy school and are
satisfied with the school experience. This
satisfaction usually leads to acquiring leadership
roles and positive identification with school. In
general, one could say success generates success,
failure leads to subsequent failure and alienation.
The components of the DOSC are similar to the
components in the Tinto (1987) model of retention of
33
students in higher education. The
Identification/Alienation subscale and Leadership and
Initiative subscale both deal with the need for social
integration into the school environment. The Academic
Interest and Satisfaction constructs coincide with the
academic integration needed successfully to attain a
degree. The Level of Aspiration subscale deals with
the need for a student to have goals and intentions
about his or her academic career in order to achieve.
Research on academic self-concept and its
predictive ability with students in the area of
achievement continues to be growing. The
psychological construct of academic self-concept is
slowly gaining in acceptance as self-concept is being
conceptualized as multidimensional rather than
unidimensional. Measuring self-concept in specific
areas rather than as a general measure should increase
the correlations found between academic self-concept
and achievement.
Research on Academic Self-Concept and Achievement
There has been a longstanding belief that a
positive self-concept should correlate with high
grades in school, yet the research which has been done
on this relationship has been confusing and
confounded. The reasons for this confusion may lie in
the many different terms used to describe self-
34
concept, the lack of reliable and valid measures of
self-concept, the outcome measures used to describe
academic achievement, and the statistical analyses
employed.
In a meta-analysis completed by Hansford and
Hattie (1982), the general conclusion was that the
research on academic self-concept and achievement
indicated a positive correlation (.42). The
correlation with general self-concept and academic
achievement however, was much lower (.18). Several
problems were encountered by Hasford and Hattie in
this analysis, which may limit their findings. These
problems were with the terms used to define self-
concept, the number of different measurement tools
administered to measure self-concept, the number of
different achievement/performance tests employed, and
the differences in the age of the students used in the
study. While searching the literature on self-concept
they found 15 different terms used to describe self-
concept and 58 tests used to measure this construct.
Many of the tests employed in the studies had no
reliability reported, comprised only a few items, and
represent developmental efforts by the authors of the
published study. There were 61 different measures of
academic achievement found in the literature. The age
range of the students studied were preschool through
35
college level. With these problems in mind, in the
following section a review will be presented of the
current literature in the area of self-concept, locus
of control and academic achievement.
Recent Studies About Self-Concept. Locus of Control,
and Academic Achievement
There have been numerous studies which have
looked at self-concept, locus of control, and other
non-cognitive factors as predictors of academic
achievement. Most of these studies have used
elementary and secondary school aged children (Bachman
& O'Malley, 1986; Byrne, 1986; Lyon & MacDonald, 1990;
Shavelson & Bolus, 1982). The results have not been
conclusive. Shavelson and Bolus (1982) and Lyon and
MacDonald (1990) found a positive correlation between
academic self-concept and academic achievement. In
contrast Byrne (1986) and Bachman and O'Malley (1986)
found no significant correlation between the same
variables.
At the college level there have only been a few
studies which have been concerned with the predictive
power of academic self-concept and academic locus of
control on academic achievement. Gerardi (1990)
studied 98 freshmen engineering students attending a
New York University and found academic self concept to
be a valid predictor of academic success. Academic
36
self-concept was fairly highly correlated with college
GPA (.57) whereas high school GPA correlated at a much
lower rate (.20) with college GPA. In his sample of
98 students, 87% was from either African-American or
Hispanic descent. Academic self-concept in this study
was measured by Brookover's Self-concept of Ability
Scale (SCA) (Brookover, Thomas, & Paterson, 1964).
Tracey and Sedlacek (1985) carried out research
using their Non-Cognitive Questionnaire (NCQ) on non-
cognitive variables affecting academic success. They
found that of the eight variables examined, overall
positive self-concept and realistic self-appraisal of
academic ability were assessed to be predictive of
academic success for both African-American and
Caucasian students. The study sampled approximately
2500 students at a large eastern state university. Of
the 2500 students, 400 were African-American.
In a similar study by Boyer and Sedlacek (1988)
again using the NCQ scoring high on a positive self-
concept measure and having a support person were found
to be predictive of academic success with
international students at the college level. This
study was completed at a large, Eastern state
university with a sample of 248 students. College GPA
and continuing enrollment at the university were the
outcome measures. Boyer and Sedlacek hypothesized
37
that there were more factors involved in retention of
students the farther one moves away from the typical
Caucasian Euro-American student. Therefore, minority
students who were reared in the United States would
have more factors which could influence their
retention in higher education than would Caucasian
students, but fewer factors than would international
students.
House (1992) found a significant relationship
between academic self-concept and students'
persistence in college measured during the fourth and
eighth semester. There were 2343 students in the
study, 41 were African-American, 36 were Hispanic, 36
were Asian American, and the remainder were Caucasian.
Academic self-concept was measured by only three
items.
Wilhite (1990) found academic self-concept and
locus of control to be significant predictors of
academic achievement in a college course. The
subjects of this study were 184 college students
enrolled in a psychology course. The measure of
academic self-concept was the Self-Concept of Academic
Ability Test (Brookover, Erickson, & Joiner, 1967),
and the locus of control measure was the Adult
Nowicki-Strickland Internal-External Control Scale
38
(Nowicki & Duke, 1974). There was no indication of
the ethnicity of students in the study.
There have been a few studies which have showed
no predictive ability of academic self concept in
students' academic success (Arbona & Novy, 1990). For
example, Arbona and Novy investigated the predictive
power of Tracey and Sedlecek's (1984) NCQ (as reported
earlier). They found no significant relationship
between the factors on the questionnaire and academic
achievement of minority college students.
The major problems with evaluating research in
the area of academic self-concept, locus of control,
and academic achievement and its relationship to
academic achievement can be summed up in three main
categories. First, there has not been any general
agreement on what academic self-concept, locus of
control, and academic achievement is, much less any
consensus on ways in which to measure these
constructs. Second, there has been no standard way to
define academic achievement. Third, there has been no
consensus on how to analyze the data. However, one
can conclude from the more recent research that some
of these problems are beginning to be addressed.
Several researchers have begun to define and to try to
measure academic self-concept (Marsh, 1984, 1986;
39
Michael et al., 1989). Specific questionnaires are
beginning to surface which operationalize parts of
Tinto's (1987) retention theory, and multivariate
statistics are becoming more widely used in areas
where variables tend to be correlated.
Locus of control theory. The concept of locus of
control originated in Rotter's (Rotter, Chance &
Phares, 1972) Social Learning Theory. Locus of
control refers to a generalized expectancy pertaining
to the connection between personal characteristics and
or actions and experienced outcomes. This causal
relationship between what is experienced in a person's
life and what the person believes caused the event
influences the person's behavior. People can
experience outcomes of events as being controlled by
the external environment or by internal
characteristics of the individual. Rotter has labeled
individuals to be externals or internals. A person
who has an external locus of control believes that he
or she has little control over what happens to him or
her. The person's life is controlled by luck, or
other people rather than by effort or ability put
forth by the individual. An individual with an
internal locus of control has been described as being
able to make autonomous decisions (Sherman, 1973) and
having a sense of well being (Lefcourt, 1982).
40
The research in the area of locus of control has
emphasized a multidimensionality to the concept. This
multidimensionality is shown through the need to
tailor the locus of control test instrument to the
particular population being studied and to the concern
the population is addressing (Lefcourt, 1982).
Literature which has pertained to locus of
control and ethnicity has been bound in controversy.
Some studies have reported minorities to have external
locus of control which has been empirically linked to
an inferior approach to life (Battle & Rotter, 1963;
Zytkoskee & Strickland, 1971). Yet, often the
experiences in their lives have been such that control
of important aspects of their lives have not been
within their reach because of poverty,
underrepresentation in government, and access to
inferior schools (Covington, 1992). For this reason,
students in the present study are not labeled as
external or internal to explain locus of control,
instead emphasis is placed on whether they believe
effort and responsibility for academic achievement is
due to others or to themselves. Beliefs about effort
and responsibility lead one to the study of
achievement motivation theory.
41
Achievement motivation theory. Achievement motivation
has often been linked to success in school because it
deals with the "why" of behavior. Why does one
student choose to strive for the A in a class while
others exert no effort to achieve an A? There are two
different conceptualizations of achievement
motivation. One perspective views achievement
motivation as a learned drive, whereas the other views
motivation as a goal or incentive which draws a
student, rather than drives a student.
Atkinson (1957, 1964, 1987) developed a
sophisticated theory regarding achievement motivation
as a learned drive. This theory holds that human
achievement is the result of conflict on an emotional
level. The conflict is between the striving for
success with the hope of experiencing pride and the
fear of failure accompanied by experiencing shame or
humiliation. A person has a need to succeed, but
often paired with that is the fear that one will not
be successful and will be labeled a failure.
Achievement motivation as explained by Atkinson is a
complex interaction of multiple motives which
influence a student's behavior in a given achievement
situation. He has concluded that achievement behavior
is influenced by motivation, approach vs. avoidance,
probability of success, and its incentive or
42
attractiveness. This theory has been labeled the
"Expectancy X Value Theory."
A student not only will be influenced by the
striving for success or fear of failure, but also will
take into consideration the probability of success and
the value of the possible success. Therefore,
achievement motivation is a complex interaction of
emotion, incentive, and probability, that can be
influenced by other motives the students may
experience for acceptance and by the need for
socialization.
Over the last few decades this achievement
motivation theory has been challenged by cognitive
attribution theory and self-worth theory. In the
1970's cognitive attribution theorists, led by Weiner
(1974, 1990), reinterpreted Atkinson's (1964,
1987,theory. Cognitive attribution theorists posed
the belief that achievement is affected more by
cognition than by emotions. Weiner proposed that how
students perceive the causes of their prior successes
and failures is a more likely determinant of whether
or not they will either choose to work on, or persist
in a certain task. Attributional theory holds that a
person's belief about a situation affects the outcome
of the situation. If students believe their past
successes are due to high ability, they are more
43
likely to undertake similar challenges in the future
because they believe they can succeed again. By
contrast, if students believe their prior successes
are due to luck, they are less likely to try again
because they do not attribute their success to an
internal characteristic such as high ability or
effort.
Weiner proposed four major causes of academic
achievement: (a) ability, (b) effort, (c) task
difficulty or ease, and (d) luck. These four causes
can be classified along three dimensions. The first
dimension is locus of causality which specifies the
cause of the event to be internal (within the person)
or external (outside the person). Ability and effort
are considered to be internal characteristics whereas
task difficulty and luck are considered to be external
characteristics.
The second dimension is that of stability which
classifies the four causes to be stable or transient.
Task difficulty and ability are seen to be stable
while effort and luck are judged to be transient or
changeable. The third dimension reflects
controllability. This dimension helps to
differentiate causes that are identical in the two
dimension schema (locus of causality and stability)
but are treated differently in the real world. This
44
difference can be seen in considering effort and
illness. These events are both classified as internal
and unstable. However, if one fails because one is
ill one is treated differently, as it is perceived one
does not have control over illness. Although three of
the four causes (excluding luck) have been shown
likely to affect academic achievement, the one which
may be most useful to educators is effort.
Attributional theory places great emphasis on
effort in achievement. If a student attributes his or
her success or failure to effort the student is more
likely to try harder in the future to succeed. If the
student fails, the student can try harder in the
future and do better. Also, it has been noted that
teachers often reward effort. Teachers are more
likely to give a student a higher grade if they
perceive the student put forth effort to succeed.
Generally, the theory suggests that students who
attribute their failures of achievement to low effort
- - which can be viewed as an unstable, internal locus
of control - - will be more willing to spend more
effort in the future to try to succeed. Attributional
theory also postulates that it is not failure which
causes students to have low self-concepts but rather
to the factor or factors to which they attribute the
cause of the failure. For example, if students
45
attribute the failure to low ability - - an internal,
stable characteristic - - they are more likely to let
failure influence their academic self-concept.
Some of the questions which have been asked
regarding why some students are willing to put forth
effort and others are not even if they have failed in
the past can be answered by Covington's self-worth
theory (1984). Achievement and the notion of ability
have been shown to influence a student's concept of
self. Students often equate high achievement with
high ability. Self-worth theory contends that in
order to protect a student's self-worth it is often
better not to put effort into school. If students put
forth effort and fail at school, that failure is a
threat to their self-worth. Self-worth theory assumes
the search for self-acceptance is the highest human
priority, and in school, self-acceptance is linked to
students' ability to achieve competitively.
For students of color, the dilemma of school
achievement is even more confusing. Minority students
must juggle the problems of ability, the amount of
effort to apply in academic endeavors, and the
decision to accept or reject the white person's
definition of success. The American educational
system promotes competition and the individual effort
of students to be the best that they can be.
46
By contrast, it has been suggested that the
African-American and Hispanic cultures promote group
effort and cooperation (Fordham & Ogbu, 1986) rather
than individual competition. Many minority students
are caught in the struggle between academic excellence
and ridicule from family and friends. These students
are called names because they have bought into the
values of the white dominant culture (Covington,
1992) .
It is not surprising that many minority students
decide to give up academic endeavors even before they
reach college age. It is unknown how this added
pressure affects the perception held by these students
regarding their self-worth, but one can imagine the
confusion for these students. More research needs to
be completed to ascertain how minority students assess
their academic self-concept and whether scores on
tests which measure academic self-concept can be used
to predict which minority students will experience
academic difficulty in college.
Statement of the Problem
Although, the retention rates of minority
students in higher education have been dismal, this
fact has not led to a better understanding of why
students of color are not successful in college. Many
47
students of color experience academic difficulty in
college which cannot be explained by lack of cognitive
ability. Many of these students enter college with
high scholastic aptitude tests scores and high school
grade point averages.
Research in the area of minority retention in
higher education has been lacking. Many researchers
who have been trying to compare minority students to
majority students have explained the differences in
achievement as being due to some type of educational
deficit in the minority population. The academic
achievement of minority students needs to be
researched to understand why they are graduating from
college at a much lower percentage rate than that
realized by majority students. In order to understand
this phenomenon research needs to be directed at those
students most at risk.
As already indicated, academic self-concept and
achievement motivation have been linked to academic
success. Similarly, academic self-concept and social
self-concept have also been linked to academic
success, but their relationship has been found to be
less clear. It is important to understand how all of
these constructs and achievement motivation are linked
to academic success. This understanding may help
educators predict which minority students will
48
experience academic difficulty during their freshman
year.
Prediction models which can be developed through
research may help educators address the issue of
retention rates for minority students. In order to
develop these prediction models research needs to be
completed which focuses directly on minority students.
The present study uses Tinto's (1987) theoretical
model of retention and the Marsh-Shavelson (Marsh,
1984, 1986; Shavelson & Bolus, 1982) model of self-
concept to try to understand minority academic
achievement in higher education at a predominately
white urban university.
Research Questions
1. What was the reliability of scores on the
measurement instruments: Dimensions of Self-Concept,
Form H (DOSC-H) (Michael et al, 1989), Self-
Descriptive Questionnaire III (SDQ III) (Marsh et al,
1983), and the Intellectual Achievement Responsibility
(IAR) (Crandal et al, 1965) used in the present study
with the sample of university minority students?
2. To what extent, if any, were established cognitive
measures - - high school GPA, and scores on the Verbal
and Mathematics subtest of the College Examination
Board Scholastic Aptitude Test (SAT) - - able to
predict students into the correct classification of
49
subgroups labeled successful, marginally successful,
or unsuccessful during their freshman year?
3. To what extent, if any, were scores in each of
these affective measures — academic self-concept as
represented by the Dimension of Self-Concept (DOSC
Form H) (Michael, Smith, & Michael, 1989) and by the
Self-Descriptive Questionnaire III (SDQ III) (Marsh,
Relich, & Smith, 1983) and attributions of effort as
indicated by the Intellectual Achievement
Responsibility (IAR) (Crandal, Katkovsky, & Crandall,
1965) - - valid in the prediction of the placement of
students in subgroup classifications labeled
successful, marginally successful, or unsuccessful
during their freshman year?
4. To what extent, if any, were scores on a
combination of affective and cognitive measures able
to accurately predict the placement of students into
subgroups labeled successful, marginally successful,
or unsuccessful during their freshman year?
5. To what extent, if any, were scores on the DOSC
Form H or the SDQ III valid in the prediction of the
placement of students into subgroups labeled
successful, marginally successful, or unsuccessful
during their freshman year? If only one test was used
by educators, which one would be a better predictor of
50
which students would experience academic difficulty
during their freshman year.
51
Chapter II
Method
This study used a correlational research design
which utilized discriminant function analysis in order
to obtain a prediction equation which could be used to
forecast which students were at risk of experiencing
academic difficulty during their freshman year in
college. This research study focused on minority
(African-American and Hispanic) students because the
retention rate for these students in higher education
is lower than that for other ethnic groups.
This chapter includes information about the (a)
subjects used in the study, (b) details regarding each
of the self-administered questionnaires, (c)
procedures used during the study, (d) type of data
analysis used to analyze the data, (e) the limitations
of the study.
Sample
The subjects in this study were 87 minority
students who were admitted to the University of
Southern California for the fall of 1993. The sample
comprised 53 Hispanic and 34 African-American students
of whom 57 were male and 30 were female. All subjects
were full-time students registered for at least 12
units. No transfer students were used in this study.
Subjects were selected from two programs on campus
52
which had a high number of minority students involved
in them: the Minority Engineering Program and the
Undergraduate Access Program.
The Minority Engineering Program is one which
provides academic, emotional, and financial support to
minority engineering students at the University of
Southern California. The program is available to all
students in engineering who are African-American or
Hispanic.
The Undergraduate Access Program is a program
which provides academic and emotional support to
students who have been admitted to the University of
Southern California on a special admission provision.
Although the students did not meet the normal
admissions requirements, they were seen to have
special abilities and potential.
Instrumentation
There were three self-report questionnaires used
in this study: the Dimensions of Self-Concept Form H
for college students (DOSC-H) (Michael, Smith, &
Michael, 1989), the Self Description Questionnaire III
(SDQ III) (Marsh, Smith, & Barnes, 1983) and the
Intellectual Achievement Responsibility Questionnaire
(IAR) (Crandall, Katkovsky, & Crandall, 1965) Short
Form for students from 6th to 12th grade. The first
two questionnaires were designed to measure academic
53
self-concept and the third questionnaire to measure
academic locus of control. All three questionnaires
have been validated on numerous populations of
students although the bulk of the research in this
area has used elementary and secondary students.
Dimensions of Self-Conceot Form H
The DOSC-H is an 80-item questionnaire which is
answered on a likert type scale (never, seldom, about
half the time, very often, always). The questionnaire
was specifically developed to measure non-cognitive
factors which are associated with academic self-
concept. The DOSC was intended to be used to identify
students having perceptions about themselves which
could lead to low academic motivation and achievement.
The questionnaire has five factor subscales
hypothesized to measure five constructs of the same
name: Level of Aspiration, Anxiety, Academic Interest
and Satisfaction, Leadership and Initiative, and
Identification vs. Alienation.
The subscale, Level of Aspiration, describes
behaviors which link achievement levels and
perceptions of potential when looking at academic
endeavors. The subscale on Anxiety reflects the
behaviors associated with emotional instability and
with exaggerated concern for fear of failure and loss
of self-esteem linked with academic performance. The
54
subscale, Academic Interest and Satisfaction
represents the joy of learning and pleasure involved
with academic experiences. The subscale, Leadership
and Initiative portrays the behaviors associated with
starting projects and carrying them through to
completion accompanied by the enjoyment of being
identified as someone who demonstrates mastery of
knowledge. The subscale, Identification vs.
Alienation measures the perceptions and behaviors
associated with feeling accepted into the community in
contrast to feeling isolated and left out of the
community. Each factor has 16 item statements which
contribute to its score. The range in reliability of
scores in previous studies varied between .79 and .89.
The standard error of measurement of each subscale
approximates 3.0. Construct validity was evidenced by
factor analytic structure which showed that there were
five distinct factors measured by this instrument and
that the items in each scale were more closely
associated with the one scale to whom they were
intended to belong than to any other subscale.
Criterion-related validity was demonstrated by
concurrent validity relative to cognitive functioning
measured by achievement test scores (Michael, Smith, &
Michael, 1989).
55
Intellectual Achievement Responsibility (IAR)
The IAR Short Form B is a 20-item self-report
questionnaire which was designed to measure a
student's perception of responsibility and effort
needed to achieve in a school setting. The stem of
the question is followed by two possible statements
that would complete the sentence. The student must
choose one of the two statements to complete the
sentence. The IAR is scored in the internal
direction. Three scores are generated, the I+, I-,
and the total score I. The total score I is obtained
by adding the 1+ to the I-. Test-retest reliability
of scores for a sample of 70 students in grade 9 was
reported to be .65 (Crandall, Katkovsky, & Crandall,
1965).
Self Description Questionnaire III (SDQ III)
The SDQ III is a 140-item self-concept
questionnaire designed to measure the
multidimensionality of self-concept. There are four
subscales of academic self-concept, 8 subscales of
nonacademic self-concept, and a general self-concept
subscale. The 13 self-concept subscales are as
follows: Mathematics, Verbal, Academic, Problem
solving, Physical Ability, Appearance, Same Sex Peers,
Opposite Sex Peers, Parents, Spiritual/Religion,
56
Honesty, Emotional, General Esteem. Two subscales
were omitted for this study because they were not seen
as pertinent. The two subscales were those labeled
Honesty and Spiritual/Religion.
The questionnaire allows the students to respond
to statements with one of 8 responses: Definitely
False, False, Mostly False, More False Than True, More
True Than False, Mostly True, True, Definitely True.
Each scale has 10 to 12 items of which half is
negatively worded. The reliability of scores was
reported to range between .76 and .95 with a median
score of .89 in previous studies (Marsh, 1984, 1986,
1987, 1990) . The standard error of measurement was
shown to be about one point above or one point below
the observed score.
Cognitive Variables
Verbal and Mathematics scores of the College
Board Scholastic Aptitude Test (SAT-V and SAT-M)
(Educational Testing Service, 1948-1994) and high
school grade point average (HSGPA) were down-loaded
off the Student Information System by the students'
identification number. These scores were used in the
discriminant function analysis to ascertain their
predictive power when using college GPA and number of
units completed as the outcome measures. Prior
research in this area has suggested that these
57
cognitive measures lack the ability to predict
academic achievement with minority students (Duran,
1986).
Procedure
All students who completed the questionnaires
were participants in either the Minority Engineering
Program or the Undergraduate Access Program at the
University of Southern California in the fall of 1993.
The procedures used in this study were approved by the
Institutional Review Board of the University of
Southern California. The students were administered
the DOSC-H, IAR, and SDQ III in a group setting. The
testing period lasted for about 1 hour and 2 0 minutes
or until all students were finished. Students who
were not at the group administration were given the
packet of questionnaires to complete at home and
return to the program administrator.
The students were then tracked throughout the
school year and at the end of the year, their total
number of units completed and college GPA were
recorded. These outcome measures were matched to the
student1s record through the use of student
identification numbers. The student identification
numbers were then changed to a four digit case number
to provide confidentiality to the student's
information. The student's records contained the
58
following information: Gender, Ethnicity, SAT scores,
High School GPA, Scale scores on the DOSC-H (five
subscales), Scale scores on the SDQ III (ten
subscales), and I+, and I- scores on the IAR.
Data Analysis
This study used discriminant function analysis
techniques. Prediction equations were generated from
the discriminant function analysis to help forecast
which type of student would be at risk for academic
difficulty during the freshman year.
The discriminant analysis generated prediction
equations which placed students into groups based on
certain scores on the criterion variables. The
criterion variables were combined into linear form to
produce a functional equation which used mean values
of the criterion variables to select students into
different groups which were as statistically distinct
as possible. The criterion variables were given
weights as to the amount of variance they provided to
the groupings. The discriminant analysis implements a
step-wise algorithm and the F-test criterion to select
variables for inclusion to the model. The first
variable in the model provided the most variance to
the equation.
The outcome variables for this study were college
GPA and number of units completed. The students were
59
grouped by college GPA into three groups:
Unsuccessful Group (Group 0) earned below a 2.0 GPA,
Marginally Successful Group (Group 1) earned a 2.0
through 2.59 GPA, and Successful Group (Group 2)
earned a 2.6 or above GPA. At the end of the
analysis, a classification matrix was generated to
evaluate the number of students correctly placed in
each group.
Limitations to the Study
There were several limitations to this type of
study. First, the subjects used in the study were not
randomly selected, but were participants in programs
which provided them with academic, social, and
emotional support during their freshman year in
college. But it should be noted that even with this
type of supportive program, 23 students received below
a 2.0 cumulative GPA at the end of their freshman year
in college.
A second limitation of this study was that all
subjects attended the same large, private,
predominately white, four year university. This
constraint would limit the generalizability of results
to other settings, although the research model
employed would provide a means by which a university
looking at the unique characteristics of its
population of students could ascertain those variables
60
most important for its model of classifying students
into various criterion subgroups.
A third limitation to the study was the use of
three self-report instruments to measure the non-
cognitive variables in the study. Because the
constructs of academic self-concept and academic locus
of control are not only psychological in nature and
but are phenomena which are not easily identified,
most researchers have decided to utilize self-report
instruments because there appear to be no more valid
measures to use at this time.
A fourth limitation to this study was the size of
the sample, as it was not large enough to perform
cross-validation procedures on the four discriminant
analyses performed. If the sample was not large
enough to complete a cross-validation procedure, the
original prediction accuracy of the classification
matrix of the discriminant function could be biased in
a positive direction.
61
Chapter III
Results
In this chapter the results of the analysis
between cognitive and affective variables used to
predict academic achievement with freshman minority
students at a predominately white private university
are presented. The study sought to explain and
predict which students are at high risk for
experiencing academic difficulty during their freshman
year using cognitive and affective variables
separately and together. The chapter is divided into
five parts, each of which addressed a specific
research question.
Findings Of The Study
Research Question #1
#1. What was the reliability of scores on the
measurement instruments: Dimensions of Self-Concept,
Form H (DOSC-H) (Michael et al, 1989), Self-
Descriptive Questionnaire III (SDQ III) (Marsh et al,
1983) , and the Intellectual Achievement Responsibility
(IAR) (Crandall et al, 1965) used in the present study
with the sample of university minority students?
One of the major problems with research in the
area of student retention and academic self-concept
has been use of measures yielding scores of
questionable reliability. To establish the
62
reliability coefficients for scores on each scale on
each questionnaire the reliability procedure provided
by SPSS (SPSS, 1992) was performed on all three test
instruments, the DOSC-H, SDQ III, and the IAR. The
results are shown in Table 1.
Table 1
Descriptive statistics for the scales on the
DOSC-H, SDQ III, and IAR
Scales Mean Standard
Deviation
Alpha
DOSC-H
ASP 60.427 10.4554 .8917
ANX 41.9775 10.2702 .8764
AIAS 47.7978 8.5867 .8481
LAI 47.4045 9.1437 .8502
IA 53.8764 6.9753 .8304
SDQ III
MATH 56.7528 16.0788 .9347
VERBAL 51.5393 11.7212 .8109
PROB 55.3708 10.9528 .8123
ACAD 58.0674 10.7003 .8458
SAMESEX 59.7640 9.0114 .7458
OPPSEX 58.6629 13.1615 .8817
PARENT 58.6854 13.3012 .8713
EMOT 53.2135 11.8172 .8207
GENSELF 79.3371 15.1117 .9290
IAR
INT + 7.2237 1.7820 .3242
INT - 6.5294 2.1077 .5796
All scales on the DOSC-H and the SDQ III reached
a satisfactory level of internal reliability. The two
IAR scales did not reach a satisfactory level of
internal reliability, however, the two scales for the
63
IAR remained in the analysis because no other measure
of responsibility for academic achievement was
administered. For the reason outlined above, the
results concerning these two scales on the IAR may
need to be interpreted with caution.
Research Question #2
#2. To what extent, if any, were established
cognitive measures - - high school GPA, and scores on
the Verbal and Mathematics subtest of the College
Examination Board Scholastic Aptitude Test (SAT) - -
able to predict students into the correct
classification of subgroups labeled successful,
marginally successful, or unsuccessful during their
freshman year?
In this section the three cognitive variables
were entered into a stepwise discriminant function
analysis to ascertain whether cognitive variables
alone could successfully predict which freshmen would
experience academic difficulty during their first year
of college. Academic difficulty was defined as
earning a cumulative GPA below 2.0.at the end of the
freshman year. Table 2 gives the groups means and
standard deviations for the cognitive variables SAT
Mathematics, SAT Verbal, and High School GPA. As only
85 of the 87 students had the complete cognitive data
64
set for this analysis, the sample size for this
research question was only 85 students.
Table 2 shows a surprising trend for the SAT
Mathematics and SAT Verbal scores for those students
who were unsuccessful (Group 0) during their first
year in college.
Table 2
Group descriptive statistics for the three group
discriminant analysis using cognitive variables
Group Means HSGPA SATMATH SATVERB
Group 0 3 .11 537.27 438.18
Group 1 3 . 04 508.21 422.85
Group 2 3.36 532.85 432.85
Standard
Deviations HSGPA SATMATH SATVERB
Group 0 .49 108.67 80.45
Group 1 .48 103.42 72.51
Group 2 .50 101.53 103.99
Many individuals would expect that high SAT
Mathematics and Verbal scores correspond to an
outcome of excelling in college. Table 2 shows that
this expectation may not be true for minority students
at a predominately white university. The unsuccessful
students (Group 0) had the highest average scores for
both SAT Mathematics and SAT Verbal. The marginally
successful group (Group 1), those students who had a
65
cumulative GPA of 2.0 to 2.59, scored the lowest on
both SAT Mathematics and SAT Verbal scores.
Table 3 gives the results of the Wilks' lambda
and the univariate F ratios for the cognitive
predictor variables. This table shows which variables
if used alone would show a significant difference
between groups. The only cognitive variable which
reached significance was the High School GPA. The
Scheffe procedure identified that the significant
differences lay between Group 1 (marginally successful
students) and Group 2 (successful students) on High
School GPA.
Table 3
Wilks' Lambda and Univariate F ratios for cognitive
variables
Variable Wilks'
Lambda
Univariate
F ratio Sig
Between
Groups
HSGPA .9137 3 .869 .0248 1 & 2
SATMATH .9848 . 614 .5434
SATVERB .9926 . 197 .8211
The stepwise discriminant function analysis
selects the variable with the most discriminant power
and enters that variable into the equation. The
results are listed in Table 4. Only High School GPA
had enough variance to enter the discriminant
equation. Both SAT Mathematics and SAT Verbal tests
66
did not meet the minimum requirement to enter the
equation. Table 4 shows the results of the
discriminant function analysis starting with the
pooled within-group correlations between
discriminating variables and canonical discriminant
function. No rotation was completed because only one
variable was entered into the equation and only one
function could be generated because SAT Mathematics
and SAT Verbal test scores did not explain enough of
the variance at the .05 level of significance to be
included in this discriminant function analysis.
Table 4
Canonical discriminant functions of cognitive
variables
Canon After Wilks'
Fen Eigen Corr Fen Lambda Chi Sig
1 .0944 .2936 0 .9138 7.39 .024
Pooled within-group correlations between
discriminating variables and canonical
discriminant functions
Function 1
HSGPA 1.000
SATMATH 0.481
SATVERB 0.277
Canonical discriminant function evaluated at
group means (group centroids)
Group Function 1
Group 0 -0.17040
Group 1 -0.30985
Group 2 0.35499
67
Table 4 summarizes the information provided by
the discriminant function. The canonical discriminant
function using just High School GPA gave one
significant discriminant function. The group
centroids show the discrimination is between the
unsuccessful and marginally successful group versus
the successful group. In other words, those students
who received below a 2.59 cumulative GPA and those
students who earned a 2.6 cumulative GPA or above.
The best way to discern whether or not the
discriminant function can successfully discriminate
between the three groups is by a classification matrix
of the variables. The classification matrix shows
what percentage of the students were correctly
classified into the intended group. The results are
shown in Table 5.
Table 5
Classification results for cognitive variables
Actual group
No. Of
Cases
Predicted grp membership
0 1 2
0 22 4 9 9
18.2% 40.9% 40.9%
1 28 2 17 9
7.1% 60.7% 32.1%
2 35 1 8 9
2.9 % 31.4% 65.7%
% Of "grouped" cases correctly classified: 51.76%
Cases by chance: 3 3%
68
Table 5 shows the results of the classification
matrix using cognitive variables. The discriminant
function analysis using only cognitive variables
predicted only 18.2% of the students who would
experience academic difficulty during their freshman
year. Being able to predict correctly those students
who would fall in the Unsuccessful Group (Group 0)
category is the most important concern. The purpose
of an early warning system would be to identify
correctly those students who would be on academic
probation after their first year in college. The
cognitive factors alone did not successfully predict
those students who would have academic difficulty
during their freshman year.
Research Question #3
#3. To what extent, if any, were scores in each of
these affective measures - - academic self-concept as
represented by the Dimension of Self-Concept (DOSC-H)
(Michael et al, 1989), and by the Self-Descriptive
Questionnaire III (SDQ III) (Marsh et al, 1983), and
attributions of effort as indicated by the
Intellectual Achievement Responsibility (IAR)
(Crandall, Katkovsky, & Crandall, 1965) valid in the
prediction of the placement of students in subgroup
classifications labeled successful, marginally
69
successful, or unsuccessful during their freshman
year?
The third research question utilizes affective
variables to predict academic success and failure in
freshman minority students at a predominately white,
urban university. The affective variables fall into
three different categories: (a) academic self-concept
scales, (b) social self-concept scales, and (c)
internal responsibility for positive and negative
events.
The affective variables were entered into a
stepwise discriminant analysis to ascertain which
variables added the largest proportion of variance to
the function equations. Table 6 shows the descriptive
statistics for the affective variables used in the
analysis.
All 87 students were used in this analysis as all
the students had completed the questionnaire measuring
the affective variables. The means and standard
deviations show that the three groups scored similarly
on several of the affective scales. The Unsuccessful
students scored highest on the scales measuring
Internal responsibility for positive events (1+), Same
Sex Peers, Verbal Self-Concept, Opposite Sex Peers,
and Problem Solving Self Concept subscales. The
70
Successful students scored highest on the subscales
measuring Academic Self-Concept, Math Self-Concept,
Level of Aspiration, and Self-Concept regarding
Parental Relationship. The Marginally Successful
students scored highest on the subscales measuring
Identification vs. Alienation and Internal
Responsibility regarding Negative Events (I—).
Table 6
Group descriptive statistics for the three group
discriminant analysis using affective variables
Group Means
GPA Group IAR+ ACADEM SAMESEX VERBAL
Group 0 7.34 57.34 61.39 55.86
Group 1 7.34 58.03 60.20 51.13
Group 2 7. 05 60.00 59.17 50.17
GPA Group MATH PARENT OPSEX PROBSOLV
Group 0 59.39 59.30 63.95 58.82
Group 1 52.55 56.55 59.79 55.27
Group 2 59.74 60.37 54.22 53.77
GPA Group IA ASP IAR-
Group 52.08 58.60 6.73
Group 1 55.34 60.24 6.89
Group 2 54.11 61.60 6.11
Standard Deviations
GPA Group IAR+ ACADEM SAMESEX VERBAL
Group 0 1.94 13.53 7.53 11.59
Group 1 1.96 8.02 8.23 12.00
Group 2 1.51 9.89 9.32 11.53
GPA Group MATH PARENT OPSEX PROBSOLV
Group 0 14.74 11.72 13.06 11.76
Group 1 19.58 12.99 11.53 10.40
Group 2 12.53 14.94 13.66 10.69
GPA Group IA ASP IAR-
Group 0 7.48 10.93 2.07
Group 1 5.84 9.31 1.63
Group 2 7.09 11.29 2.38
71
Table 7 indicates the Wilks' Lambda and the
Univariate F ratios for the affective predictor
variables. Only the variable Opposite Sex Peers
revealed a significant difference between the groups.
The Scheffe procedure showed the significant
differences were found between Group 0 (Unsuccessful
students) and Group 2 (Successful students). The
results of the analysis using affective variables are
shown in Table 7.
Table 7
Wilks' Lambda and univariate F ratios for affective
variables
Variable Wilks' Univariate Between
Lambda F ratio Sig Groups
IAR+ .99356 .2722 .7624
ACADEM .98774 .5214 .5956
SAMESEX .98885 .4737 .6243
VERBAL .95977 1.761 . 1782
MATH .95589 1.938 . 1503
PARENT .98468 .6536 .5228
OPSEX .90988 4.160 .0189 0 & 2
PROBSOLV .96520 1.514 . 2259
IA .96606 1.475 .2346
ASP .98689 .5579 . 5745
IAR- .97071 1.267 .2870
Table 8 provides a summary of the discriminant
function indicating which variables entered the
equation first. Only six variables remained at the
end of the analysis. Those variables included:
Opposite Sex Peers, Math Self-Concept, Identification
vs. Alienation, Self-Concept regarding Parental
72
Relationship, Internal Responsibility regarding
Negative Events, and Problem Solving Self-Concept.
Level of Aspiration entered the analysis but was
removed when the variance it supplied was shared by
other variables.
Table 8
Summary table for affective variables in discriminant
analysis
Step
Action
Ent Rem
Wilks'
Lambda Sig
Between
Groups
1 OPSEX .90988 .0189 0 1
2 MATH .86505 .0166 0 1
3 IA .78741 .0030 1 2
4 PARENT .76172 . 0042 1 2
5 ASP .74217 .0065 1 2
6 IAR- .71861 .0080 1 2
7 ASP .73625 .0052 1 2
8 PROBSOLV .71591 .0072 1 2
Table 9 displays the canonical discriminant
functions using affective variables, the rotated
standardized discriminant function coefficients, and
the correlations between the rotated canonical
discriminant functions and discriminating variables.
As is evident from Table 9, both canonical
discriminant functions were significant. The
variables used in the rotated standardized
discriminant function coefficients in Function One
comprise: Opposite Sex Peers, Problem Solving Self-
73
Concept, and Internal Responsibility regarding
Negative Events. In Function Two the variables
consisted of: Identification vs. Alienation, Math
Self-Concept, and Self-Concept regarding Parental
Relationship.
Table 9
Canonical discriminant function of affective variables
Fen Eigen
Canon
Corr
Aft
Fen
Wilks 1
Lambda Chi Sig
1 .2195 .4242 0 .7159 27.2 .007
2 .1454 .3563 1 .8730 11.0 .050
Rotated standardized discriminant
coefficients
function
Function 1 Function 2
OPSEX .71935 .22576
PROBSOLV .51322 .01945
IAR- .48751 -0.01597
IA -.37788 -1.02630
MATH -.32691 .72168
PARENT -.27950 .52372
Correlations between rotated canonical discriminant
functions and discriminating variables
Function 1 Function 2
MATH -0.22459 0.51609
PARENT -0.17948 0.26803
OPSEX 0.67035 0.09364
PROBSOLV 0.35144 0.20625
IA -0.10100 -0.44704
IAR- 0.36035 -0.20207
IAR+ 0.18738 0.10783
ACADEM -0.03031 0.03344
SAMESEX 0.31045 0.05238
VERBAL 0.21445 -0.04046
ASP -0.15645 -0.37854
74
Table 10 shows the group centroids for the
discriminant analysis when using affective variables.
The group centroids revealed that Function One was
discriminating between the Unsuccessful students and
Marginally Successful students versus the Successful
Students. Function Two was discriminating between the
Unsuccessful and Successful students versus those
students who were Marginally Successful.
Canonical discriminant function evaluated at group
means (group centroids)
Table 11 sets forth the classification matrix
using affective variables. The percent of the
"grouped" cases correctly classified was 57.4. The
prime target group, (the Unsuccessful Students), was
correctly classified at 65.2%. This outcome was an
improvement over the analysis using cognitive
variables alone which correctly classified the target
group at 18.2%. Generally this analysis indicated
that affective variables were more accrurate
predictors of which students would experience academic
difficulty during their freshman year.
Table 10
Group
0
1
2
Function 1
0.54640
0.20081
-0.52545
Function 2
0.50066
-0.49053
0.07743
75
Table 11
Classification results for affective variables
No. of Predicted grp membership
Actual Group cases 0 1 2
Group 0 23 15
65.2%
5
21.7%
3
13.0%
Group 1 29 5
17.2%
14
48.3%
10
34.5%
Group 2 35 7
20.0%
7
20.0%
21
60.0%
% Of "grouped" cases correctly classified: 57.4%
Cases by chance: 33%
The affective variables were able to predict almost
50% more students into the target group, Group 0
(Unsuccessful Students) than the cognitive variables.
Research Question #4
#4. To what extent, if any, were scores on a
combination of affective and cognitive measures able
to accurately predict the placement of students into
subgroups labeled successful, marginally successful,
or unsuccessful during their freshman year?
The fourth research question utilizes both
cognitive and affective variables to predict academic
success and failure in freshmen minority students at a
private, urban, predominately white university. Both
cognitive and affective variables were used in order
to hopefully create an optimal discriminant equation
which could maximally predict which students would
76
experience academic difficulty during their freshman
year. The variables were entered into the stepwise
discriminant analysis. Only 85 of the 87 students had
a complete data set when cognitive and affective
variables were employed. Therefore, only 85 students
were used in this analysis.
The means and standard deviations for the
cognitive and affective variables displayed a
surprising trend for the Unsuccessful Group (Group 0).
The trend revealed that this group scored the highest
on a number of cognitive and affective variables. The
Unsuccessful Group (Group 0) scored the highest
average on six of the twelve predictor variables:
Verbal Self-Concept, Math Self-Concept, Same Sex
Peers, Opposite Sex Peers, SAT Mathematics and SAT
Verbal scores. The group of students who had the best
academic self-concept of themselves achieved the
lowest college grade point average. Also, this group
of students scored the highest on the SAT Mathematics
and Verbal. Table 12 shows the results for the
descriptive statistics using affective and cognitive
variables together. The higher the score on the self-
concept variable the more positive the students felt
about themselves.
77
Table 12
Group descriptive statistics for the three group
discriminant analysis using cognitive and
affective variables
Group Means
GPA Group
Group 0
Group 1
Group 2
IAR+
7.27
7.39
7.05
ACADEM
58.09
57.78
60.00
VERBAL
56.22
50.85
50.17
MATH
59.90
51.85
59.74
GPA Group SAMESEX OPSEX PARENT IA
Group 0
Group 1
Group 2
61.54
60.50
59.17
63.36
59.60
54.22
59.09
56.64
60.37
52.36
55. 39
54.11
GPA Group
Group
Group
Group 2
ASP
59.45
60.03
61.60
HSGPA
3.10
3.03
3.36
SATVERB
438.18
422.85
432.85
SATMATH
537.27
508.21
532.85
GPA Group PR0BS01V IAR-
Group 0 58.50 6.72
Group 1 55.03 6.89
Group 2 53.77 6.11
Standard Deviations
GPA Group IAR+ ACADEM VERBAL MATH
Group 0 1.95 13.36 11.73 14.87
Group 1.98 8.05 12.12 19.58
Group 2 1.51 9.89 11.53 12.53
GPA Group SAMESEX OPSEX PARENT IA
Group 0 7.67 13.04 11.95 7.54
Group 1 8.23 11.70 13.22 5.94
Group 2 9. 32 13.66 14.94 7.09
GPA Group ASP HSGPA SATVERB SATMATH
Group 0 10.39 0.48 80.45 108.67
Group 1 9.41 0.48 72.51 103.42
Group 2 11.29 0.49 103.99 101.53
GPA Group PROBSOLV IAR-
Group 0 11.93 7.54
Group 1 10.51 5.94
Group 2 10. 69 7.09
78
Table 13 gives the Wilks' Lambda and Univariate F
ratios for the predictor variables.
Table 13
Wilks' lambda and univariate F ratios for cognitive
and affective variables
Variable Wilks' Univariate Sig Between
Lambda F ratio Groups
IAR+ .9931 0.282 0.754
ACADEM .9898 0.418 0.659
VERBAL .9542 1.967 0.146
MATH .944 2 . 393 0.097
SAMESEX .9870 0.540 0.584
OPSEX .9195 3.586 0. 032 0 & 2
PARENT .9859 0.585 0.559
IA .9715 1.201 0.306
ASP .9920 0.3297 0.720
HSGPA .9137 3 .869 0.024 1 & 2
SATVERB .9952 0.1976 0.821
SATMATH .9852 0.6144 0.543
PROBSOLV .9697 1.279 0.283
IAR- .9716 1.198 0. 307
Only Opposite Sex Peers and High School GPA afforded
significant differences between the groups. The
Scheffe procedure was performed to ascertain where the
differences between the groups lay. The significant
difference for Opposite Sex Peers lay between the
Unsuccessful Students and the Successful Students
subgroups. The significant difference for High School
GPA fell between the Marginally Successful Students
and the Successful Students subgroups.
Table 14 presents the summary of the discriminant
function analysis indicating which variables entered
79
the equation first and which variables did not enter
the equation. As can be seen in the table the first
variable to enter the equation was Opposite Sex Peers.
The next variable to enter the equation was
Mathematics Self-Concept and then Identification vs.
Alienation.
Table 14
Summary table for affective and cognitive variables in
discriminant analysis
Step
Action
Ent Rem
Wilks'
Lambda Sig
Between
Groups
1 OPSEX .91956 .0321 0 1
2 MATH .86405 .0180 0 1
3 IA .79156 .0043 1 2
4 HSGPA .75020 .0032 1 2
5 PARENT .72833 .0047 1 2
6 IARNEG .70610 .0062 1 2
7 ASP .68031 1.0460 1 2
8 VERBAL .65453 1.0466 1 2
High School GPA was the fourth variable to enter
the equation. Self-Concept regarding Parental
Relationship, Internal Responsibility regarding
Negative Events, Level of Aspirations, and Verbal
Self-Concept were the last four variables to enter the
equation. The variables which did not enter the
equation were Academic Self-Concept, SAT Mathematics,
SAT Verbal, Same Sex Peers, Problem Solving Self-
Concept, and Internal Responsibility for Positive
Events.
80
Table 15 sets forth the canonical discriminant
functions for the cognitive and affective variables,
the rotated standardized discriminant function
coefficients, the correlations between rotated
canonical discriminant functions and discriminating
variables. Both canonical discriminant functions were
significant.
Table 15 revealed the coefficients and loadings
which could be used to help explain the differences
between the groups used in the analysis. The rotated
standardized coefficients indicate Function One to be
using the variables Opposite Sex Peers, High School
GPA, and Internal Responsibility regarding Negative
Events.
Function Two is using the variables
Identification vs. Alienation, Math Self-Concept,
Level of Aspirations, Self-concept regarding Parental
Relationship, and Verbal Self-Concept. Table 15 also
indicates the variables which afforded the variance to
the function equations after a varimax rotation was
executed. The correlations between the rotated
discriminant function and the discriminating variables
provide additional information regarding which
variables held the most predictive power.
81
Table 15
Canonical discriminant functions of affective and
cognitive variables
Canon Aft. Wilks'
Fen Eigen Corr Fen Lambda Chi Sig
1 .2768 .4656 0 .6545 33.2 .0068
2 . 1876 .4054 1 .8357 14.09 .0496
Rotated standardized discriminant function
coefficients
FUNCTION 1 FUNCTION 2
OPSEX .65994 .16787
HSGPA -.63248 -.13264
IAR- .41007 .06362
IA -.28352 -1.35245
MATH .17912 .78895
ASP .05354 .62062
PARENT -.28572 .41244
VERBAL .26255 .35868
Correlations between rotated canonical discriminant
functions and discriminating variables
Function 1 Function 2
IAR+ .18192 .09984
VERBAL .30590 .28268
MATH -.23568 .48401
PARENT -.18219 .19828
OPSEX .58444 .06879
HSGPA -.60466 .28774
IA -.0481 -.34543
ACADEM -.11448 .19144
SAMESEX .31060 -.16603
SATMATH -.15267 .33213
SATVERB .07028 .18086
ASP -.18277 .00751
PROBSOLV .27929 .11437
IAR- . 33616 -.16096
Table 16 shows the group centroids for the
discriminant analysis utilizing affective and
cognitive variables.
82
Table 16
Canonical discriminant function evaluated at the
group means (group centroids)
Group Function 1 Function 2
0 0.58787 0.64317
1 0.23383 -0.58545
2 -0.55659 0.06409
The group centroids revealed that Function One
was discriminating between those students who were
Successful and those students who were Marginally
Success and Unsuccessful. Function Two discriminated
between those students who were Unsuccessful and those
who were Marginally Successful and Successful.
Table 17 presents the classification matrix using
both cognitive and affective measures.
Table 17
Classification results for affective and cognitive
variables
No. of Predicted grp membership
Actual group cases 0 1 2
Group 0 22 15
68.2%
3
13.6%
4
18.2%
Group 1 28 6
21.4%
13
46.4%
9
32.1%
Group 2 34 7
20.0%
10
28.6%
18
51.4%
% Of "grouped" cases correctly classified: 54.12%
Cases by chance: 33%
The percent of "grouped" cases correctly
classified was 54.12%. The prime target group, which
83
consisted of those students in the Unsuccessful Group
(Group 0), was the highest group correctly classified
at 68.2%. Although Table 17 indicated a significant
improvement over the percentage of those students who
would have been predicted by chance, there were still
too many students who have been classified in the
wrong group, although the Unsuccessful Group (Group
0), the target group, had the highest percentage
correctly classified. When only the cognitive
variables were used, it was difficult to predict which
students would be unsuccessful. When the affective
and cognitive variables were combined the Hhit rate"
for the Unsuccessful Group increased by 50%.
Generally, the analysis indicated that the
affective measures and cognitive measures combined
together rather than cognitive measures alone were
more valid predictors for those students who have been
unsuccessful during their freshman year. However, the
affective variables alone had a similar "hit rate" for
those students who were unsuccessful during their
first year. The overall classification rate was
similar for all three discriminant analyses.
Research Question #5
#5. To what extent, if any, were scores on the DOSC
Form H or the SDQ III valid in the prediction of the
placement of students into subgroups labeled
84
successful, marginally successful, or unsuccessful
during their freshman year?
The fifth research question regarding which test
instrument yielded a more valid prediction
classification is addressed in this section. The
DOSC-H was designed to help education administrators
predict which students were at risk for academic
difficulty. The SDQ III was designed to measure self-
concept and to show that academic self-concept and
academic achievement are related. The DOSC-H was
tested first. A discriminant function analysis was
executed which yielded the following data which is
displayed in Table 18.
Table 18
Group descriptive statistics for the three group
discriminant analysis using the DOSC-H
Group
Means
IA ASP LAI ANX AIAS
Group 0
Group 1
Group 2
52.08
55. 34
54.11
58. 60
60.24
61.60
47.91
46.79
47.88
40.69
41.00
43.31
46. 60
48.58
48.00
Standard
Deviations IA ASP LAI ANX AIAS
Group 0
Group 1
Group 2
7.48
5.84
7.09
1093
9.31
11.29
9.64
8.45
9.52
9.04
11.21
10.47
9.87
6.56
9.00
Table 18 indicates the five subscales on the
DOSC-H with their means and standard deviations.
85
There did not seem to be much discriminating power
between the groups because the means of the different
scales were so similar.
The Wilks' Lambda and univariate F ratio revealed
no variable on the DOSC-H with a significant F ratio.
The relevant data are displayed in Table 19.
Table 19
Wilks' lambda and univariate F ratio for DOSC-H
Variable Wilks'
lambda
Univariate
F ratio
Sig
IA .9660 1.475 .2346
ASP .9868 .5579 .5745
ANX .9862 .5876 .5579
LAI .9969 .1270 .8809
AIAS .9915 .3569 .7009
Table 20 consists of the summary table, the
rotated standardized discriminant function
coefficients, correlations between rotated
discriminant functions and discriminating variables.
Table 21 show the group centroids. The only variable
from the DOSC-H subscales which entered the equation
was the Identification vs. Alienation subscale, but
this subscale did not generate a significant
discriminant function. No other variable could supply
enough variance to enter the equation when using only
the five scales on the DOSC-H.
86
Table 20
Summary Table for variables in discriminant analysis
for DOSC-H
Step
Action
ent rem
Wilks' Sig
lambda
Between
Groups
1 IA .9660 .234 1 2
Canonical discriminant function for DOSC-H
Fen eigen
Canon After Wilks'
Corr Fen lambda
Chi Sig
1 .0351 .1842 0 .9661 2.90 .23
Pooled within groups correlations between
discriminating variables and canonical discriminant
function
Function 1
IA 1.000
ASP .6740
AIAS .6165
LAI .3802
ANX -0.1596
Table 21 shows the group centroids for the
discriminant analysis using the DOSC-H. There was
only one Function gererated and it discriminated
between the Unsuccessful students and those students
who were in the Marginally Successful and Successful
group.
87
Table 21
Canonical discriminant functions evaluated at
group means (group centroids)
Group Function 1
Group 0 -.2790
Group 1 .1990
Group 2 .0184
The classification results revealed that the
target group, Unsuccessful Group (Group 0), was
correctly predicted at 43.5%. The Marginally
Successful Group (Group 1) was correctly forecasted at
65.5%, but the Successful Group (Group 2) was
correctly predicted at only 11.4%. Overall, the DOSC-
H was not successful at forecasting the students into
the intended groups.
Table 22
Classification results for DOSC-H
Actual group
No. of
casesO
Predicted
1
grp. membersh ip
2
0 23 10
43.5%
11
47.8%
2
8.7%
1 29 8
27.6%
19
65.5%
2
6.9%
2 34 13
37.1%
18
51.4%
4
11.4%
% Of "grouped cases correctly classified: 36.7%
Cases by chance: 33%
88
The SDQ III questionnaire has nine subscales
which were used in this study to help predict which
students would experience academic difficulty during
the freshman year. Table 23 gives the descriptive
statistics for the SDQ III subscales.
Table 23
Grouped descriptive statistics for three group
discriminant analysis using SDQ III
Group
Means ACADEM SAMESEX VERBAL MATH PARENT
Group 0 57.34 61.39 55.86 59.39 59.30
Group 1 58.03 60.20 51.13 52.55 56.55
Group 2 60.00 59.17 50.17 59.74 60.37
Group
Means OPSEX GENSC EMOT PROBSOLV
Group 0 63.95 81.73 57.56 58.82
Group 1 59.79 78.41 54.79 55.27
Group 2 54.22 79.05 49.60 53.77
Standard
Deviations ACADEM SAMESEX VERBAL MATH PARENT
Group 0 13.53 7.53 11.59 14.74 11.72
Group 1 8.02 8.23 12.00 19.58 12.99
Group 2 9.89 9.32 11.73 12.53 14.94
Standard
Deviations OPSEX GENSC EMOT PROBSOLV
Group 0 13.06 14.87 9.69 11.76
Group 1 11.53 16.33 10.54 10.40
Group 2 13.66 14.58 12.06 10.69
Table 23 indicates the variables on the SDQ III
subscales which have the highest mean score for each
group. The Unsuccessful Group (Group 0) which
89
consisted of those students with a cumulative GPA of
less than 2.0 had the highest scores on Same Sex
Peers, Opposite Sex Peers, Verbal Self-Concept,
Emotional Stability, and Problem Solving Self-Concept
measures.
Table 24 shows the Wilks' Lambda and univariate F
ratio. According to the analysis, the only variables
which yielded significance are Opposite Sex Peers and
Emotional Stability. The Scheffe procedure indicated
that the significant differences on the two measures
were both between the Unsuccessful Group and the
Successful Group of students.
Table 24
Wilks' lambda and univariate F ratios for SDQ III
Variable Wilks' Univariate Sig Between
lambda F ratio Groups
ACADEM .9877 .521 .5956
SAMESEX .9888 .473 .6243
VERBAL .9597 1.761 . 1782
MATH .9558 1.938 .1503
PARENT .9846 .653 .5228
OPSEX .9098 4.160 .0252 0 & 2
GENSC .9920 .335 .7162
EMOT .9134 3.982 .0223 0 & 2
PROBSOLV .9652 1.514 .2259
The discriminant analysis used seven variables
the two functions. The three academic self concept
scales were: Mathematics Self-Concept, Academic Self-
concept, and Verbal Self-Concept. They are all listed
in Table 25.
90
Table 25
Summary Table for SDQ III
Step
Action
ent rem
Wilks'
lambda
Sig Between
groups
1 OPSEX .9098 .0189 0 1
2 MATH . 8650 .0166 0 1
3 ACADEM .8250 .0140 1 2
4 EMOT .7658 .0049 0 1
5 VERBAL .7219 .0029 1 2
6 PARENT . 6913 . 0027 0 1
7 GENSC .6719 . 0038 0 1
Canonical discriminant functions of SDQ III
Canon Aft Wilks' Chi Sig
Fen Eigen Corr Fen Lambda
1 .3362 .5016 0 .6719 32.2 .003
2 .1138 .31971 1 .8978 8.7 .189
Rotated standardized discriminant function
coefficients
Function 1 Function 2
EMOT .95343 -.09027
GENSC -.60443 .15879
OPSEX .62399 .13995
PARENT -.42690 .29662
MATH -.22557 1.07672
ACADEM -.09307 -1.10436
VERBAL .25237 .66059
Correlations between rotated canonical discriminant
functions and discriminating variables
Function 1 Function 2
ACADEM -.19314 -.02208
VERBAL .19743 .38697
MATH -.29000 .54308
PARENT -.19919 .27088
OPSEX .49707 .22483
GENSC .02662 .22895
EMOT .51149 .14319
SAMESEX .25794 .21329
PROBSOLV .17186 .26088
91
The four other scales of self-concept were
Emotional Stability, Self-concepts regarding Parents,
Opposite Sex Peers, General Self-concept. As shown in
Table 25 the variables Opposite Sex Peers, Emotional
Stability, General Self-Concept, and Self-Concept
regarding Parental Relationship contributed the most
variance to Function 1 and Mathematics Self-Concept,
Academic Self-Concept, and Verbal Self-Concept afford
the most variance to Function 2. The group centroids
revealed that Function 1 discriminated between the
group of Successful students versus the Unsuccessful
and Marginally Successful students. Function 2
discriminated between the Unsuccessful students versus
the Marginally Successful students and the Successful
students.
The correlations between the rotated discriminant
function and the discriminating variables provided
additional information regarding which variables added
the most variance to the equations. A varimax
rotation was computed whenever applicable.
Table 2 6 shows the group centroids for the
discriminant anlysis using the SDQ III. Function One
discriminates between the Successful Students (Group
2) and those students who were in the Marginally
Successful (Group 1) and the Unsuccessful Students
(Group 0). Function Two discriminates between the
92
Unsuccessful Students (Group 0) and the Marginally
Successful (Group 1) and Successful Students (Group
2).
Table 26
Canonical discriminant function evaluated at group
means (group centroids)
Group Function 1 Function 2
Group 0 .57848 .55872
Group 1 .34373 -.36562
Group 2 -.66499 -.06422
Table 27 reports the classification results based
on using the SDQ III to help predict which students
would experience academic difficulty during their
freshman year in college.
Table 27
Classification Results for SDQ III
Actual group
No. of
Cases
Predicted grp.
0 1
membership
2
Group 0 23 14
60.9%
5
21.7%
4
17.4%
Group 1 29 10
34.5%
13
44.8%
6
20.7%
Group 2 34 5
14.3%
4
11.4%
26
74.3%
% Of "grouped" cases correctly classified: 60.9%
Cases by chance: 33%
The classification results for the Unsuccessful
Group (Group 0) was a correct classification of 60.9%
of the cases, the Marginally Successful Group (Group
93
1) had the lowest correct classification results at
44.8%, and the Successful Group (Group 2) yielded the
highest correct classification results at 74.3%.
Overall, the SDQ III had the highest percentage
of cases correctly classified (60.9%). The "hit rate"
on the target group was 8% less than that for the
combination of the cognitive and affective variables,
but was still fairly high at 60.9%. The SDQ III was
able to predict with high accuracy those students who
would be successful during their freshman year
(74.3%).
Summary
This chapter presents the results of the data
analyses corresponding to the five research questions
which were proposed in the current study. The goal of
this study was to determine whether measures of self-
concept and beliefs about effort and responsibility
regarding academic achievement could be used to
predict which minority freshmen at a predominately
white, urban university will experience academic
difficulty during their freshman year.
The first research question sought to examine the
internal consistency of the scales on all three
measurement instruments utilized. This question was
considered important because scores on so many of the
instruments used in the research in retention in
94
higher education do not exhibit high levels of
internal consistency.
The second research question pertained to the
cognitive variables which are used to admit students
to most universities. A discriminant analysis was
completed which indicated that when only cognitive
variables were considered High School GPA was the most
useful predictor of academic difficulty.
The third research question was concerned with
which affective variables (academic and social) could
be used to predict which students were at risk for
academic difficulty. The affective variables proved
to be better predictors of those students who would
experience academic difficulties during their first
year in college than were the cognitive variables.
The fourth research question combined the
affective variables and cognitive variables which have
been shown to be related to academic achievement.
This combination of variables had the highest "hit
rate" for the prime target group, the Unsuccessful
students (68.2%) but was not appreciable more useful
than when affective variables were used alone (65.2%).
The fifth research question examined two self-
concept questionnaires which have been devised to help
administrators predict those students who would
experience academic difficulty (DOSC-H) or determine
95
their academic and social self-concept (SDQ III). The
DOSC-H was not able accurately to predict which
minority students would encounter academic difficulty.
The SDQ III had moderate success at predicting which
students would experience academic success and failure
during their freshman year. The overall "hit rate"
for the SDQ III was the highest of all the
discriminant analyses completed (60.2%).
96
Chapter IV
Discussion
This final chapter synthesizes the findings
regarding minority students who have encountered
academic difficulty during their freshman year in
college at a predominately white, urban university.
This chapter will follow a similar pattern to chapter
three where each of the research questions is
addressed individually. It focuses on practical
implications as well as on implications for theory and
future research, and ends with a conclusion and
recommendations regarding research with minority
populations in higher education.
Analysis of the Findings
Internal Consistency of Questionnaires
The scores on the SDQ III had the highest
reliability coefficients overall for the tests given.
The scores on the subscales on Math Self-Concept and
General Self-Concept both were above a .90. Scores on
Opposite Sex Peer, Self-Concept regarding Parental
Relationship, Academic Self-Concept, Emotional
Stability, Verbal Self-Concept, and Problem Solving
Self-Concept all had coefficients between a .88
and .81. The only subscale with scores yielding a
coefficient below a .80 was Same Sex Peer which had a
97
coefficient of .74 which is still satisfactory for
research purposes.
For the DOSC-H the highest reliability
coefficient was obtained for scores on the subscales
Level of Aspiration and Anxiety with the coefficient
scores being .89 and .87 respectively. The scales
Identification vs. Alienation, Academic Interest and
Satisfaction, and Leadership and Initiative had
coefficients of .85, .85 and .83. These high
internal reliability coefficients indicated that the
DOSC-H is a reliable instrument for use with college
students.
Scores on the IAR afforded the lowest reliability
coefficients. This outcome might have been due to the
fact that the questionnaire is dichotomous in nature.
Scores on instruments with dichotomous items would be
expected to have lower internal consistency. These
low internal reliability scores make this instrument
questionable for use in research.
As the sample of students in this study was not
the typical of students normally used in research, it
was especially important to measure internal
consistency reliability of scores with the test
instruments administered.
The SDQ III, DOSC-H, and the IAR were selected
specifically for this research project because of the
98
extensive analysis conducted on them which verifies
their appropriateness as measures of self-concept and
beliefs about self. Of the three, however, the SDQ
III appears to be the strongest in terms of its
internal consistency.
Cognitive Variables
Cognitive variables have been used for years to
admit students to college with the belief that they
can predict success in college. Recently, however,
cognitive variables, especially SAT Mathematics and
SAT Verbal test scores, have been shown not to be
valid predictors of academic success for minority
students (Astin, 1982, Duran, 1986).
This study used the three cognitive variables
available to most college administrators to execute a
discriminant function analysis predicting academic
difficulty. The only variable which captured enough
variance to enter the discriminant equation was High
School GPA. The findings revealed that the higher the
High School GPA the more likely the student was to not
encounter academic difficulty in college. For a
number of years High School GPA did not hold as much
credibility for those making college admissions
decisions as the SAT test scores because High School
GPA is influenced by a number of factors which have
nothing to do with academic ability. For example,
99
High School GPA can be influenced by the personality
characteristics of teachers, as well as difficulty of
course material. It has been suggested, however, that
High School GPA reflects study skills and habits which
can carry a student through new learning situations.
By comparison, SAT scores may measure a student's
academic ability at the time he or she completed the
examination, but it does not provide a measure of the
student's motivation to achieve in college.
Overall, cognitive variables are not as valid as
affective variables for prediction of academic success
for minority students at a predominately white, urban
university. In fact, the mean SAT Mathematics and SAT
Verbal scores were highest for those students who had
earned the lowest cumulative college GPA. High
School GPA which was the most useful predictive
cognitive variable was not able correctly to identify
the students in the most important target group in the
study - - those who were in academic difficulty. When
only cognitive variables in the discriminant analysis
were used the classification matrix could only
correctly identify 18% of the students who had earned
below a 2.0 cumulative GPA their freshman year in
college. Out of the 22 students on academic probation
(below a 2.0 cumulative GPA) the classification matrix
100
correctly identified only 4 students who had earned
below a 2.0 cumulative GPA.
Affective Variables
In the context of the third research question,
affective variables alone were used to determine
whether they could predict which students would
experience academic difficulty during their freshman
year. The affective variables fell into three
categories (a) academic self-concept, (b) social self-
concept, and (c) internal responsibility for positive
and negative academic events.
The discriminant analysis produced two
significant function equations. The first equation
discriminated between those students who earned above
a 2.6 cumulative GPA and those students who had earned
below a 2.6 cumulative GPA. The second equation
discriminated.between those students who had earned
between 2.0 - 2.59 cumulative GPA and those students
who had earned below a 2.0 and above a 2.6 cumulative
GPA. However, the group centroids for the second
equation were almost equally distanced or spread
apart. The first equation used the variables Opposite
Sex Peers, Problem Solving Self-Concept, and Internal
Responsibility regarding Negative Events. The second
equation used the variables Identification vs.
101
Alienation, Math Self-Concept, and Self-Concept
regarding Parental Relationship.
The classification matrix correctly classified
65.2% of those students who would experience academic
difficulty - - earning below a 2.0 cumulative GPA - -
during their freshman year. The overall "hit rate"
was 57.47% of the cases correctly classified. The
"hit rate" for those students who were Unsuccessful
was greatly improved when using affective variables
compared to employing cognitive variables alone.
The results of the discriminant analysis can be
interpreted as follows: The students who earned below
a 2.59 cumulative GPA were more interested in the
opposite sex and perceived that they had better
relationships with members of the opposite sex, felt
better about their problem-solving skills, and
accepted more responsibility for negative academic
events such as failing a test. It also indicated that
those students who had earned between a 2.0 and 2.59
cumulative GPA identified with the university more
than the other students, had the lowest mathematics
self-concept, and the poorest relationship with their
parents.
An interesting finding was the important
prediction power of the variable Opposite Sex Peers.
102
Research in the area of retention of students in
higher education has not been directed toward the
study of opposite sex relationships. It would be
interesting to uncover the impact dating may have on a
student's academic endeavors especially for minority
students who may feel more alienated at a
predominately white university than a non-minority
student.
In conclusion, the discriminant analysis using
affective variables alone demonstrated moderate
success as compared to using cognitive variables
alone. The classification matrix correctly
categorized 65.2% of those students who would
experience academic difficulties during their first
year in college. It was interesting to discover that
three of the six variables employed in the
discriminant function equations were social self-
concept variables.
Cognitive and Affective Variables
Both cognitive and affective variables were used
in the discriminant analysis to ascertain which
combination of variables can more accurately predict
academic success and failure. The cognitive variables
used in this study were High School GPA, SAT
Mathematics, and SAT Verbal test scores. The
affective variables used in this study fell into three
103
categories: (a) academic self-concept, (b) social
self-concept, and (c) beliefs a student had regarding
responsibility about the need for effort and ability
in academic situations.
The two discriminant function equations which
were generated for this analysis were both
significant. Function One discriminated between the
Successful students and the groups of the Unsuccessful
and Marginally Successful students when use was made
of the variables Opposite Sex Peers, High School GPA,
and Internal Responsibility for Negative Events.
Function Two discriminated between Marginally
Successful students and Unsuccessful and Successful
students when the variables Identification vs.
Alienation, Math Self-Concept, Level of Aspiration,
Self-Concept regarding Parental Relationship, and
Verbal Self-Concept were employed. However the group
centroids for Function Two were fairly equally spaced
at .64317, -.58545, and .06409.
The classification matrix correctly identified
68% of the students in the Unsuccessful Group (Group
0), the most important target group, although it
misidentified 54% of the Marginally Successful Group
(Group 1), and 49 % of the Successful Group (Group 2).
These percentages were possibly biased in a positive
104
direction because the sample was not large enough to
perform a cross-validation procedure in the analysis.
The results of the discriminant analysis can be
interpreted as follows: Function One discriminated
between those students who had earned above a 2.6
cumulative GPA and those students who had earned below
a 2.6 cumulative GPA. The variables used in the
function were High School GPA, Opposite Sex Peers, and
Internal Responsibility for Negative Events. The
profile for those students who had earned a cumulative
GPA of 2.6 or above were those students who had earned
the highest High School GPA, were the least
comfortable with peers of the opposite sex, and took
the least amount responsibility for failures. Marsh
(1986) stated that it is important for students to
accept responsibility for academic success and to
attribute this success to ability or effort. However,
if students accept responsibility for academic
failures, this action can lower their academic self-
concept. Wiener (1989) took issue with Marsh and
declared that only if a student attributes failures to
lack of ability will it affect the student's concept
of self.
In this study those students who had earned the
highest cumulative GPA in their freshman year,
105
accepted responsibility for academic failures the
least. They attributed the academic failures to
external events, such as the test being too hard or
the instructor being a poor teacher.
Function Two utilized the variables
Identification vs. Alienation, Mathematics Self-
Concept, Level of Aspiration, Self-Concept regarding
Parental Relationship, and Verbal Self-Concept to
discriminate between the students who scored below a
2.0 or above a 2.6 cumulative GPA and those students
who scored between a 2.0 and 2.6. cumulative GPA. The
group centroids in this function (.64317, -.58545,
and .064 09) showed that all three groups were almost
equally separated. This function indicated that those
students who expressed the most confidence regarding
verbal ability, imparted that they were the most
alienated on campus, and expressed fewer aspirations
earned the lowest cumulative GPA. On two of the five
variables - - Mathematics Self-Concept and Self-
Concept regarding Parental Relationship - - those
students who had received the highest cumulative GPA
and those students who had scored the lowest
cumulative GPA were very similar. These conflicting
findings could not be easily explained.
One important finding in this discriminant
analysis was the strong contribution of the Opposite
106
Sex Peers variable which was also found when the
analysis used affective variables only. This variable
might point to an area of study which needs further
research. The Opposite Sex Peers variable may measure
not only how comfortable the student is with the
opposite sex, but also how likely the student is to
spend time dating - - an expenditure of time which may
seriously compete with time needed to study.
According to Astin (1975), serious dating among
college students may negatively impact college grades.
In conclusion, the discriminant analysis using
cognitive variables, and social and academic self-
concept variables demonstrated moderate success at
predicting which students would experience academic
difficulty during their freshman year. The
classification matrix correctly identified 68% of
those students in the Unsuccessful Group (Group 0).
This rate of success needs to be duplicated, as there
was no cross-validation procedure completed.
SDQ III vs. DOSC-H
Much of the research in retention of students in
higher education has emphasized author developed
instruments which lack reliability and often employ
one-item scales to measure a particular concept. In
order to overcome this shortcoming in previous
research, the present study utilized instruments for
107
which there is a growing body of empirical data
regarding their characteristics and utility.
Both instruments, DOSC-H and SDQ III, were used
in separate discriminant analyses to ascertain whether
on their own they could be used to predict students
who would encounter academic difficulty. Since the
IAR did not prove to be highly reliable the measure
was not used separately in any discriminant analysis.
The DOSC-H did not turn out to be a useful predictor
of those who would experience academic difficulty in
this sample of minority students. The DOSC-H, which
has been used with other samples of college students,
was able, however, to predict academic success
(Caracosta & Michael, 1986).
The SDQ III was a more accurate predictor of
those students who encountered academic difficulty
than was the DOSC-H. When used alone, the SDQ III
accurately predicted 60% of those students who would
experience academic difficulty during their first year
of college and 74% of those students who would earn
above a 2.6 cumulative GPA. The success of the SDQ
III might be attributed to the specificity and
relatively larger assortment of variables such as
three different measures of social interaction, one of
emotionality, and several measures of academic self
108
concept. The increased specificity of the instrument
may increase the predictive validity of its scores.
Practical Implications
Out of the 87 students sampled in the study,
there were 23 students who had earned below a 2.0
cumulative GPA during their first year in college. Of
the 23 students, 15 of them were Engineering students.
The Engineering students at the University of
Southern California have a very rigorous curriculum.
Most of the students have calculus, chemistry, one
engineering class, and composition during their first
semester. The second semester, they take second
semester calculus, physics, another engineering class,
and one general elective class. For most of the
students the work load is more than they had imagined.
Many have a difficult time completing all the work
given to them. This circumstance may be why High
School GPA is a valid predictor variable because it
indicates not only academic ability but also study
skills and persistence.
It has also been noted that students who have a
heavy course load may not be able to spend as much
time in social activities as other students. Being a
minority student in a predominately white campus may
exacerbate the student's alienation and isolation - -
109
factors in turn which may cause these students to
spend more time socializing than studying. These
findings suggest that it may be beneficial for
educators to organize study groups for students which
could provide the students with some socializing as
the students complete their studying.
Implications for Theory
The data from this study call into question the
two theories which served as its theoretical
foundation. Tinto's (1987) model of retention of
students in higher education emphasizes the need for
social integration as well as academic integration in
order for students to succeed in college. Marsh and
Shavelson's model (Marsh, 1986, 1990, 1992; Shavelson
& Bolus, 1982) of self-concept stresses the connection
between academic self-concept and academic
achievement.
In the present study, the group which was
struggling academically (cumulative GPA below 2.0)
scored highest on the following variables: Opposite
Sex Peers, Same Sex Peers, Verbal Self-Concept,
Mathematics Self-Concept, SAT Mathematics and SAT
Verbal measures. The scores on these measures
demonstrated that high academic ability, social
confidence, and high academic self-concept did not
predict success to any substantial degree in college
110
for minority students in a predominately white, urban
university. Perhaps Tinto's (1987) hypothesis can
illuminate what these students are experiencing in
college, at least in regard to the need for
socialization. Tinto suggested that students who
become overly involved in either formal or informal
organizations on campus may run the risk of academic
difficulty leading to academic probation.
There are at least three possible interpretations
of the data ascertained in the present study. First,
that the findings are specifically related to the
sample of minority students at a predominately white
university. For these students, social variables may
be as important if not more important than academic
variables when trying to predict who will succeed in
college. The feelings of alienation in minority
students on a predominately white campus are
frequently cited in research (Loo & Rolison, 1986;
Malinckrodt, 1988; Smedley, Myers, & Harrell, 1993) as
having been associated with academic difficulty.
A second possible interpretation of the data is
associated with the time at which students answered
the questionnaires regarding their self-concept. The
students answered the questionnaires before they began
their college experience. The data in this study may
have been affected by the reference group hypothesis.
Ill
This hypothesis states that a student's self-concept
is formed by comparison of one's self to other
students in one's reference group. The students
sampled in this study answered the questionnaire
before they experienced college, and used their high
school peers as their reference group rather than
their college peers. Many of the students may not
have attended highly competitive high schools or been
exposed as much to a more competitive peer group as
they would be in college. If these questionnaires had
been answered by students after they had experienced
several weeks of school, the scores on some of the
academic and social scales may have been different.
A third possibility may be that because the
questionnaires used in this study were not developed
with a minority population in mind, the questions
might not address certain aspects regarding academic
and social self-concept which would be important for
them. According to Covington (1992), minority
students may base their academic self-concept on
different aspects of academic life rather than on just
grades and on the manner in which they perform on
tests. Ogbu (1989) suggested that minority students
may base their academic self-concept on events, such
as their ability to negotiate in a white dominated
system, rather than scores on tests.
112
In conclusion, the theories which directed the
focus of the present study may need to be revised when
used with minority populations. Social variables are
very important for minority students at a
predominately white university, although it may be
possible for these students to become overly involved
in social activities which could be detrimental for
academic success (Tinto, 1987). These students need
social support but not social distractions. It is
important to note that the variable which contributed
the most variance to any of the analysis was the
subscale Opposite Sex Peers. Those students who
scored highest on the subscale Opposite Sex Peers,
which means those students who felt they had the best
relationship with the opposite sex, had the lowest
cumulative GPA. It may be that those students who
scored highest on the Opposite Sex Peers subscale,
spent more time dating in college, and therefore had
less time available for studying. The dating behavior
of students has not been addressed in retention
research, but it may be an area which needs to be
explored in the future.
The reference group hypothesis may explain some
of the unusually high scores on several of the
academic and social self-concept variables for those
113
students who earned the lowest cumulative GPA in
college. This hypothesis states that students use the
other students in their reference group against whom
to compare themselves when they form their concepts of
self. As the students had not experienced college
when they answered the questionnaire, they were still
using their high school reference group as a
comparison group. Also, the questionnaires
administered in this type of study need to be
operationalized with minority populations in mind.
Implications for Future Research
Retention research on minority students is just
beginning to gain momentum. More researchers are
trying to understand what variables are important to
help retain minority students in higher education.
One key element to all the research is the importance
of social variables which can affect academic
achievement. It is extremely important to realize
that not all minority students enter higher education
with academic deficits. However, they can experience
academic difficulty because of feelings of social
isolation and alienation which are more pronounced for
them than for majority students.
Researchers need to understand what minority
students experience in predominately white campuses
and to try to find some way to identify which students
114
are most at risk for academic difficulty. The
cognitive variables with which the students were
admitted to the university are not highly accurate
predictors of which students will experience academic
difficulty during their freshman year.
Future research in the area of retention of
minority students in higher education needs to look at
non-cognitive variables which are likely to impact
students education. These non-cognitive variables
include the social, emotional, and academic beliefs
held by students which influence how they will adjust
to college life. These variables, however, will need
to address the specific concerns which are encountered
by minority students in college.
The present research did not examine the
different ethnic groups separately because the sample
was too small. Future research needs to look at
ethnic groups individually to determine whether there
are some variables which are more predictive for one
group than for another. One problem for researchers
to overcome is that of the small sample sizes
available for these ethnic groups at predominately
white universities.
Conclusions and Recommendations
Each educational research project is one small
piece to a complex puzzle facing the higher
115
educational system in this country. One of the most
urgent problems facing higher education in the United
States is the low retention rates of the minority
students on its campuses. Minority students are being
admitted to universities only to have their dreams
taken away from them through subsequent academic
failure. This failure is often not due to lack of
academic ability, as substantiated by the high scores
on SAT tests for many students on academic probation.
This study illuminates the connection between
social variables and academic achievement, revealing
the complexity of their interaction. If students
spend too much time trying to socialize with others on
campus, they may be taking time away from their
academic studies. Conversely, if they don't have some
degree of satisfactory social connection, they may
become estranged from the campus, and increase their
likelihood of dropping out. The present study has
indicated that the minority students who experienced
academic failure during their first year in college
were more socially involved than those who were
academic successes.
Retention programs on college campuses can try to
ameliorate these problems, but first it is important
to know what the students are experiencing during
their freshman year and what they perceive could help
116
them make a better adjustment to college. More
qualitative and quantitative research needs to be
completed to address these problems. Qualitative
research needs to asks questions regarding how
minority students define academic self-concept and how
dating practices can help and hinder academic
achievement. Quantitative research needs to specify
what types of social interaction can help students be
more successful in college.
This research study has hopefully provided at
least a limited contribution to the body of knowledge
regarding the retention of minority students in higher
education. First, the findings serve to confirm in
part the results of several previous studies which
revealed that high school GPA was a more valid
predictor of academic success for minority students
than were SAT scores.
Second, the findings indicated that non-cognitive
measures were more accurate predictors than were
cognitive ones for academic success of minority
students in a predominately white, urban university.
This finding also suggests that some students can feel
self-confident about themselves academically and still
experience academic difficulty.
Third, the findings revealed that social
variables are very important for minority students in
117
a predominately white, urban university. More
research needs to be conducted to understand the
importance of interactions with the opposite sex for
students in higher education. This area of retention
is one which has not been addressed by researchers.
Last, the findings suggest that more attention
needs to be placed on research in the area of
retention of minority students in a predominately
white, urban university. In particular, more
attention needs to be placed on measurement
instruments which are devised especially for minority
students. The instruments used in this study were not
devised to address specific issues which minority
students experience in predominately white, urban
campuses.
In conclusion, this study has identified an area
of study which has been ignored by researchers
interested in retention management. This area of
study involves social variables which may have a major
impact on academic success, such as serious dating and
overinvolvement in school sponsored organizations.
Future research efforts may need to incorporate these
social variables in their studies to truly understand
the complex factors which may lead to academic
difficulty during the freshman year.
118
References
Allen, W. (1988). Black students in U.S. higher
education: Toward improved access, adjustment,
and achievement. Urban Review. 20, 165-188.
Arbona, C. & Novy, D. (1990). Noncognitive
dimensions as predictors of college success
among black, Mexican-American, and white
students. Journal of College Student
Development. 31, 415-422.
Astin, A. (1970). The methodology of research on
college impact, part one. Sociology of
Education 43, 223-254.
Astin, A. (1975). Preventing students from
dropping out. San Francisco: Jossey-Bass.
Astin, A. W. (1977). Four critical years. San
Francisco: Jossey-Bass.
Astin, A.W. (1982). Minorities in American higher
education. San Francisco: Jossey-Bass.
Astin, A. (1986). Achieving educational
excellence. San Francisco: Jossey-Bass.
Atkinson, J.W. (1957). Motivational determinants
of risk-taking behavior. Psychological Review.
64, 359-372.
Atkinson, J.W. (1964). An introduction to
motivation. New Jersey: Van Nostrand.
Atkinson, J.W. (1987). Michigan studies of fear
of failure. In F. Halisch & J. Kuhl (Eds.)
Motivation intention and volition (pp.47-60).
Berlin: Springer.
Bachman, J.G. & O'Malley, P.M. (1986). Self-
concepts, self esteem, and educational
experiences: The frog pond revisted (again).
Journal of Personality and Social Psychology.
50, 35-46.
Battle, E., & Rotter, J. (1963). Children's
feelings of personal control as related to
social class and ethnic group. Journal of
Personality. 31, 428-490.
119
Bean, J.P. (1980). Dropout and turnover: The
synthesis and test of a causal model of student
attrition. Research in Higher Education. 12,
155-187.
Bean, J.P. (1983). The application of a model of
turnover in work organizations to the student
attrition process. Review of Higher Education.
6, 129-148.
Bean, J.P. (1985). Interaction effects based on
class level in an explanatory model of college
student dropout syndrome. American Educational
Research Journal. 22, 35-64.
Bean, J.P. (1990). Why students leave: Insights
from research. In Hossler, J.P Bean, J.P.
(Eds.) The Strategic Management of College
Enrollments. San Francisco: Jossey-
Bass .
Bentler, P., & Speckart, G. (1979). Models of
attitude-behavior relations. Psychological
Review. 86, 452-464.
Boyer, S.P. & Sedlacek, W.E. (1988). Noncognitive
predictors of academic success for
international students: A longitudinal study.
Journal of College Student Development. 29,
218-223.
Braithwaite, R. (1955). Scientific explanation.
Cambridge, England: Cambridge University
Press.
Brookover, W., Erickson, E., & Joiner, L. (1967)
Self-concept and ability and school achievement
III (U.S. Office of Educational Cooperative
Research Report. Project No. 2831). East
Lansing: Michigan State University, Office
of Research and Publications.
Byrne, B.M. (1986). Self-concept/academic
achievement relations: An investigation of
dimensionality, stability, and causality.
Canadian Journal of Behavioral Science. 18,
173-186.
120
Cabrera, A.F.; Nora,A. & Castaneda, M.B. (1993).
Structural equations modeling test of an
integrated model of Student Retention. Journal
of Higher Education. 64, 123-139.
Caracosta, R. & Michael, W.B. (1986). The
construct and concurrent validity of a measure
of academic self-concept and one of locus of
control for a sample of university students.
Educational and Psychological Measurement. 46,
735-744.
Cassel, R.N. (1990a). The Life Style Analysis
Test (LFSTYLE). Chesterfield, Missouri:
Psychologist & Educators, Inc.
Cassel, R.N. (1990b). The Independence Versus
Regression Test (BALANCE). Chesterfield,
Missouri: Psychologist & Educators, Inc.
Covington, M. (1984). The self-worth theory of
achievement motivation: findings and
implications. Elementary School Journal. 85,
5-20.
Covington, M. (1992). Making the grade. New York:
Cambridge Press.
Crandall, V.C., Katkovsky W., Crandall, V.J.
(1965). Children's belief in their own control
of reinforcements in intellectual-academic
achievement situations. Child Development. 36,
91-109.
Donovan, R. (1984). Path analysis of a
theoretical model of persistence in higher
education among low-income black youth.
Research in Higher Education. 21, 243-259.
Dowaliby, F., Garrison, W. & Dagel, D. (1993).
The student integration survey: Development of
a early alert asssessment and reporting system.
Research in Higher Education. 34, 513-531.
Duran, R. (1986). Prediction of Hispanic college
achievement. In Olivas, M. Latino College
Students. New York: Teachers College Press.
Durkheim, E. (1961). Suicide. (J. Spaulding and
G. Simpson, Trans.) Glencoe: The Free Press.
121
Educational Testing Service. (1948-1994). College
Entrance Examination Board. Scholastic
Aptitude Test. Princeton, NJ: Education
Testing Service.
Ethington, C.A. (1990). A psychological model of
student persistence. Research in Higher
Education.31. 279-293.
Fishbein, M., & Ajzen, I. (1975). Belief,
attitude, intention, and behavior: An
introduction to theory and research. Mass.:
Addison-Wesley.
Fordham, S., & Ogbu, J.U. (1986). Black student's
school success: Coping with the burden of
"acting white." The Urban Review. 18, 176-206.
Gerardi, S. (1990). Academic self-concept as a
predictor of academic success among minority
low-socioeconomic status students. Journal
of College Student Development.31. 402-407.
Graham, S. (1988). Can attribution theory tell us
something about motivation in blacks?
Educational Psychologist. 23, 3-21.
Hattie, J. (1992). Self-concept. Hillsdale, NJ:
Erlbaum.
Hare, B. (1985). Stability and change in
self-perception and achievement among black
adolescents: a longitudinal study. The Journal
of Black Psychology. 11, 29-42.
Hansford, B.C. & Hattie, J.A. (1982). The
relationship between self and
achievement/performance measures. Review of
Educational Research.52. 123-142.
Holmes, R. (1992). Comparing positive and
negative factors related to social development
for predicting college GPA. College Student
Journal. 26, 14-19.
Hossler, D. (1984). Enrollment management: An
integrated approach. New York: College
Entrance Examination Board.
122
Hossler, D. & Bean, J.P. & Associates. (1990).
The Strategic Management of College
Enrollments. San Francisco: Jossey-Bass.
House, J.D. (1992). The relationship between
academic self-concept, achievement-related
expectancies, and college attrition. Journal
of College Student Development. 33, 5-10.
Krotseng, M. (1992). Predicting persistence from
the student adaptation to college
questionnaire: Early warning or siren song?
Research in Higher Education. 33, 99-111.
Lefcourt, H.M. (1982). Locus of control: Current
trends in theory and research. Hillsdale, NJ:
Erlbaum.
Loo, C. & Rolison, G. (1986). Alienation of
ethnic minority students at a predominately
white university. Journal of Higher Education.
57, 58-77.
Lyon, M.A. & MacDonald, N.T. (1990). Academic
self-concept as a predictor of achievement for
a sample of elementary school students.
Psychological Reports. 66, 1135-1142.
Mallinckrodt, B. (1988). Student retention,
social support, and dropout intention:
comparison of black and white students.
Journal of College Student Development. 17,
60-64.
Marsh, H. W.; Relich, J; & Smith, I. (1983).
Self-concept: The construct validity of
interpretations based upon the SDQ. Journal of
Personality and Social Psychology. 45, 173-187.
Marsh, H.W. (1984). Self-concept: The application
of a frame of reference model to explain
paradoxical results. Australian Journal of
Education. 28, 165-181.
Marsh, H.W. (1986). Global self-esteem: Its
relation to specific facets of self-concept and
their importance. Journal of Personality and
Social Psychology.51. 1224-1236.
123
Marsh, H.W. (1987). The big-fish-little-pond
effect on academic self-concept. Journal of
Educational Psychology. 79, 280-295.
Marsh, H.W. (1990). Causal ordering of academic
self-concept and academic achievement. Journal
of Educational Psychology. 82, 646-656.
Marsh, H.W. (1992). Content specified of relations
between academic achievement and academic
self-concept. Journal of Educational
Psychology. 84, 35-42.
Marsh, H. W., & Hocevar, D. (1985). The
application of confirmatory factor analysis to
the study of self-concept: First and higher
order factor structures and their invariance
across groups. Psychological Bulletin. 97,
565-582.
Marx, R.W. & Winne, P.H. (1978). Construct
interpretations of three self-concept
inventories. American Educational Research
Journal. 15, 99-108.
Mclnerney, D.M. (1988). Cross-cultural studies of
achievemnt motivation: Educational
implications and research directions for the
future. Sydney, Australia: Macarthur
Institute of Higher Education.
Metzner, B.S. & Bean, J.P. (1987). The estimation
of a conceptual model of nontraditional
undergraduate student attrition. Research in
Higher Education. 27, 15-39.
Michael, W.B., Smith, R.A. & Michael, J.J. (1989).
Dimensions of Self-concept (DOSC^: A technical
manual (revised). San Diego: EdiTS.
Norusis/SPSS Inc. (1988). SPSS-X advanced
statistics guide. Chicago: SPSS Inc.
Ogbu, J. (1989). The individual in collective
adaption: A framework for focusing on academic
underperformance and dropping out among
involuntary minorities. In Weis, L.; Farrar,
E., Petrie, H., Issues, Dilemmas, and
Solutions. Albany: State University of New
York Press.
124
Olivas, M.A. (1986). Latino college students.
New York: Teachers College Press.
Pascarella, E.T. & Terenzini, P.T. (1980).
Predicting voluntary freshman year
persistence\withdrawal behavior in a
residential university: A path analytic
validation of Tinto's model. Journal
of Educational Psychology. 75, 215-226.
Pascarella, E.T. & Terenzini, P.T. (1980).
Predicting freshman persistence and voluntary
dropout decisions from a theoretical model.
Journal of Higher Education. 51, 60-75.
Poole, S. (1989). Predicting academic success
in college for students from disadvantaged
backgrounds using locus-of-control. self-
concept and selected standard data. Michican:
Dissertation Services.
Price, J. (1977). The Study of Turnover. Ames:
Iowa State University Press.
Rosenburg, J. (1965). Society and the adolescent
self-image. Princeton, NJ: Princeton
University Press.
Rotter, J.B., Chance, J.E., & Phares, E.J. (1972)
Applications of a social learning theory of
personality. New York: Holt, Rinehart &
Winston.
Sedlacek, W.E. (1987). Black students on white
campuses: 20 years of research. Journal of
College Student Personnel. 28, 484-495.
Sedlacek, W. & Brooks, G.C., Jr. (1976). Racism
in American education: A model for change.
Chicago: Nelson-Hall.
Servis, K. (1993). Freshman Cohort Analysis.
Office of Academic Records and Registrar:
University of Southern California.
Sherman, S.J. (1973). Internal-external control
and its relationship to attitude change under
different social influence techniques. Journal
of Personality and Social Psychology. 26(3),
23-29.
125
Shavelson, R.J., Hubner, J.J., & Stanton, G.C.
(1976). Validation of construct
interpretations. Review of Educational
Research. 46, 407-441.
Shavelson, R.J., Bolus, R. (1982). Self-concept:
the interplay of theory and methods. Journal
of Educational Psychology. 74, 3-17.
Smedley, B.D.; Myers, H.F.; Harrell, S.P. (1993).
Minority-status stresses and the college
adjustment of ethnic minority freshmen. Journal
of Higher Education. 64, 435-452.
Soares, L.M., & Soares, A.T. (1982). Factor
analysis of the academic self. Paper presented
at the American Educational Research
Association, New York.
Song, I.S., & Hattie, J.A. (1985). Relationships
between self-concept and achievement. Journal
of Research in Personality. 19, 365-372.
Spady, W. (1970). Dropouts from higher education:
An interdisciplinary review and synthesis.
Interchange. 1, 64-85.
Spady, W. (1971). Dropouts from higher education:
Toward an empirical model. Interchange. 2,
38-62.
Stern, G.G. (1970). People in context. New York:
Wiley.
Stewart, D.M. (1991). Higher education. In D.W.
Hornbeck & L.M. Salamon (Eds). Human capital
and American's future (pp. 193-219).
Baltimore: John Hopkins Univeristy Press.
Terenzini, P.T. & Pascarella, E.T. (1977).
Voluntary freshman attrition and patterns of
social and academic integration: A test of a
conceptual model. Research in Higher Education.
6, 25-43.
Terenzini, P.T. & Pascarella, E.T. (1978). The
relation of student's precollege
characteristics freshman year experience to
voluntary attrition. Research in Higher
Education. 9, 337-366.
126
Terenzini, P.T. & Pascarella, E.T. (1980). Toward
the validation of Tinto's model of college
student attrition: A review of recent studies.
Research in Higher Education. 12, 271-282.
Tinto, V. (1975). Dropout from higher education:
a theoretical synthesis of recent research.
Review of Educational Research. 45, 89-125.
Tinto, V. (1987). Leaving college: Rethinking
the causes and cures of student attrition.
Chicago: Chicago Press.
Tinto, V. (1988). Stages of student departure:
reflections on the longitudinal character of
student leaving. Journal of Higher Education.
59, 438-455.
Tracey, T. & Sedlacek, W. (1984). Non-cognitive
variables in predicting academic success by
race. Measurement and Evaluation in Guidance.
16, 171-178.
Tracey, T. & Sedlacek, W. (1985). The
relationship of noncognitive variables to
academic success: A longitudinal comparison by
race. Journal of College Student Personnel.
405-410.
Tracey, T. & Sedlacek, W. (1987). Prediction of
college graduation using noncognitive variables
by race. Measurement and Evaluation in
Counseling and Development. 19, 177-184.
Weiner, B. (1974). Achievement motivation and
attributional theory. New Jersey: General
Learning Press.
Weiner, B. (1990). Attribution in Personality
Psychology. In L. Pervin (Ed.), Handbook of
Personality. New York: Guilford Press.
White, T.J. & Sedlacek, W.E. (1986). Noncognitive
Predictors. The Journal of College Admissions.
20-23.
127
Wilhite, S. (1990). Self-efficacy, locus of
control, self-assessment of memory ability, and
study activities as predictors of college
course achievement. Journal of Educational
Psychology. 82, 696-700.
Zytkoskee, A, & Strickland, B. (1971). Delay of
gratification and internal versus external
control among adolescents of low socioeconomic
status. Developmental Psychology. 4, 93-98.
Abstract (if available)
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
Asset Metadata
Core Title
00001.tif
Tag
OAI-PMH Harvest
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-oUC11256557
Unique identifier
UC11256557
Legacy Identifier
9617117