A MULTIVARIATE ANALYSIS OF THE
RELATIONSHIP OF ACADEMIC APTITUDE,
SOCIAL BACKGROUND, ATTITUDES AND

VALUES TC} COLLEGIATE PERSISTENCE

Thesis for the Degree OI ph. D.
MICHIGAN STATE UNIVERSITY

Stanley 0. Ikenbzerry
1960

I mum; III" LII I! I III III I III II III I

6 713

This is to certify that the
”I i
thesis entitled

A HUM'IVARIATE ”ELISE OF THE MIOIISHIP ORACADEHIC
APTITUDE, SOCIAL BACKGROUND, ATTITUDE} AND
VALUE TO COLIMIATE PESISTENCE

I" i

presented by
Stanley 0. Ikonbm

has been accepted towards fulfillment
of the requirements for l

m.— degree in W

.. 1 . V
Major professor
Datew lb 1960

0-169

 

 

A.MUDTIVARIATE ANALISIS OF THE RELETIONSHIP OF ACADEMIC
APTITUDE, SOCIAL BACKGROUND, ATTITUDES AND

VALUES T0 COLLEGIATE PERSISTENCE

BX
Stanley 0. Ikenberry

A THESIS

Submitted to the School for Advanced Graduate Studies
of Michigan State University of Agriculture and
Applied Science in Partial Fulfillment of
the requirements for the degree of

DOCTOR OF PHILOSOPHY
Department of Education

1960

Chapter

I.

II.

III.

TABLE OF CONTENTS

TABLE OF CONTENTS . . .
LIST OF TABLES. . . . .
LIST OF ILLUSTRATIONS .
ACKNOWLEDGEMENTS. . . .
ABSTRACT. . . . . . . .
THE PROBLEM . . . . . .

The Importance of the

Problem
The IndiVidual Student. 0 o
The SOCiety o o o o o o o 0
The Educational Institution

Purpose of the Study. . . . . .
Rationale of the Study. . . .
Summary of the Purpose of the

and the Underlying Rationale. . . .
The Hypothesis to be Tested . . . . .

Study

THE REVIEW OF’THE LITERATURE. . . . . . .

Academic Aptitude and Student Withdrawal

from C0116geo o o o o o o o o o o o o o

Attitudes and Values of Students‘Who

Withdraw from COllege o o o o o o o o o 0

Social Backgrounds of Students Who‘Withdraw
from College. 0 o o o o o o o o o o o o o

A Comprehensive Study of Student‘Withdrawal
from COIlegeo o o o o o o o o o o o o o o

summary 0 o o o o o o o o o o o o o o o o 0

THE MHOIDIDGY O O O O O O O O O O O O O O 0

Definition of the Population. . . . . . . .
Selection and Classification of the Sample.

ii

Page

ii

vii

viii

he

<D\O \nkn {erthb

H

H
N

H
4:-

24

32

37

Chapter

IV.

V.

InStrumentation o o o o o o o o o o o o 0
College Qualification Test. . . . . . .
Test of Critical Thinking, Form G . . .
Michigan State University Reading Test.
Inventory of Beliefs, Form I. . . . . .
Rokeach's Dogmatism Scale, Form E . . .
Differential Values Inventory . . . . .
The 300131 Status Index 0 o o o o o o o

COlleCtion Of Data. 3 o o o o o o o o o o o o o o

The Statistical Model and Computation Procedures.
Description of Multiple Discriminant Analysis .
Assumptions of the Statistical Model. . . . . .

Summary 0 o o o o o o o o o o o o o o o o o o o 0
THE ANALYSIS OF DKTA. o o o o . . . o . . o o o . 0

Results of the Multiple Discriminant Analysis
The TeSt Of the Hypothesis. 0 o o o o o o o o 0

Interpretation of the Significant Discriminant

mnetions O O O O O O O O O O O O O O O O O O O O O 0
Interpretation of the First Discriminant Function .
Interpretation of the Second Discriminant Function.
Interpretation of the Third Discriminant Function .

Group Differences On the Significant Discriminant
FunCtionS o o o o o o o o o o o o o o o o o o o o
Discrimination Between‘Withdrawal Groups
and Enrolled Groups 0 o o o o o o o o o o o o
Discrimination Between Above and Below Average
Achievement Groups. 0 o o o o o o o o o o o o

Discrimination Between Males and Females. . . .e

Group Differences on the Third Discriminant
Function. 0 o o o o o o o o o o o o o o o o 0

Summary 0 o o o o o o o o o o o o o o o o o o o 0
SUMMARY AND CONCLUSIONS 0 o o o o o o o o o o o o 0

Summary 0 o o o o o o o o o o o o o o o o o o o o
The MethOdOIOgy o o o o o o o o o o o o o o o o
The TeSt Of the HypOtheSiSo o o o o o o o o o o
The Nature of the Three Significant Discriminan

FunCtiCnS o o o o o o o o o o o o o o o o o 0
Group Differences on the Three Significant
Discriminant FunCtionSo o o o o o o o o o o 0

iii

Page

79
80
82

85
89
91
94

97

99
103
103
103
106
106

108

Chapter Page

Conclusions and Generalizations . . . . . . . . . . . 110
Differences Between‘Withdrawals and Enrollees,
Within Sex and Achievement Categories . . . . . . 110
Differences Between‘Withdrawals and Enrollees,
Regardless of Sex and.Achievement . . . . . . . . 110
Differences Among the Various Withdrawal Groups . . 113
Discussion of the Contribution of the Various
Instruments in Differentiating Among Groups . . . 114

Implications for Further ResearCh o o o o o o o o o o 1 15
BIBLIW’RAPH-Yo o o o o o o o o o o o o o o o o o o o o o 118
APPENDIX

A. Group Means for All Ten Groups and For the
Total Group on the Measures Employed in the Study. 122

B. Matrix A and Matrix'W of the Determinantal
Equation'A-AWIV=0..............125

C. Standard.Deviations of the Measures Employed
in the Study for the Total Group . . . . . . . . . 127

D. Conventionalized Discriminant Coefficients
Corresponding to the Six Discriminant Functions. . 129

iv

Table No.

1

10

11

LIST OF TABLES

Number of Students in Each Group, Classified
by Collegiate Persistence, First Term Grade
POIUL Average, and Sex 0 o o o o o o o o o o o o o 0

Pearson Product Mbment Correlation Coefficients

of College Qualification Test, Total Score, With
Grades in Communication Skills, Natural Science,
and.Grade Point Averages for Males and Females . . .

Pearson Product-Moment Correlation Coefficients

of Test of Critical Thinking Scores With Grades

in Communication Skills, Natural Science, and

Term Grade Point Averages for Males and Females. . .

Pearson ProductAMoment Correlation Coefficients

of Michigan State University Reading Test‘With
Grades in Communication Skills and Natural

Science, and'With Term-End Grade Point Averages

for Males and Females. o o o o o o o o o o o o o o 0

An Intercorrelation Matrix of the Inventory of
Beliefs, Rokeach's Dogmatism Scale, College
Qualification Test, and.Michigan State University
Reading Test for Male and Female Freshman Students

Intercorrelations of Biographical Variables
Related to the Socio-Economic Position of the
Fanily o o o o o o o o o o o o o o o o o o o o o o 0

Correlation Coefficients of the Biographical
Variables With the First Factor. 0 o o o o o o o o 0

Means and Standard Deviations of Father's Educa-
tion, Mether's Education, and Father's Occupation
For the Sample Of the StUdyo o o o o o o o o o o o 0

Pearson ProductAMOment‘Within Group Correlation
Coefficients Among the Variables Used in the Study .

Latent Roots, Chi Square Values, Degrees of
Freedom and Statistical Significance Levels
for Each of the Six Discriminant Functions . . . . .

‘Weighted Conventionalized Discriminant Coefficients
of the First Discriminant Function . . . . . . . . .

V

Page No.

1+9

52

65

66

75

78

81

Table No. Page No.

12 ‘Weighted Conventionalized Discriminant Coefficients
of the Second Discriminant Function . . . . . . . . . . . 83

13 ‘Weighted Conventionalized Discriminant Coefficients
of the Third Discriminant Function. . . . . . . . . . . . 85

1h Discriminant Function Values for Selected Combinations
of Variables and For the Final Score on the Third
Discriminant Function for Each Group and For The
TOtal Group 0 o o o o o o o o o o o o o o o o o o o o o o 87

15 Mean Discriminant Scores on the First and Second
Discriminant Functions. 9 o o o o o o o o o o o o o o o o 89

16 Group Means on the Third Discriminant Function. . . . . . 97

vi

Figure No.

1

LIST OF ILLUSTRATIONS

Page No.

Differences Among Group Means Plotted in the
Two Dimensional Space of the First Two
Significant Discriminant Functions . . . . . . . . . . . 90

Differences Between the‘Withdrawal and Enrollee
Groups,‘Within Similar Sex and.Academic Achieve-

ment Categories, Plotted In the Two Dimensional

Space of the First and Second Discriminant Functions . . 92

Differences Between Above and Below Average

Achievement Groups, Within Similar Collegiate

Persistence and Sex Categories, Plotted in The

Two Dimensional Space of the First and Second

Discriminant Functions . . . . . . . . . . . . . . . . . 95

Differences Between Males and Females,‘Within

Similar Collegiate Persistence and.Achievement

Categories, Plotted In The Two Dimensional Space

of the First and Second Discriminant Functions . . . . . 96

vii

ACKNOWLEDGEMENTS

The assistance of the members of the doctoral committee of the
writer is gratefully acknowledged. ‘William'w. Farquhar, chairman of
the committee, was of invaluable assistance from the initial planning
stages of the study through its completion. The writer is also in-
debted to professors Paul L. Dressel, Archie O. Haller, and Halter F.
Johnson for guidance in the formulation of the study and for critical
examination of the manuscript. He wishes also to express especial
gratitude to Joseph L. Saupe for consultation and guidance in the de-
sign of the study and the statistical methodology involved.

The study reported herein was part of a larger investigation of
"Critical Thinking, Attitudes, and Values in Higher Education," spon-
sored.joint1y by Michigan State University and the United States
Department of Health, Education, and Welfare, Office of Education.
The collection of data on the scale required by the design of the
study would have been impossible without the support of these two
groups. The writer is grateful to Paul L. Dressel, principal investi-
gator and Irvin J. Lehmann.research associate, of the ”C.T.A.V." study,
for initial encouragement to explore the problem of student withdrawal

from college and for continual guidance.

viii

A MULTIVARIATE ANALYSIS OF THE RELATIONSHIP OF ACADEMIC
APTITUDE, SOCIAL BACKGROUND, ATTITUDES AND

VALUES TO comma PERSISTENCE

By

Stanley 0. Ikenberry

AN ABSTRACT

Submitted to the School for Advanced Graduate Studies
of Michigan State University of Agriculture and
Applied Science in Partial Fulfillment of
the requirements for the degree of

IDCTOR OF PHILOSOPHY

Department of Education

1960

Approved

 

Stanley 0. Ikenberry

ABSTRACT

The purpose of this investigation was to contribute to the know-
ledge about students who withdraw from college prior to the completion
of the freshman year. In addition to differentiating between students
who remained and students who withdrew from college, the study was de-
signed to differentiate between withdrawals of above and below average
first term achievement, and between male and female withdrawals.

The population of the study consisted of full-time, native born,
new entering freshmen at Michigan State University, fall, 1958. FTom
the 2,746 students in the population, the 303 students who withdrew
prior to the completion of the 1958-59 academic year formed the sample
of ”withdrawals" for the study. A random sample of 250 students was
drawn from the remainder of the population for purposes of comparison,
and formed the "enrollee" sample for the study. A further classifica-
tion by first term achievement and sex was made in both the withdrawal
and enrollee samples. This classification scheme yielded ten groups.

The following measures were used in the study: (1) The College
Qualification Test; (2) Test 9;; Critical Thinki , _F_‘_g_r_r_q g; (3) m.
gag §_t_a_1_;_e_ University Reading Test; (4) The Inventory 9_f_ Beliefs, F3513
I; (5) Rokeach's Dogn_at°sm §-c_a_l_e_, 39m 2;; (6) The Differential V_a_lg_1_e_s_
Invento ; and (7) A social status index. The first three measures
dealt with various aspects of intellective ability; the fourth and

fifth were measures of attitudes of stereotypy and dogmatism; the sixth

Stanley 0. Ikenberry

'was a measure of traditional Protestant ethic values; and the final in-
dex was a combination of the social background variables of father's
and mother's educational level and father's occupational prestige. All
testing was completed during the first week of the fall term, 1958.

Multiple discriminant analysis, a statistical technique of de-
fining linear combinations of variables which maximize differences among
groups and minimize differences within groups, was used to analyze the
data. The first linear combination, or discriminant function, maximizes
the dispersion among groups, and subsequent functions maximize disper-
sion among groups with the effects of previous functions removed.

The analysis of data yielded three significant discriminant func-
tions and thus the null hypothesis, "There is no difference in intellective
ability, social background, attitudes, and values among groups of students
classified by collegiate persistence, first term grade point average,
and sex,” was rejected. The first function accounted for over 60 per
cent of the variance as defined by the variables, and was primarily an
intellective function. The second function accounted for approximately
25 per cent of the total variance and was interpreted as a social-cul-
tural function, maximizing sex differences. The third function accounted
for about 5 per cent of the total variance and was interpreted to be a
social background function, with the usual correlates of intelligence
and sex removed.

Within achievement and sex categories enrollees were consistently

higher than withdrawals on the intellective function. This function,

Stanley 0. Ikenberry

however, discriminated primarily between above and below average achieve-
ment groups. 'Hhen collegiate persistence groups were classified by
achievement level and sex, withdrawals were heterogeneous in intellective
ability.

Within achievement and sex groups, enrollees were higher on the
second, social-cultural, function than withdrawals. The main discrimi-
nation of the second function was, however, between sex groups, females
being higher than males. The third function, a social background func—
tion with sex and ability factors removed, discriminated between students
who withdrew from college and students who remained enrolled, regardless
of achievement level or sex. This function may have represented an in-
direct measure of Social-psychological motivation and other circumstan-
tial factors resultant from different social backgrounds or environments

and related to college attendance and persistence.

xii

CHAPTER I

THE PROBLEM

During the past decade an increasing volume of educational research
has been directed toward the study of students who withdraw from school.
‘While the major portion of the research has been concerned.with withdrawal
from secondary school, an increasing number of researchers have directed
their efforts toward the examination of students who withdraw from college.

From previous research it is possible to define certain educational
levels at which disproportionate numbers of students withdraw. The fresh-
man year in college is one such period. Various studies have indicated
that approximately ten to fifteen per cent of an entering freshman class
may be expected to withdraw from college sometime during the freshman
year. An additional ten to fifteen percent may be expected to withdraw
.during the summer vacation.1

The research reported herein was directed toward the study of stu-
dents who withdraw from college between September and June of their first

academic year.
Importance of the Problem

The importance of the study of student withdrawal from college is
best illustrated through consideration of the implications of withdrawal

 

1Robert Iffert, Retention and Withdrawal 9; College Students, U.S.
Department of Health, Education, and Welfare, Office of Education Bulletin,
1958, No. 1. ‘Washington: United States Government Printing Office, 1957,
p. 16.

for the individual student, for society, and for the functioning of the

educational institution.
The Individual Student

It is possible to suggest only a few of the many possible implica-
tions the act of withdrawal from college may have for individual students.
The total significance of withdrawal from college is to a degree unique'
for each student. Some broad implications, however, may be advanced.

Increasing numbers of vocations in American society require college
attendance and/or graduation as a minimal qualification for entry. In
other vocations, aspirants purportedly attend college because of an alleged
advantage of the college graduate in advancing in the particular vocation.
‘While many thoughtful educators are critical of the excessive vocationalism
present in collegiate curricula, the evidence of such an emphasis is well

established.2

If the current trend in the direction of increased educa-
tional requirements for vocational entry continues, and if the role of
colleges and universities in vocational training does not decline, it is
likely that the vocational implications of withdrawal from college will
continue to be of appreciable significance for individual students.
Ebually, if not more important, are the implications of withdrawal
from college for the non-vocational life of students. The educational

experiences of a college education are designed to broaden and deepen

understandings and interests considered by scholars through the ages to

 

2mm J. MoGrath and Charles H. Russell, 51-3 Liberal Arts Colleges
Becoming Professional Schools?, New York: Institute of Higher Education,
Columbia University, 1959.

be important to the full development of human potential. It is difficult
to weigh such outcomes of a college education, especially those more im-
portant and lasting effects which are further developed and modified over
the years subsequent to graduation from college. It can be stated, how-
ever, that early withdrawal results in student forfeiture of college ex-
periences designed to increase understanding and appreciation of the
cultural heritage imbodied in the physical, social and aesthetic world.
The research findings of this study should assist in obtaining a clearer
understanding of the student who withdraws from college and may be use-
ful in helping such students obtain maximum benefit from the college

experience.

The Society

Colleges and universities are social institutions and the functions
performed by such institutions must be viewed in relationship to the
characteristics and needs of society. American society has for many
years been moving in the direction of a non-agrarian, industrialized
and specialized economy. There is an increasing demand for greater num-
bers of individuals with more advanced skills and knowledge. It is evi-
dent from examination of current national and international problems that
the need for a highly educated citizenry has never been greater. Students
who are capable of additional educational development but withdraw from
the experiences specifically designed to assist and direct such develop-
ment represent a loss to society. Research concerning student withdrawal
may contribute to the reduction of student attrition and thus increase

the quantity and quality of the educated citizenry.

The Educational Institution

 

The study of student withdrawal from college is important to
colleges and universities for several reasons. First, it is important
because of the reaponsibilities of the institution to the individual
student and to society. Secondly, student attrition represent an in-
direct but general criteria of success of the educational program.

That is, a high proportion of student attrition may be reflective of in-
adequate admission programs, inappropriate curricular offerings, poor
orientation programs, or to any number of factors resultant from insti-
tutional malfunction. The study of student withdrawal from college may
contribute to the improvement of one or more of these institutional
functions.

The third, and increasingly more important reason why the study
of student withdrawal from college is important, is that of institutional
efficiency. As increasing numbers of students make application for ad-
mission to American colleges and universities, the efficient use of the
human and physical resources of the institution will become more and
more important.3 'The need for efficiency would imply that colleges and
universities have effective admission and selection programs designed
to admit those students most likely to profit from and succeed in the
educational program.

If the characteristics of the student who withdraws from college

during the freshman year can be identified, it is possible that measures

 

3Clarence B. Lindquist, College ggd,University Faculties, Recent
Personnel ggd'Instructional Practices, U.S. Department of Health, Educa-
tion andfwelfare, Office of Education Bulletin, 1959, No. 27,‘Washington,
D.C.: United States Government Printing Office, 1959.

can be taken to assist students who may be likely to withdraw from col-
lege in achieving maximum benefit from the educational program. As
needless student withdrawal decreases, institutional efficiency should

increase.
Purpose of the Study

The purpose of this study, broadly stated, is to contribute to a
more complete understanding of the nature of the student who withdraws
from college during the freshman year. The study is designed to define
some of the major factors which differentiate between those students who
remain in college through the end of the freshman year and those students
who withdraw during the same period. ‘Moreover, the study is designed
to differentiate among different classifications of students who withdraw
during the freshman year. Differences between students who achieve at
above and below average academic levels prior to withdrawal from college
‘will be studied. Possible differences between males and females who

‘withdraw from college will also be examined.

Rationale of the Study

 

Researchers have for some time explored the hypothesis that stu-
dents differ in the extent to which they are capable of performing the
kinds of intellective behavior generally required of college students.
This capability or intellective ability is assumed to be inseparably re-
lated both to hereditary differences and to differences in personal ex-

periences or environment prior to admission to college.

6 L

Various instruments have been developed which purportedly measure

intellective ability and over the years several of these instruments have
been found to positively correlate with success in college as measured
by grades.“

It is reasonable, therefore, to hypothesize that those students
who find it either difficult, or impossible, to perform the kinds of
intellective behavior demanded of college students will be more likely
to withdraw from college. In attempting to define those factors which
differentiate between students who remain in college and those who withdraw,
it is especially pertinent therefore to include some measure or measures
of intellective ability. One would also hypothesize that intellective
ability would be a crucial variable in differentiating between withdrawals
attaining high achievement levels and those of low achievement.

In addition to intellective ability there is reason to examine a
possible relationship between social background and persistence in col-
lege. Hollinshead has presented data which indicate a positive relation-
ship between college attendance and family income.5 He theorized that
the relationship between the various measures of socio-economic background
and college attendance reflected three phenomena. First, he suggested
that the relationship may reflect a difference in the ability to attain

the necessary levels of academic achievement. Secondly, the relationship

 

“Willard G.‘Warrington and Joe L. Saupe, "Development and Appli-
cations of’Tests of General Mental Ability,“ Review‘g§,Educational Research,
29815-25. 1959.

5Bryon s. Hollinshead, W_}_19_ Should 92 T_o_ Colle s, New York:
Columbia University Press, 1952, p. 37.

between socio-economic background and college attendance may be indica-
tive of a lack of the financial resources necessary for college attendance.
The third phenomenon suggested was that the student from the lower socio-
economic levels may not perceive college attendance as important to
current goals or future aspirations. The adolescent from the middle or
upper socio-economic level home may experience pressures from parents,
peers, and significant others to attend college. Moreover, his voca-
tional aspirations may require college attendance.

It is possible for the adolescent from the lower socio-economic
level home to experience pressures in the opposite direction. The parents
of the lower socio-economic home may not have attended college, and.may
feel it unnecessary for the son or daughter to do so. Many of the peer
group of the adolescent may not have attended college. Finally, it may
be possible (and in some cases necessary) to achieve vocational aspira-
tions immediately upon graduation from high school without college
attendance.

All of the foregoing possibilities suggest that a comprehensive
study of collegiate withdrawal should consider the social backgrounds
of students as a potentially significant predictive variable.

Some investigators have suggested the possibility that various
attitudes and values held by students are related to decisions to with-
draw from college. Dressel and Mayhew reported that the rigid, authori-
tarian student is more likely to withdraw from college than the flexible,

non-authoritarian student.6 It was theorized that the student who is

 

6Paul L. Dressel and Lewis B. Mayhew, General Education: gzplora-

tions ig,Evaluation,‘Washington, D.C.: American Council on Education, 195“.

rigid, authoritarian, and is stereotyped in his belief system has diffi-
culty in adjusting to the collegiate environment. The lack of rigid
structure in some college courses, the diversity of associates in the
residence halls, the presentation of new ideas and new patterns of thought
in the classroom, and similar factors may constitute a somewhat unpleasant
or even threatening environment to students who are rigid or stereotyped
in attitudes and personality development.

In addition to the above dimension, it is suggested by some edu-
cators that the failure and/or withdrawal of some college students is
due to a lack of seriousness of purpose, dedication to scholarship and
academic ideals, or initiative. The assertions are suggestive of possible
differences in the value patterns of students.

Spindler has hypothesized differences in American values and
character structure revolving around the acceptance of the values con-
tained in the Protestant ethic.7 Acceptance of such a value system
would.imply high regard for thrift, morality, religion, achievement and
individuality. ‘Many'individuals, Spindler asserted, do not fully accept
the values contained in the Protestant ethic, but place more emphasis on
values of sociability, group conformity, and moral relativism.

There are several reasons why such values might be related to per-
sistence in college. The most obvious, of course, is that students who
are more serious or dedicated in purpose, and more inclined to be self-
directive and resist group pressures may be able to achieve at a higher

level in college due to commitment to such a value system.

 

7George Spindler, ”Education in a.Transforming American Culture,”

Harvard Educational Review, 25:150-156, 1953.

It is also possible that the collegiate community tends to be
committed to the ”Protestant ethic." There is some evidence to indicate
that age is related to commitment to the values contained in the

Protestant ethic.8

If this generalization applies to the college com-
munity, professors and officials of the university would have values
which are nearer to the Protestant ethic values than many entering fresh—

men. Such a discrepancy may result in subsequent conflict and withdrawal.

Summary of the Mose of the Study
and the underlying Rationale

In the foregoing pages it was stated that the purpose of this
study is directed toward a study of various factors which differentiate
between students who remain in college through the end of the freshman
year and students who withdraw from college during the same period.

An equally important aim of the study is to differentiate among
different classifications of students who withdraw from college. Failure
to make distinctions among students who withdraw from college in refer-
ence to previous achievement or basic differences such as sex produces
results which are less precise than desired.9

After consideration of several possible factors which may be re-
lated to collegiate withdrawal, it was concluded that a comprehensive
study of the problem should consider such variables as intellective

ability, social background, attitudes, and values.

 

8Richard Prince, ; Study 93; 3132 Relationship Between Individual
Values ggdnAdministrative Effectiveness ig,thg_8chool Situation, (Un-
published Dissertation5, Chicago: University of Chicago, 1957.

9See the review of literature in Chapter II.

10

his approach to the problem of collegiate withdrawal reported herein
is different from previous investigations in two respects; comprehensive-
ness and methodolog. In the first respect, several investigators have
examined collegiate withdrawal in relation to academic aptitude, or atti-
tudes, or values, or social background. lone, to date, has considered
all such factors in approaching the problem. The problem presented here-
intekesintoaccountelloftheseespects.

Hethodologicslly, thestudyisuniqueinthetthisisthefirst
known ett-pt to use s. multivariate analysis technique in the study of the
problem of collegiate withdrawal. This technique, to be discussed in
Chapter Three, is especially valuable in defining the relative contribu-
tion or importance of variables in the differentiation among students who
withdrew and students who remain enrolled in college.

.. .o I

Following directly hem the previously stated purposes and re-
tionsle of the study, the subsequent lupothesis was formulatedx.

It is possible to differentiate among groups of students

classified by collegiate persistence, first term freshmen

grade point average, and sex, on the basis of intellective

ability, social background, attitudes, and values.

for the purposes of the statistical test, the above hypothesis
was stated in null form, as follows:

There are no differences in intellective ability, social

background, attitudes and values, among groups of students

classified by collegiate persistence, first term freshmen

grade point average, and sex.

The null hypothesis tests the assumption that all of the groups

of students msy be considered members of the same parent population and

11

no differences exist among groups in intellective ability, social back-
ground, attitudes, and values.

If, subsequent to the analysis of data, the null hypothesis is
rejected, the differences among groups will be exalined. Special atten-
tion will be directed toward apparent differences between students who
withdraw fro- college and those who remain in college through the end
of the freshen year. Attention will also be directed at different
classifications of withdrawals, such as ales and females who withdraw,
and students who withdraw from college with above and below average
achieve-ant records.

Thus, the importance of the study, the goals and objectives, and
specific directions of inquiry have been set forth. In the following
chapter, a critical review of the literature is presented and the find-
ings of previous investigators will be reported and smarised.

Inthethirdchapteranaccountofthenethodologyofthestudy
is presented, including a description of the sample and sapling procedure,
the instrmentation, the processing of data, and the analysis procedures.

In the fourth chapter the results of the analysis are reported.

A sumaary of the purpose of the study, the relevant research, the
methodology of the study, the findings of the study, and the conclu-
sionstobedrawnkclthestudyarepresentedinthefinal chapter.

CHAPTER II

THE REVIEW OF'THE LITERATURE

The review of the literature contains a critical examination of
research on the problem of student retention and withdrawal from college.
The findings of several studies are presented and the design, methodology,
techniques of analysis, and generalizations are examined with reference
to a set of evaluative criteria.

A systematic review of journal publications and unpublished doc-
toral dissertations over the past ten year period was made. An evalua-
tion of the research was approached by applying a set of criteria to the
studies included in the review. These criteria represent some of the
basic considerations which should be taken into account in the design
and methodology of research.1 The criteria used in the tentative evalua-
tion of the literature were as follows:

I. Definition of the Problem
A. Was the problem clearly defined?
B. was the problem stated in the form of testable hypotheses?

II. Statement of’Methodology

A. were all pertinent and distinguishing characteristics
of the sample adequately described, such as methods of
selection, size, and other pertinent characteristics

related to the statement of the problem?

 

William w. Farquhar and John D. Krumboltz, ”A Check List for
IEValuating Experimental Research in Psychology and Education,” Journal
21; Educational Research, 52:353-354, 1959.

13

B. Was a control group randomly selected and adequateLy de-
fined? ‘Was the population from which the control group
was drawn made explicit?

C. were the methods and sources of data collection clearly
stated and adequate to the problem?

III. Analysis of the Data

A.‘Was some statistical design employed in the investigation
and was it appropriate to the problem under study and the
data collected?

B. Did the analysis of data result in some statement of
probability?

IV. The Findings

Ao‘Were the findings consistent with the problem presented?

B. Did the investigator generalize within the limits of '
the study?

1. Did he generalize within the universe sampled?
2. Did he generalize within the evidence available in
his findings?
3. Did he generalize within the structure of his
hypotheses or within the stated purposes of the study?
Several limitations are apparent in any attempt to evaluate pub-
lished research. Professional journals frequently impose restrictions
on the length of manuscripts in order to increase the volume of research
studies published. This restriction in turn frequently requires that
certain details of the experiment be omitted from the published report.

It is therefore possible to criticize research unjustly. As this reviewer

1#

had no alternative but to accept the published report as reasonably repre-
sentative of the research, an advance apology is due those investigators
who are recipients of unjust criticism.

It is common for research studies in the area of college student
withdrawal to direct attention to several facets of the problem. Some
of the studies include student given reasons for withdrawal, various
background characteristics of those who withdraw, the academic ability
of withdrawals, and other related phenomena. Categorization of studies
is therefore difficult.

It ie,however, advantageous to present the research findings in
some orderly fashion. The review has therefore been divided into the
following categories: (1) student given reasons for withdrawal from
college; (2) the academic ability of students who withdraw from college;
(3) social background of those who withdraw from college; and (h) atti-

tudes and values of college student withdrawals.
Academic Aptitude and Student Withdrawal From College

The relationship between the intellectual ability of students
and subsequent withdrawal from college has received careful study by
several investigators. Margaret Ruth Smith of‘Wayne State University
conducted a study of all freshman students, both part-time and full-time,
who entered Wayne State University in the fall of 1953.2 The total group

of 949 entering freshman students was subsequently divided into three

 

2Margaret Ruth Smith, .4; Study 9__f_ First Year Drag-Outs at Wme
,State Universit , Detroit, Michigan: wayne State University, 1957
(Mimeographedi.

15

sub-groups on the basis of the length of their persistence in college.
Group I consisted of those students who dropped out during or at the end
of the figgt semester. Group II consisted of those students who drOpped
out during or at the end of the sgggnd semester. Group III was a random
sample from the total population of 1075 students who remained ig|college
for both semesters and returned to Wayne State for the beginning of the
sophomore year.

Data regarding students were collected from University records,
individuals on campus, and from a questionnaire mailed to students.
Sixty per cent of the first semester drop-outs and 88 per cent of the
second semester drop-outs returned the questionnaire.

Of the 949 students who entered wayne State University from high
school in September, 1953, thirty-three per cent dropped out at some
time during their first year or at the end of the year. Sixteen per
cent dropped out during or at the end of the first semester and seven-
teen per cent dropped out during or at the end of the second semester.
A larger proportion of part-time students dropped out during the first
year than regular or full-time students. Seventy-five per cent of group
one were in good academic standing at the time of departure, in contrast
to sixtyatwo per cent of group two and eighty per cent of group three
who were in good academic standing at the time of withdrawal. ‘When the
groups were compared in terms of number of academically superior students,
(those above a ”B" average), group three (students enrolled) had eighteen
per cent, group two had thirteen per cent, and group one had ten per cent.

Students who remained in college showed a slight superiority on

the total score of the American Council gf’Education Psychological

16

Examination. There was, however, no relationship between time of with-
drawal and the total score on the _A_._C_.§,_ Psychological Examination.

The problem under study by Smith was clearly defined. Broad
questions were presented to guide the investigation, but no hypotheses
were stated. The sample was adequately described. One of the strongest
points of the study was the fact that the investigator studied the na-
ture of the ”non-respondents“ to the questionnaire. Non-respondents
were compared to respondents on a variety of data. Thus, the investi-
gator was able to speak with some confidence concerning the represents.-
tiveness of the portion of the ample who returned the questionnaire.

Adequate control groups were selected and defined. Methods of
data collection were made explicit.

The study did have some limitations. First, Smith included both
full-time and part-time students in the sMple. Thirty-eight per cent
of the students who withdrew from Wayne during or at the end of the
first year were part-time students. The inclusion of part-time students
in the sample limited the ability to generalize the findings of the study
to the general student population of most colleges and universities.

There was an indication that a statistical model and statistical
techniques were used in the analysis of the data. The nature of this
model, the instances of its use, and the probability‘of difference state-
ments between groups were 92; stated.

Elma W. Bragg tested the null hypothesis that there are no sta-
tistically significant differences in the performance of those students

who withdraw from college and those students who remain in relation to

1?

certain selected criteria considered to be related to withdrawal.3

The population under study consisted of the ”day class" of an
anonymous university. From the original freshman enrollment of 72# stu-
dents, #19 students (58 per cent) withdrew during the first two years of
college. Three hundred five students, or 42 per cent, remained enrolled
through the beginning of the third.year. Random samples of 100 cases
were drawn from both the group who remained and the group who withdrew.
Comparison of unmatched groups served as the statistical design and
tests of the critical ratio and chi square were used as the statistical
techniques.

The findings of the study indicated that there was a statistically
significant difference in the academic performance of students who con-
tinued and students who withdrew from college on the criterion of high
school grade point average. There was also a significant difference be-
tween the two groups in relation to academic performance during the first
semester in college as measured by the first semester grade point aver-
age. In both instances, those students who remained in college tended
to have a higher high school grade point average and a higher grade point
average the first semester in college.

The quantitative section of the American Council 22,Education
ngchological Examination did not discriminate between those who with-
drew and those who remained enrolled. The linguistic section of the

A.C.E. discriminated between the two groups at the twentieth percentile,

 

3mm w. Bragg, "Study of Student Withdrawal at 'W.U.'," 2113
Journal gflEducational Ps cholo , 473199-230, April, 1956.

18

but not at the fiftieth or seventy-fifth percentiles. The Cooperative
English Eggs, total score, discriminated between those who continued
and those who withdrew at the twentieth and fiftieth percentiles, but
not at the seventy-fifth percentile. There was no relationship between
sex and continuation in college, nor between age and collegiate persistence.
The problem was clearly defined by Bragg, and a testable hypothesis
was formulated. The characteristics of the sample, including the method
of selection, and the nature of the control group were made explicit.
Methods of collection and sources of data were clearly explained. The
instruments were described, although no reliability statement was given
for any of the instruments.
The critical ratio was used to test differences in means, except
in the cases of the gagggg Psychological EXamination and the Cooperative
English.gg§t_in which the investigator used the chi square statistic to
test for differences in the per cent scoring above the selected percen-
tiles of 20, 50, and 75. The t.test of significance of difference between
two means is a more exact test of difference than is the critical ratio.
The investigator reported no limitations of the data which would have
necessitated use of the critical ratio in preference to the t,test.
Moreover, the transformation of the scores received on the A&E=§&
szchological Examination and the Coopgrative English.zg§t,into fre-
quencies scoring within a given percentile was of questionable value.
An over-all, three by two chi square test would not have yielded the
differences the investigator reports; thus, a separate chi square test
must have been made on each percentile grouping. These separate tests

would reduce the confidence one could place in the resulting statements

19

of probability of difference. The use of the t.test in this instance
would have given a more accurate statement of differences. The t_test,
however, would necessitate leaving the data in score form, and would
prohibit tests of significance within various percentile groups.

Halladay and Andrew report a study of student withdrawal from
various selected colleges in.Arkansas.# The sample consisted of the
1955-56 freshman classes of nine participating Arkansas colleges. Any
student who failed to re-enroll for the fall term, 1956. was considered
a “drop-out.” A total of 1,06? drop-outs were included in the sample.

The findings, as reported by the investigators, indicated that
the largest number of drop-outs took place after the end of the first
academic year. The percentage of male and female drop-outs was essen-
tially the same. Thirtyesix per cent of all drop-outs earned a passing
grade prior to withdrawal, while 64 per cent of the drop-outs did not
achieve at an acceptable level or left college before any grades were
assigned. The investigators reported that the analysis of scores of
freshman drop-outs on ”an ability test" indicated that 64 per cent of
the drOp-outs scored below the national media on the test. They also
indicated that 74 per cent scored.below the national median on ”an
English achievement test."

By way of evaluation, the Halladay and Andrew study possesses
many serious limitations. The problem could have been more clearly

stated and the specific goals or hypotheses of the study explicitly

 

“1). Whitney Halladay and Dean 0. Andrew, "Drop-outs from Arkansas
Colleges,“ Personnel Eng;Guidance Journal, 3?:212-13, November, 1958.

20

defined. ‘While the sample size was given, the exact nature of the sam-
ple including information as to the sex distribution, distribution among
the colleges, and collegiate class level was not stated.

No control group was selected. National norms were used for com-
parative purposes. National norms do not provide data or criteria neces-
sary to test differences between students who withdrew from Arkansas
colleges and students who remained enrolled. No description was given
of the instruments used in the study. No statistical design was dis-
cussed and no statements of probability of difference were given. Recog-
nizing the limitations of the study, the generalization of the findings
should be restricted.

Paul Hunger conducted a study on withdrawal from college of stu-
dents who were admitted to the University of‘Toledo from the upper,
middle, and lower thirds of their high school graduating classes.5 Scores
on the thg_§tgtg.Psychological Examination, administered as an entrance
examination, and grade point averages for the first semester in college
were used as criterion variables. Each student was classified as to the
length of persistence in college. After assigning students to college
persistence groups, the test scores and the first semester grade point
averages of the various groups were compared.

The analysis of data indicated that there was no significant re-
lationship between the scores on the thg.§tgtg_Psxchological Examination

and persistence in college for students who had graduated in the upper,

 

5Paul F. Hunger, "Student Persistence in College,” Personnel ggd
Guidance Journal, 353241-43, December, 1956.

21

middle, and lower third of their high school classes. That is, with
high school rank held constant, there was no significant difference in
measured academic aptitude between those students who withdrew from
college and those who remained. There were, of course, differences
among the scores of upper, middle and lower groups themselves.

There was a significant relationship between persistence in col-
lege and the first semester grade point averages of the three groups.
Hunger generalized that a more accurate prediction of attrition could
be made through utilization of both the rank in the high school graduat-
ing class and the first semester grade point average than could be made
by either of the variables alone. The investigator stated that it was
likely that middle third students would remain in college to graduation
if they had just slightly better than a "C“ average in the first semes-
ter; upper third students if they almost had a "B” average; and lower
third students if they had barely a "C” average in the first semester.

Although no hypotheses were given, the problem.was clearly stated.
Sample selection and size were made explicit. The control groups were
obtained in accord with the design of the study. The methods of data
collection were given. The group means, standard deviations, and des-
criptive information were presented in regard to the instruments used.
The analysis of variance and.t,test were used as statistical techniques
and were appropriate to the data. The investigator'was careful to
generalize within the population sampled and within the evidence avail-
able. .A high degree of confidence could be placed in the findings.

Another approach was made to the problem of collegiate withdrawal

by Line and Pitt who studied a group of 1,99# freshmen of the University

22

of‘Wisconsin.6 All of the students were high school graduates and resi-
dents of the state of Wisconsin.

From September, 1948, to June, 1952, students would have had 0p-
portunity to register for a total of eight semesters of study. The mean
number of semesters for which students were registered was 5.2. Twenty-
eight per cent of the students were registered for two semesters or less.
Only thirty-eight per cent were registered for all eight semesters. The
greatest loss of students was during the freshman year, during the first
two registration periods. The second greatest loss was during the first
registration period of the second year. Student withdrawal subsequent
to the first semester of the second year was relatively small.

The findings of the study indicated that the higher the percentile
rank of a student in his high school graduating class, the more semesters
he was likely to be registered during the eight semester period. The
product moment coefficient of correlation between these two variables was
.37 which is significant at the .01 level of confidence. Percentile ranks
on the American Council g§_Education Psychological EXamination for students
correlated .288 with the number of semesters registered. The mean per-
centile rank on the gagggg for all students in the sample was #9.3 and
ranged from 3h.“ for students registered only one semester to 57.5 for
students who were registered all eight semesters. The correlation of
the first semester grade point average of students with the length of

residence in college was .5#1 which is significant at the .01 level of

 

6L. J. Line and Hy Pitt, ”The Staying Power and Rate of Progress
of University of‘Wisconsin Freshmen," College ggdlUniversity, 29:86-99.

23

confidence. The higher the first semester grade point average of stu-
dents, the more likely they were to remain in college.

The problem was clearly defined, although no hypotheses were pre-
sented. The size and selection of the sample were described and con-
trasting groups, serving a control function, were formulated. Methods
and sources of data collection were given. The mean and standard
deviation for all the instruments were listed but no estimate of instru-
ment reliability was given. The analysis of data did result in a state-
ment of probability of relationships. Generalizations were made within
the limits of the pepulation sampled and in accord with the evidence of
the study.

The statistical model used in this study was of unusual interest.
The correlation of ability and achievement indices with length of colle-
giate persistence was a novel approach and produced interesting results.
No statement was made concerning the nature of the distribution of colle-
giate “persistence scores." It is quite likely, though, that the dis-
tribution would have been far from normal and.would have approached a U
shape. The possible effects of such a U shape distribution in inflating
the coefficients of correlation should have been discussed.

Finally, as supportive research, dissertations by Goble of Indiana
University and Hanks of the University of Arkansas are presented. Goble
of Indiana University compared the graduates of the 1950 freshman class

at Miami University to those who did not graduate.7 Students coming from

 

7Robert Irvin Goble, A Study 9_1_:‘_ 9:13 Student DroE-Out Problem gt
Miami Universit , (Unpublished Ph.D. dissertation, Indiana University,

W7.

24

the higher quartiles of the high school graduating class were found to
be more likely to graduate. The most frequently given reasons for with-
drawal included military service for men and marriage for women.

Hanks, at the University of Arkansas, studied freshman student
withdrawal and found withdrawals to have significantly lower scholastic
aptitude scores, less mastery of the English language and lower high
school grade point averages than students who remained in college beyond

the end of the freshman year.8

Attitudes and Values of Students

Who Withdraw from College

Partially because a limited amount of research has been conducted
on the relationship between student attitudes and values and withdrawal
from college, studies concerned with student given reasons for withdrawal
have been included in this section. Student opinions may, however, be
indicative of attitudes and values.

Roger Ibshino reported a study of freshman student withdrawal.9
The population consisted of 98 students who were assigned to the investi-
gator when they were first enrolled in college. Data were collected

from counseling sessions and interviews, together with a tabulation of

selected data from the file of each of the student advisees. Of the

 

8Charles Hanks, A Comparative Study 9;; Factors Related t3 Reten-
tion gag; Withdrawal 93:. Freshman Students gt. 3133 University _o_f_ Arkansas,
(Unpublished Ph.D. dissertation, University of Arkansas, 195a).

9Roger Yoshino, "College Drop-outs at the End of the Freshman
Year,“ Journal 9; Educational Sociolo , 32:1+ZJ+8, September, 1958.

25

total of 98 students, 45 had.withdrawn during or at the end of the
freshman year. The fifty-three students who remained in college served
as the control group.

The results of the unstructured interview showed that students
offered the following reasons as influential in the decision to drop
out of school: lack of preparation in high school, 51 per cent; inade-
quate finances, 39 per cent; no clear cut field of interest, 33 per
cent; poor study habits, 29 per cent; had to take courses they were not
interested in, 27 per cent; marriage, 22 per cent; discouragement with
low grades, 18 per cent; planned to attend college for only one year,

18 per cent; unhappy personal adjustment, 16 per cent; lack of academic
ability, 13 per cent; misconception of what to expect in college, 11
per cent; not enough courses in which they had an interest, 11 per cent;
could not make ”house" grades, 11 per cent; transfer to another college,
11 per cent; homesick, # per cent; and family prOblems at home, 4 per
cent.

There was a significant difference in high school grade point
average and in American Council gg'Education Psychological Examination
scores between students who withdrew and students who remained in col-
lege. Students who remained in college tended to be higher on both
measures. There was also a significant difference on the above two measures
between those students who dropped out for academic reasons and those
who withdrew because of other reasons.

The most interesting finding of the study was that there was no
significant difference between those students who withdrew from college

and those students who remained in college on the ratings given by high

26

school officials to the following questions: (1) How well does he get
along with others? (2) Does he need constant prodding or does he go
ahead with.work without being told? (3) How'emotionally stable is he?
(h)‘What kind of a citizen is he? (5) Has he a program of definite
purposes in terms of which he distributes his time and energy? (6)‘What
are his chances of success in college?

Although the problem was adequately defined, no testable hypotheses
were presented. Methods of data collection were made explicit and were
adequate to the problem presented. A.t,test and chi square anaLysis
were reported as the statistical tools utilized and were apprOpriate to
the nature of the data. There is some question as to the nature of the
sample. The rate of withdrawal for the end of the freshman year of #6
per cent was an unusually high figure. Other investigators have con-
sistently reported less than twenty per cent attrition garigg.the fresh-
man year..

The author did overgeneralize beyond the data presented in the
discussion of the findings. The generalizations were made that students
who matriculate into college have better than average personalities; and
that freshmen are immature in a number of ways, and need guidance and
support from their families, instructors and upperclassmen. The design
of the study, the instruments used, and the actual data collected.wou1d
not appear to be supportive of such a statement.

Harry A. Grace reported one of the few attempts to examine the
relationship between personality characteristics and withdrawal from

college.10 He tested the hypothesis that independent-responsible students

 

1OHarry A. Grace, ”Personality Factors and College Attrition,"
1292.61 M .°_f. ____Educati°n. 35:36-40. July. 1957.

27

would be least likely to drop out of college and dependent-irresponsible
students most likely to leave.

Freshman students who entered the college during 195# and 1955
‘were studied. Every student who withdrew from college for whatever rea-
son was considered in the attrition group. A total of 107 students from
both classes withdrew before April, 1957. A control group of 107 stu-
dents who continued was matched with the attrition group with regard to
measures of academic aptitude.

The Tails}; Manifest Anxiety §_c_:a_l_e_ and Gough's scales measuring
dominance and reaponsibility were used as test instruments. A measure
of men's participation in college life was formed through the use of a
list of men who received athletic numerals or sophomore varsity letters.
Participation in the voluntary military training program was also used
as a measure of participation in college life.

The analysis of data, according to the investigator, indicated
that those who remained in college earned more college credits. The men
who remained enrolled were more active in campus varsity athletics and
AFROTC than those who withdrew. The investigator claimed that the data
supported the hypothesis that independent-reSponsible students would be
lgagt’likely to drop out of college and dependent-irresponsible ones
most likely to leave. The investigator then confused the issue by stat-
ing that when attrition and control groups were compared, only women
who remained in college were more responsible.

The problem was clearly defined and testable hypotheses were
formulated. A control group was selected and matched according to

level of ability with the attrition group. Methods of data collection,

28

however, were not made explicit. Statistical techniques were apparently
used, but no mention was made of the particular techniques used. The
analysis did result in a statement of probability of group differences.

Because the entering classes of two separate years were used,
some members of the pOpulation were enrolled in college for longer periods
of time than others, and thus would have had.more opportunity to withdraw
from college. If, as the author claims, there were original differences
between the two entering freshman classes in dominance and responsibility,
this would have possibly caused the subsequent withdrawal group to be
atypical on the same measures.

Notwithstanding all of the possible reasons for attaining invalid
but statistically significant results, the investigator did not present
any data to support the main hypothesis of the study; The mere fact
that within the withdrawal group a slightly larger number of students
appear to be dependent-irresponsible rather than independent-responsible
is no support for the hypothesis that independent-responsible students
are least likely to drop out of college. A.comparison with a control
group of those who remain in college would be necessary to make any such
assertion. Due to the above limitations, it is impossible to accept the
conclusions of the investigator.

C. A. Berry and.A. L. Jones conducted a study to reveal reasons
students gave for withdrawal from Grambling Cbllege during the 1953-54
academic year. ‘A total of 151 freshman students severed connections

with Grambling College during or at the end of the year.11 A questionnaire

 

110. A. Berry and A. L. Jones, "Factors Involved in the‘Withdrawal
of Students from Grambling College At or Before the End of Their Fresh-
man Year,” Journal gfiNeggo Education, 25:No. b:4#5—h7, 1956.

29

‘was mailed to each of the 151 former students. Follow-up letters were
sent to the 128 students who did not reply to the first questionnaire.
In toto, 71 students, or 47 per cent of the total withdrawal group,
returned the questionnaire.

Of those who returned the questionnaire, #9 per cent had com-
pleted two semesters, 27 per cent had completed one semester, and 2a
per cent had withdrawn before the first semester had ended. Of the 76
per cent who had completed at least one semester at Grambling College,
15 per cent had attained honor status and 76 per cent earned.between a
”C" and "B" grade point average.

The majority of the students expressed positive feelings and at-
titudes toward Grambling College. Results of the questionnaire indicated
that lack of finances was the outstanding reason precipitating the with-
drawal of students. Other reasons included were dislike of the college,
marriage, "to join the Armed Service,” family difficulties, pregnancy,
and health.

The purpose of the Berry and Jones study was stated in general
terms, and the sample and methods of data collection were described.

Recognizing the difficulty involved in attempting to receive a
near 100 per cent return from mailed questionnaires, the ability to
generalize is limited when less than half of the original sample is re-
tained in the final analysis. Moreover, the investigators did not make
any statement as to the nature of the remaining 53 per cent of the origi-
nal sample who did not return the questionnaire. The assumption of
similarity of reSpondents and non-respondents without some comparative

data would seem unjustified when studying collegiate withdrawal.

30

No control group was selected and apparently no statistical
model was applied to the data. Thus, no statement of probability of
difference between those who remained in college and those who withdrew
from college during or at the end of the freshman year could be made.

Holmes studied voluntary freshman withdrawal from the College
of Liberal Arts of Syracuse University in order to obtain student given
reasons for withdrawal and other characteristics of the withdrawals.12
Out of the entering freshman class of 1050 students in the College of
Liberal Arts, it was found that 169 students withdrew by the end of the
freshman year, including 55 males and 114 females.

Questionnaires were mailed to all students who withdrew from
college and completed questionnaires were received from #h per cent of
the sample. Fifty-six of the female withdrawals (49 per cent) and 19
of the males (3h per cent) returned the questionnaire. Freshman student
population norms of the College of Liberal Arts at Syracuse University
were used in the comparison of biographical characteristics of students
who withdrew with the characteristics of the total pOpulation.

The findings of the study indicated that the students who with-
drew were more likely to come from.the local area (i.e., the local city,
county, and state) than the total freshman population. Various aspects
of the services of the college came under criticism in the reSponses of

the questionnaire, including the food services of the college and academic

 

12Charles H. Holmes, "Why They Left College: A Study of Voluntary
‘Withdrawal from the College of Liberal Arts at Syracuse University,‘I
College §g§,Universit , 34:295-300, September, 1959.

31

counseling. The investigator stated that the main student given reason
for withdrawal was simply that students wanted to attend another college
or university. Some 63 of the 75 (respondents) listed the fact that
they did transfer to another college.

The Holmes study has definite limitations. A clearer statement
of the problem, including a statement of hypotheses or a discussion of
specific aims of the study would have been desirable. The main limita-
tion of the study was the sex bias of the sample, in addition to the
inability to obtain a larger percentage of returns on the questionnaires.
The large percentage of respondents who indicated they transferred to
another institution points up the likelihood of some bias in question-
naire returns and a resultant bias in the sample used in the analysis.
The number of males in the final analysis was 19, and thus limited the
ability to generalize to collegiate withdrawals in general. ‘While the
freshman student population was used for comparative purposes, the selec-
tion of a control group consisting of students who remained in college
would have been desirable. No statistical design was reported and no
statements of probability of differences were given.

‘Wooster and Stover studied the entering freshman class of a
college of education to determine the reasons for withdrawal of those
students who dropped out of the college.13 A letter was sent to each
student who withdrew, asking the reasons for withdrawal from the college

of education. A reply card was enclosed. Of the 55 per cent who returned

 

13t}. F. Wooster and‘W.‘W. Stover, "Lost: Students,“ Educational
Research Bulletin, 37:85-90, April, 1958.

32

the card, marriage was the reason given by more than half of the re-
spondents. Transfer to another institution, dissatisfaction with
college, "personal" reasons, military service, and work were also given
as reasons for withdrawal.

The sample of the‘Wooster and Stover study should have been more
clearly defined. Also, the prominence of marriage as a reason for with-
drawal would indicate a strong sex bias in the sample. No mention is
made, however, of the number of males and females in the sample.

No control group was selected. No mention was made of the use
of any statistical model in the analysis of the data and, of course,
no statements of probability of differences were given.

The investigators generalized beyond the findings of the study
if the data reported may be considered complete. A recommendation was
made for individual counseling of students and it was asserted that in-
dividual counseling would reduce the number of withdrawals. While this
is a logical hypothesis, and.worthy of study, the design of the study

and the data collected do not merit such generalization.

Social Backgrounds of Students

Who Withdraw From College

The relationship between collegiate persistence and various
social background factors such as level of parental income and educa—
tion, social status of the father's occupation, and other sociological
variables which may influence the social importance and motivation of
college persistence is of particular relevance to the general problem

of college student withdrawal.

33

W. L. Slocum conducted a study concerning the withdrawal from
college of students who entered the State College of Washington as fresh-
men in the fall of 1951, 1952, and 1953.14 Transfer students were ex-
cluded. Drop-outs were defined as those students who were not enrolled
on April 25, 1955. A sample of the total population under study was
randomly drawn. This yielded a sMple of 1,019 students which included
1465 enrolled students and 5 51+ drop-outs. Completed questionnaires were
obtained for 99 per cent of the enrolled students and 81 per cent of
the drop-outs. Data were gathered from the counseling center records
and from the office of the registrar.

The findings of the study indicated that there was seldom a single
cause for withdrawal as perceived by students. Most of the drop-outs
identified two or more problems as important or very important in con-
nection with the decision to drop out of school. Furthermore, the data
of the study indicated that the enrolled students met many of the same
problems as the drop-outs. Eighty-two per cent of the drop-outs identi-
fied two or more important reasons for withdrawal, and more than two
reasons were identified by 61 per cent.

Fifty-one per cent of the drop-outs were in academic difficulty
at the time of withdrawal. Only 31 per cent, however, acknowledged
that low grades had been either important or very important to them in
connection with their decisions to withdraw. This finding was inter-

preted as an indication that low grades frequently may be regarded by

 

14W. L. Slocum, ”Social Factors Involved in Academic Mortality,”
College and University, 32:53-61», 1956.

34

students as the reflection of other problems which operate to depress
grades rather than as an indication of lack of intellectual capacity.
There was, however, a significant difference between students who with-
drew from college and students who remained in relationship to rank in
high school graduating class and scores on the American Council 9}; Ed};-
M Psychological Examination.

Both women who were enrolled and women who withdrew from college
were dependent upon parents for the bulk of their financial support
while in college. The men drop-outs, however, could be described as
being more dependent upon parental support and personal savings than
males who remained in college. The males who remained in college were
more dependent on part-time summer employment.

It was found that the family background factor which had con-
siderable bearing on chances of academic survival was the educational
level of the parents. Generally, the higher the educational level of
parents, the higher the probability of survival of the student.

There was some evidence in the pattern of the responses of the
enrolled students and of the drop-outs that parental interest as per-
ceived by the student was of great importance in connection with the
decision to remain in college or to leave. This was indicated by the
fact that while 81 per cent of the enrolled students answered, "Yes,
very much so,” to the Question,”Do your parents want you to finish
college,” only 3 5 per cent of the drop-outs answered "Yes, they were
very much concerned,” to a similar question, "Did your parents encour-

age you to remain at NBC?"

35

The problem was broadly defined, although there were no stated
hypotheses. The instruments used and the methods of data collection
‘were made explicit. Slocum implied the use of statistical tests, but
did not make explicit which techniques were used, nor the instances of
use. Statements of probability of difference were given.

The most serious limitation of the study was the fact that the
investigator failed to make clear the precise nature of the sample.

Some of the students included in the sample would have had the oppor-
tunity to be seniors at the time when the sample was drawn. Other stu-
dents would have been in school for less than two years. The results
of the study would have been more precise if a more thorough explana-
tion of the nature of the sample had been given, including the length
of residence and the class level of all members of the drop-out sample.

Three doctoral dissertations were reviewed which presented data
relative to the relationship between social background and collegiate
persistence. Smoke compared graduates and non-graduates of the 1951
graduating class of Indiana University on selected variables.15 Smoke
found that the parents of graduates had more education, that the greatest
mortality occurred among students whose fathers were farmers or farm
managers, and whose mothers were managers, operatives, or service workers,
and that the greater the participation in high school activities, the
more likehy students were to graduate. Graduates also had higher academic

scores and higher reading scores than non-graduates.

 

15Eileen Smoke, A_Comp§§ison.g£,the Graduates and Non-Graduates

93; the Class _o_f_ 1251 at Indiana Univers'i'? (Unpublished-Ph.D. disserta-
tation, Indiana University, 1955).

36

Lehr, in a study of student withdrawal at Northwestern State
College, found that the first semester grade point average of those who
remain in college was significantly higher than those who withdrew from
college.16 Lehr, however, found no relationship between withdrawal
from college and age, distance of the student's home from the college,
English and reading ability, parental occupation, and level of intellec-
tual ability.

Certain non-intellectual factors and their relationship to col-
lege student withdrawal were studied by Hood.17, Groups of freshmen who
attended a summer counseling camp sponsored by the university were given
an extensive questionnaire. The responses of students who completed
the first two years of college were compared with the responses of stu-
dents who withdrew from college prior to the end of the first two years.
Those students who expected to have many friends in college and those
who had had some previous leadership experience had a lower attrition
rate than those who expected to have only a few friends and who had had
no previous leadership experience.

Among the males, age, occupational level of the father, educa-
tional level of the father, rank in high school class and varsity ath-

letics were found to be related to withdrawal from college. Those

 

léfiilton‘W. Lehr,.A Statistical Description_ of the Factors Related
to Drog:Qgts_ and NonADropaOuts at Northwestern State Colle e, (Unpub-
lished Ph.D. dissertation, University of Oklahoma, 19

lelbert Bullard Hood, Certain Non-Intellectual Factors Related
tg_Student Attrition gt_Cornell Universit , {Unpublished Ph.D. disser-
tation, Cornell University, 1957).

37

students who did not know whether it was proper for college students to
disagree with their parents, and those who answered "don't know" to
questions dealing with parental attitudes and Opinions had a signifi-

cantly higher attrition rate.

A Comprehensive Study of Student Withdrawal

From College

One of the most comprehensive and well-known attempts to study
student retention and withdrawal in colleges and universities was directed

by'Robert E. Iffert of the United States Office of’Education.18

The
study was designed to answer three questions: (1)'What is the rate of
student drop-out in relation to type of institution, economic status of
the family, motivation of the student, academic performance, amount of
self-help, participation in extracurricular activities, and residence
of the student? (2)‘What reasons do students give for transferring to
other institutions and for discontinuing college attendance? (3)‘What
implications do the characteristics of students and their mobility have
for higher educational institutions with reference to recruitment,
selection, admission, counseling, instruction, scholarship aid, and
other functions?

All of the approximately 1,600 eligible institutions in the United

States were classified.by type of institution and.by source of control.

A stratified random sample of 177 institutions was selected. 'Due to

 

18Robert E. Iffert, Retention g3djWithdrawal 9f,College Students,

U. S. Department of Health, Education andTWelfare, Office of Education,
Bulletin 1958, No. 1, washington: United States Government Printing
Office, 1957, p. 1.

38

various factors, the number of institutions participating in the study
and supplying complete data was eventually reduced to 1#7.

The students selected for study from the participating institu-
tions included full-time freshmen for the fall of 1950. Part-time
students, foreign students, students who were married at the time of
first registration, and students who were in attendance under the pro-
visions of the legislation for veterans' education were excluded.

In addition to the collection of data from the respective insti-
tutions, all students in the study sample were asked to fill out and
return a questionnaire. In total, more than 75 per cent of the gradu-
ates responded to the questionnaire and more than 60 per cent of the
non-graduates reaponded.

The rate of graduation, in regular progression from the insti-
tution of first registration, was 39.5 per cent, with 38.8 per cent of
the men and #0.5 per cent of the women graduating. This difference
between sexes in relation to rate of graduation was not statistically
significant. The investigator estimated that if all students who eventu-
ally graduated from some institution of higher learning were included,
the percentage would rise to approximately 60 per cent. Thus, it was
estimated that approximately #0 per cent of all entering freshmen never
finish the four year degree at any institution at any time.

Iffert attempted to explore the hypotheés that the stronger a
student's motivation, the better his chances of remaining in college.
It was reported, however, that every effort to find such an association
of sufficient magnitude to support the hypothesis by comparing the

ratings of reasons for going to college and the records of persistence

met with failure.

Moreover, students who withdrew from college did not identify
unsatisfactory college facilities and services by their ratings in a
‘way that would set them apart as a distinct group from those who re-
mained in college. The mean rating of many facilities and services
decreased in level of satisfaction as length of attendance increased.

It was concluded that withdrawal from college is associated not so much
‘with dissatisfactions as with inability, or unwillingness, to endure
dissatisfactions.

The most important reasons given by males for leaving college
'were military service, lack of interest in studies, financial diffi-
culties, discouragement by low grades, difficulties with college level
work, employment Opportunity, and probation for academic reasons. ‘wo-
men rated the planning of marriage, finances, full-time work, lack of
interest in studies, and discouragement by low grades as the more imp
portant reasons for leaving college.

The findings of the study indicated that there was a relationship
between family income and persistence in college. Students from higher
income families tended to remain in college longer. There was, however,
no relationship between student part-time work in college and.with-
drawal from college. Differences between the several attendance groups
did not provide convincing evidence of the probability that earning part
or all of college expenses seriously affected persistence in college.

‘When rank in the high school graduating class of college students
was examined in relation to withdrawal or persistence in college, it was
found that a significant relationship existed. The higher the rank in

the high school graduating class, the greater the prObability of graduation

40

from college. There was also a relationship between withdrawal from
college and the standing on the college placement tests. The higher
the standing on the placement tests, the greater the chance for graduap
tion from college. It was stated, however, that the rank in the high
school graduating class was a much better predictor of graduation than
the standing in the placement tests.

It was interesting to note that if admission officers had allowed
entrance to only the top fifth of the high school classes, 144 per cent
of this select group would have failed to graduate from the institution
of first registration in the normal four year period.

Considering the scope of the research, the problem was excellently
defined. Although no testable hypotheses were formulated, general ques-
tions were asked and these questions served as a guide to the collection
and analysis of the data. The size and the methods of selection of the
sample were adequately described. This was a particularly important
factor in a study attempting to make generalizations relative to all
types of institutions in all parts of the country.

Control groups were selected, in a sense, as withdrawal students
were classified as to the term of withdrawal, up to and including gradua-
tion. The analysis and reporting of the data followed closely the pur-
poses set out for the study. Generalizations were made within the limits
of the study. The limitations of the sample, the questionnaire, and
the findings of the study were made explicit.

A more detailed explanation of the statistical techniques used
in the study would have been desirable. In some instances, the analysis

of the data did result in some statement as to the probability of

41

difference. In other cases, no statement of probability of difference

was given.

Summary

The foregoing review of the literature consisted of a critical
examination of the research literature concerned with the problem of
student retention and withdrawal from college. Several studies were
presented and the hypotheses, methodology, techniques of analysis,
findings and generalizations were examined relative to a set of evalua-
tive criteria.

The main conclusions to be derived from.the review of the
literature are as follows:

1. The evidence strongly supports the conclusion that student
status on various estimates of academic aptitude is related
to persistence in college. A high level of attainment on the
various cognitive measures increases the probability of the
student remaining in college and low attainment on the cogni-
tive measures decreases the probability of remaining in
college.

2. There is some tentative evidence to suggest that measures
of academic aptitude distinguish between students who with-
draw from college during the freshman year, and students who
withdraw from college subsequent to the freshman.year.

3. The relationship of student attitudes and.values to
withdrawal from.college remains unclear. Although the limited
amount of research conducted in this area is suggestive of

the possibility of such a relationship, additional evidence

is necessary before any relationship may be claimed or
disclaimed.

4. Students offer a great variety of reasons explaining with-
drawal from.college. Each student typically has several rea-
sons for withdrawal. Interestingly, the majority of student
reasons for withdrawal from.college are not centered in the
institution, but directed toward various phenomena outside of
the institution. One investigator, in fact, reported that
there was a tendency for satisfaction with various aspects of
the collegiate environment to decrease as length of college
attendance increased.

42

5. There is sufficient evidence to suggest a relationship
between collegiate persistence and the social backgrounds of
students prior to admission to college. Parental socio-
cultural status as reflected in such factors as parental in-
come, occupation, and educational level appears to be related
to collegiate persistence.

6. There has been little effort to examine the interrelap
tionship between academic aptitude and other significant
variables in distinguishing between those students who remain
in college and those students who withdraw. It is possible,
for example, that the relationship of attitudes, values, and
social background factors to collegiate withdrawal may be due
primarily to instrument contamination with cognitive factors
and only secondarily, if at all, to the trait purportedly
measured.

7. Little attention has been directed toward possible dif-
ferences between students with high academic achievement
records who withdraw from college and students with low
achievement records who withdraw from.college.

8. The research concerned with the study of student retention
and withdrawal has various limitations which should be con-
sidered in future examination of the problem. Some of the
more common limitations of the studies reviewed in this chap-
ter‘were:

a. A clear definition of the nature of the student under
study is necessary if the results of the investigation

are to be generalized to other groups of students and

other institutions. At a minimum, it would appear neces-
sary to define the collegiate class level status, the

sex distribution of the sample, and qualifications which
may have been placed on the sample such as number of
credits carried or academic achievement prior to withdrawal.

b. In most instances, it is necessary to select a control
group which is comparable to the withdrawal group of
students. The method of selection of the control group
and the nature of the students in the control group should
be defined.with clarity.

c. If subjects are dropped from the sample because of a
lack of complete data, attention should be given to the
possibility that the sample used in the analysis may have
been biased thereby. Studies utilizing a questionnaire
commonly face this problem.

d. The use of appropriate statistical techniques allows
the investigator to generalize the specific findings of
the study to the general population.

43

e. Generalizations should be confined to the data and
findings of the study. Generalization beyond the data
may be misleading to other investigators who may wish to
conduct further research, or to those who in some manner

‘wish to apply the findings of the study.
The findings resultant from the review of the literature were
used in planning the design and methodology of this study. The account

of the methodology of the study is presented in Chapter III.

CHAPTER III

THE METHODOLOGY
Definition of the Population

The population under study consisted of Fall 1958 entering
Michigan State University freshmen. Three thousand two hundred sixteen
students at Michigan State University were classified as freshmen dur-
ing the fall registration period.

Certain restrictions, however, were placed on the population in
order to best achieve the goals of the study. The following types of
students were excluded from the pOpulation under study: (1) all stu-
dents who had previously attended any college or university; (2) all
foreign students; (3) all part-time students enrolled for less than 12
credits of study} and, (h) students whose test data were incomplete or
unusable.

'With these qualifications observed, the restricted population of
the study numbered 2,7h6 students, including 1,436 males and 1,310 fe-
males. The sample of students used in the analysis was selected from
the restricted population as defined above, and any future references
to the population of the study, or generalizations and conclusions to
be drawn from.the results of the analysis should be interpreted in terms
of the restricted pOpulation.

45

Selection and Classification of the Sample

Every student from the restricted population who withdrew from
college subsequent to the September 1958 registration period but prior
to June 17, 1959, was identified and classified as a "withdrawal” for
the purposes of this study. Three hundred three students fram the re-
stricted pOpulation withdrew during the 1958-59 academic year, and all
303 students were included in the sample of the study.

A sample of 250 students was randomly drawn from the total group
of 2,##3 students from the restricted.population who remained enrolled
through the end of the freshmanyear.1 This group of 250 students was
included in the sample of the study'and will hereafter be referred to
as ”enrollees.”

After the initial selection and classification of the sample on
the basis of collegiate persistence, all members of the sample were
classified.hy level of first term achievement. Students who achieved
first term.grade point averages of 2.00 (C) or above, were separated
from students who achieved first term grade point averages of 1.99 (C

minus) or below.2

 

1Students at Michigan State University are assigned student num_
bers. Each number consists of six digits. The student number of each
student was recorded on the International Business Machine card, and
cards were electronicly sorted on the last two digits of the student
number. Such a procedure completely randomized the cards. The top
twenty five student cards in each of the ten machine ”pockets” were
selected to comprise the sample.

2A letter grade of A, B, c, D, or F is usually reported for each
course taken by the student. A.term grade point average is computed by
the assignment of point values to each of the letter grades, letting
A=#, B=3, C=2, D=1, and.F=0. The point value of the letter grade is
multiplied'hy the number of credit hours of the course, and the sum.for
all courses is divided'hy the total number of credits for courses taken.
Such computation yields grade point averages ranging from #.00 (A) to
0.00 F .

#6

Forty-eight students of the withdrawal group withdrew from the
university during the first term, prior to the assignment of any course
grades. These students, of course, had no grade point average. Be-
cause the 0.00 grade point average of these students could not be as-
sumed to be a reasonable estimate of past or potential achievement,
students with such a grade point average were classified into a third
achievement level category for purposes of analysis.

A further classification was made on the basis of sex. The final
separation yielded a total of ten groups. The nature of these groups
is defined in Table 1.

TABLE 1
NUMBER OF STUDENTS IN EACH GROUP, CLASSIFIED

BY COLLEGIATE PERSISTENCE, FIRST TERM
GRADE POINT AVERAGE,.AND SEX

 

 

 

First Term Grade ‘Withdrew from Enrolled in
Point Average Range College College
Males Females Males Females

 

Group 1 Group 2

 

Group 3 Group # Group 7 Group 8
0.01-1.99 N=116 =68 N=55 N:34

 

Group 5 Group 6 Group 9 Group 10

 

It is apparent from Table 1 that groups one through six consist
of students who withdrew from college sometime during the freshman year

and that groups seven through ten contain students who remained in

1+7

college through the end of the freshman year. Students in groups three,
four, seven and eight are similar to the extent that all received a
first term grade point average of less than 2.0 (C), while students in
groups five, six, nine and ten received first term grade point averages

of 2.0 (C), or more. Groups ranged in size from 20 to 116 students.
Instrumentation

In order to achieve the purposes of the study, it was necessary
to obtain measurements of the academic ability, attitudes, values, and
social backgrounds of all students in the sample.

The following instruments were selected to measure the above
characteristics:

I. Academic Ability

A. The College Qualification Test (Psychological
Corporation .

B. Test of Critical Thiw in , Form G, (American
Council on Education).

C. Michigan State Universit Reading Test, Michigan
State University).

II. Attitudes
A. $113 Inventog 9_f_ Beliefs, Form .I_, (American
Council on Education .
B. Rokeach's Doggatism Scale, Form E, (Professor
Milton Rokeach, Michigan State -University).

III. Values
A. The Differential Values Invento , (Richard Prince,
University of Chicago).
IV. Social Background
A. Biographical Data Sheet, (Michigan State University).
Each of these instruments will be discussed.with special attention
given to the origin and purpose of the instrument, evidences of validity

and reliability, and other distinguishing characteristics of the instrument.

lelege Qualification Test

The College Qualification Eggtlconsists of a series of three
ability tests which may be combined into a comprehensive total score.3
The instrument was designed to be predictive of college success as
measured by grades. The three ability tests included are in the areas
of verbal facility, numerical ability, and general information.

The verbal test consists of 75 vocabulary items, 50 of which re-
quire identification of synonyms and 25 which require identification of
antonyms. There is a time limit of 15 minutes. The numerical ability
test contains 50 items drawing on arithmetic, algebra and geometry. It
was designed to measure conceptual skill and not computational or cleri-
cal speed. The time limit for this test is 35 minutes. The third test
of the College Qualification :2 E series measures general information.
The information test is composed of 75 items from a broad range of sub-
ject matter areas, half of the items dealing with physical, biological
and chemical sciences and half with social studies such as history,
economics, geography, and government. The time limit for this test is
30 minutes.

The total score of the College Qualification Eggtlconsists of the
sum of the scores on the verbal, numerical and information tests. Be-
cause of the greater general predictive power of the total score, the
single total score of the College Qualification Tg§3.was used in prefer-

ence to the three separate sub-tests.

 

3'George K. Bennett, Marjorie G. Bennett,‘Wimburn L.‘Wallace and

Alexander G.‘Wesman, College Qualification TestsI ManualI 1252, New York:
The Psychological Corporation, 1957.

#9

The "Study of Critical Thinking, Attitudes, and Values in Higher
Education" reported data regarding the relationship of College Qualifica-
tigg_Tg§t,to freshman student grades in the Basic College freshman courses
of Communication Skills, Natural Science and also to term grade point
averages.“ The findings of this study can serve as an indication of
test validity. The data are presented in Table 2.

TABLE 2
PEARSON PRODUCT MOMENT CORRELATION COEFFICIENTS OF COLLEGE
QUALIFICATION TEST, TOTAL SCORE,‘WITH GRADES IN

COMMUNICATION SKILLS, NATURAL SCIENCE, AND
GRADE POINT AVERAGES FOR MALES AND FEMALES

‘—L

 

 

 

 

Males Females

Communication Skills

First Term .66 , .60

Second Term .57 .61

Third.Term .hB .51
Natural Science

First Term .61 .6h

Second Term .55 .51

Third Term .38 .h9
Term Grade Average '

FirSt Term .60 057

Second Term .51 .41

Third Term .34 .37

 

Correlation coefficients between the College Qualification Test
scores and these achievement criteria ranged from .3# to .66. ‘While it

is evident that the magnitude of the correlation coefficients tended to

 

#Irvin J. Lehman and Stanley 0. Ikenberry, Critical Thinking,

Attitudes and Values In Higher Education: A Preliminarngepgrt.
(Paul L. Dressel, Principal Investigator),- East Lansing: Office of

Evaluation Services, Michigan State University, 1959 (Himeographed).

50

decrease from term to term, it is also evident that the test, as the
publisher claims, is broadly predictive of collegiate success when suc-
cess is defined in terms of grades. The fact that correlation coeffi-
cients tend to decrease from term to term may be due to several factors
including a narrower range of ability in the latter terms due to accel-
eration and withdrawal of some students, and possible fluctuation and
change in student ability from the time of measurement in the fall.
The evidences of validity reported by theiest publisher in the manual
are in general agreement with the data presented in Table 2.5

Estimates of the reliability of the College Qualification.lg§t,
total score, are given in the test manual and include coefficients of
.97 for males and .96 for females.6 These reliability coefficients
were determined through the Split half method, a comparison of differ-
ences between scores on the odd and the even items of the test.

A test of the reliability of the College anlification Egg;
using the restricted population of the study was conducted. Applica-

tion of KuderARichardson formula 20 yielded a reliability coefficient

0f 093- .
The Test of Critical Thinking, Form G

The Test 9}; Critical Thinkin , Form Q, was developed as a part

of the Cooperative Study of Evaluation in General Education.7 In order

 

5George K. Bennett, Marjorie G. Bennett,‘Wimburn L.‘Wallace, and

Alexander G. wesman, College Qualification Tests, ManualI 1252, New York:
The Psychological Corporation, 1957.

6Ibid.
\ 7Paul L. Dressel and.Lewis B. Mayhew, General Education: Egplora-

$§§E§.Ifl.Evaluati°Q-‘Washington, D.C.: American Council on Education,

51

to reduce critical thinking ability to a measurable entity, emphasis
was placed on problem solving ability. Questions were designed to
measure the ability to recognize the existence of a problem, to define
the problem, to select information pertinent to its solution, to recog-
nize assumptions, to make hypotheses, to draw conclusions, to judge the
validity of the conclusions, and to evaluate the conclusions in life
situations.8

‘While it is debatable that grades serve as the best ultimate
criteria for testing the validity of a measure of critical thinking,
the validity of any test is defined in terms of the purpose for which
it is used. In this study it was necessary to obtain a variety of
measures which would serve as indicators of academic ability. In such
a context, the criterion of grades is appropriate.

The Pearson product-moment correlation coefficients of the Zg§t_
2;,Critical Thinking scores with term-end grade point averages and with
grades in Communication Skills and Natural Science are presented in
Table 3.9 The correlation coefficients ranged from .21 to .57. It is
evident that those qualities and abilities measured by the Eggtugf
Critical Thinking are related to achievement in the two Basic College

Courses and to term-end grade point averages of freshmen.

 

8C00perative Study of Evaluation in General Education, Paul L.
Dressel, Director. Instructor's Manual fgg'thg_Test g§_Critical Think-

igg, Form.g, The American Council on Education, Committee on Measurement
and Evaluation, 1953 (Mimeographed).

9Irvin J. Lehmann and Stanley 0. Ikenberry, Qp.'g;t., p. 62.

52

TABLE3

PEARSON PRODUCT -MOMENI‘ CORRELATION COEFFICIENTS OF TEST OF
CRITICAL THINKING SCORES WITH GRADES IN COMMUNICATION
SKILIS, NATURAL SCIENCE, AND TERM GRADE POINT
AVERAGES FOR MALES AND FEMAIJB

 

 

 

Males Females

Communication Skills

FirSL Term 01'“? 039

Second Term .42 .45

Third Term .35 .39
Natural Science

First Term .54 .57

Second Term .39 .llllr

Third Term .31 .44
Term Grade Averages

FiI‘St Term .48 045

Second Term .37 .37

Third Term .21 .34

 

The manual for the Test 2; Critical Thinking reported estimates

 

of reliability of the instrument ranging from .71 to .89.10 A reli-
ability coefficient of .79 was found when the Kuder-Richardson formula
20 was applied to the Test 9; Critical Thinking scores of the restricted

population here under study.

Michigan State University Reading Test

The Michigan State University Reading Test was developed by
the Office of Evaluation Services of Michigan State University. The

 

10Cooperative Stucbr in Evaluation in General Education, Paul L.

Form Q, The American Council on Education, Committee on Measurement and
Evaluation, 1953 (Mimeographed).

53

items and passages included in the revised form of the test were selected
on the basis of the empirical relationship of the items to academic suc-
cess. Furthermore, the forty-five item test was designed to measure the
ability of students to comprehend thoughts expressed in reading passages
which were representative of textual materials found in several academic
areas.

The ability to read should be closely related to the ability to
do college level work. Thus, it would.be expected that there would.be
a significant correlation between grades in college and reading ability.
Data regarding Pearson product—moment correlation coefficients between
scores on the Michigan §tgtglUniversityiReading Eggt’and first, second,
and third term grades in Communication Skills and Natural Science courses
and term-end grade point averages for the three terms are presented in
Table 4.11

The correlation coefficients exhibited in Table # range from .3“
to .65. It is thus apparent that reading ability, as measured by the
instrument under consideration, is significantly related to academic
achievement of freshmen during the first year in college.

The reliability of the m Reading Leg has been estimated
by the Office of Evaluation Services of Michigan State University on
various occasions and has been found to be approximately .80. Using
the test scores of the restricted population a reliability coefficient
of .79 was obtained for the Michigan.Stgtg'University Reading Tg§t_by

application of Kuder-Richardson formula 20.

 

11Lehmann and Ikenberry, 22, gi§., p. 6#.

54

TABLE 4

PEARSON PRODUCTéMOMENT CORRELATION COEFFICIENTS OF MICHIGAN STATE
UNIVERSITY READING TEST WITH GRADES IN COMMUNICATION SKILLS
AND NATURAL SCIENCE, AND‘WITH TERM-END GRADE POINT
AVERAGES FOR MALES AND FEMALES

 

 

 

Males Females

Communication Skills

FirSt Term .65 060

Second.Term .58 .49

Third Term .56 .55
Natural Science

FirSt Term .60 .52

Second Term .50 .44

Third Term . .35 .48
Term.Trade Averages

First Term 05? 052

Second Term .48 .34

Third Term .36 .46

 

The Inventory of Beliefs. Form_;

Thg_Invento§z‘g§|Beliefs‘was developed by the Inter-College
Committee on.Attitudes, Values, and Personal Adjustment of the Coopera-
tive Study of Evaluation in General Education.12 The present form con-
sists of 120 pseudo-rational, cliche-like statements to which students
are asked to respond by means of a four-element scale: strongly agree,
agree, disagree, and strongly disagree.

Individuals who reject the majority of the statements contained

in the Inventogy 2§,Beliefs (or high scorers) are thought to be more

 

12Paul L. Dressel and Lewis B. Mayhew, General Education: Egplora-

tions Ig_Evaluation,'Washington, D.C.: American Council on Education,

1954.

55

independent, adaptive, and non-stereotypic in belief systems. A low

score is characterized as being indicative of individuals who are de-
fensive, conforming, resistant, or who are stereotyped in beliefs at

an immature level.13

It is more difficult to provide validity data for measures of
attitudes and values than for measures of academic aptitude. The in-
ventory manual reported evidence of face validity and construct validity.
Indications of concurrent validity are presented in Table 5.1“

It is apparent from Table 5 that Thg.Invento§x,g§,Beliefs is re-
lated to the two measures of academic aptitude, but that the relation-
ship is of a much smaller magnitude than the relationship between 1133
Inventogy‘g§,Beliefs and.Rokeach's Doggatism Sgglg, a measure of an
attitude or personality characteristic similar to that of stereotypy.
The magnitude of the relationship between stereotypy and dogmatism is
similar to the relationship between the College anlification Egg; and
the Michigan Stgtg‘Universitz'Reading Eggt. Thus, we can conclude that
although characteristics measured by the Inventogy 9;.Beliefs are not
independent of the phenomenon measured by the cognitive instruments,
the Inventogz appears to be measuring characteristics more closely re-

lated to factors measured by the dogmatism scale.

 

13Cooperative Study of Evaluation in General Education, Paul L.
Dressel, Director. Instructor's Manual £9}; thg Inventoyx 2f Beliefs,
The American Council on.Education, Committee on.Measurement and Evalua-
tion, 1953 (Mimeographed).

11+th and Ikenberry, 9E. 9232., p. 560

56

TABLE 5

AN INTERCORRELATION'MATRIX OF THE INVENTORY OF BELIEFS,
ROKEACH'S DOGMATISM SCALE, COLLEGE QUALIFICATION TEST,
[AND MICHIGAN STATE UNIVERSITY READING TEST
FOR MALE AND FEMALE FRESHMAN STUDENTS

 

 

Rokeach's College Michigan State
Dogmatism Qualification University
Scale Test Reading Test

 

 

 

Male Female Male Female Male Female

 

Inventory of a
Beliefs -063 -061 .33 .28 033 .29

Rokeach's
Dogmatism
Scale -019 '016 -.22 -020

College
Qualification
Test .70 .71

 

8The negative correlations between Inventogy gfIBeliefs and
Rokeach's Doggatism Scale are resultant from the scoring
procedure. The "concepts“ of stereotypy and dogmatism are
positively related.
Reliability estimates on the Inventory 2;,Beliefs, as reported
in the manual, range from coefficients of .69 to .95 with a median co-
efficient of .86. The Kuder-Richardson formula 20 was applied to the
test score data of the restricted population of the study. A reliability

coefficient of .84 was obtained.
Rokeach's Dogmatism Scale, Formgg

Milton J. Rokeach developed the Doggatism.Scale as a measure of
general authoritarianism, general intolerance, and Openness of belief

systems. Form.E of the Dogmatism Scale consists of 40 dogmatic statements

57

with which the subject is asked to indicate agreement or disagreement
on a six point scale ranging from ”agree very much" to "disagree very
much."

Rokeach has defined dogmatism in the following manner:

(Dogmatism) represents a total ideological defense against

threat and at the same time a cognitive framework for

satisfying one's need to know and comprehend the world one

lives in. In other words, dogmatic thinking and believing

makes it possible to ward off threatening aspects of reality

and at the same time give? one the satisfaction of feeling

that one understands it.

The data presented previously in Table 5 attest to the relation-
ship between the Dogmatism Scale and.Thg_Inventogy‘g£_Beliefs. The fact
that these two scales were constructed independently to measure theoreti-
cally related phenomena and are rather highly correlated tends to support
the claim for validity of both measures.

Plant, Minium, and.Myers reported oddpeven reliability coefficients
for the Doggatism Scale of .84 for males and .85 for females.16 An

internal consistency reliability of .76 was obtained, using the scores

of the restricted population of this study.

Differential Values Inventogy

The Differential Values Inventogy was developed by Richard Prince

at the University of Chicago.17 Several social theorists have studied

 

15Milton J. Rokeach, "Political and Religious Dogmatism: An Al-
ternative to Authoritarian Personality," Psychological Monographs,
70: No. 425, 1956, p. 5.

16Walter T. Plant, EdwardNW. Minium, and Celestine Myers, "An
Analysis of the Rokeach Dogmatism Scale Used/With.A Sample of American
College Students," (An unpublished paper read at the Annual Meeting of
tge'westgrn Psychological Association, San Diego, California, April 16-
1 r 1959 .

17Richard Prince, A.Study.gf the Relationship Betwegn Individual

 

 

58

the value patterns and character structure of the American society.
Prince chose to develop his instrument to measure the ”traditional"

and ”emergent" value categories outlined by Spindler.18 The items
contained in the Differential Kalggg’lnventogy‘were written for four
”traditional” value categories which included puritan morality, indi-
vidualism, work-success ethic, and future-time orientation and for the
four "emergent” value categories of sociability, conformity, relativistic
moral attitudes, and present-time orientation. These were the categories
originally set forth by Spindler, except that Spindler included "achieve-
ment orientation” in the group of traditional values and ”consideration
for others” in the emergent value area.

A person with high traditional values was described as placing
high value on personal respectability, thrift, self-denial, respect for
elders; as valuing hard.work as good in itself and necessary for success;
as placing his individual desires and ideas equal to or above the de-
sires and ideas of the group; and as being oriented toward the future
to the extent that present needs should be sacrificed for future satis-
faction and reward.19

The emergent value system on the other hand places a great deal

of importance on getting along with others and achieving group harmony.

 

Values gag_Administrative Effectiveness ig_thg_$chool Situation, (Un-
published Dissertation, University of Chicago, 1957;.

18George Spindler, "Education in a Transforming American Culture,"
Harvard Educational Review, 25:156-163, Summer, 1953.

19Prinee, 92. git" p. 42.

59

Absolutes are questionable and morality is group-determined rather than
predetermined.by any particular source. Everything one does should be
done in consideration of others and in regard for feelings and reactions
of others. Emphasis is on the present.20
Prince administered the inventory to several groups which were
assumed to be different in relationship to the value patterns in ques-
tion. An extensive item analysis was conducted to identify the items
which were most appropriate to the purpose of the instrument. In the
final form.of the Differential Xglgg§_lnventogy there are 64 items.
In each item an emergent value alternative and a traditional value al-
ternative are pitted against each other. The subject is forced.to make
a choice between one of the two alternatives. .
Prince found that high school students attending parochial schools
‘were more traditional in values than students attending public high
schools: teachers in parochial high schools more traditional in values
than teachers in public high schools; teachers more traditional in
values than students in both public and parochial high schools: adminis-
trators more traditional than teachers: and older teachers more tradi-
tional than younger teachers.21
Lehmann and Ikenberry found that students aspiring toward.more
than four years of college education were more traditional in values

than students aspiring to four years or less of college. Rural college

 

2°1bid., p. 42.
21Ibid., pp. 87-91.

60

students were found to be more traditional in values than urban students.
Students with parents of low educational level and occupational status
tended to be more traditional in values.”
These value differences identified by the Differential Egg
Inventog are in general agreement with current knowledge and theory
concerning differences in values among various social and cultural
groups. Hence, it would appear that the Differential @193 Invento
is measuring characteristics similar to those described by Prince in
the purpose of the Inventog.
An estimate of reliability, using the restricted population of
this study and the Kuder-Richardson formula 20, was computed with a re-
sulting coefficient of .75.
Michigan State University freshmen were re-tested on the Qi_f_f_e_y-
m 159322. Inventogy at the end of one year. Although some change
in values could logically be expected over a period of nine months of
college, correlation coefficient of the fall and spring scores of .61
for males and .60 for females were found.23 Coefficients of this mag-
nitude indicate a reasonable stability in the value scores over the
year.
Although the evidence for validity and reliability of the Digg-
ggtial Iii-41.2% Inventogy is not as extensive as some other measures used
in this study, the inventory appears to have a sound theoretical base
with more evidences of validity and reliability than most value instruments.

 

2"aLehmann and Ikenberry, 22. git” pp. 32-36.
23Ibid., p. 74.

61

The Social Status Index

Previous research in the area of college student withdrawals has
provided evidence of a relationship between the social backgrounds of
adolescents and college attendance. For this reason it was necessary
to obtain some indication of the social backgrounds of the members of
the sample. The social prestige of the father's occupation, and the
level of education attained by the father and mother were considered
to be indicative of the general social status of the family.

Accordingly, data indicating the level of schooling completed
by the father and mother and a description of the father's occupation

were obtained from each of the students in the restricted pOpulation.

The educational level of the father and mother were each classi-

fied according to the following form:

1. Attended grade school (grades 1 through 8) but did not finish.

2. Completed grade school through grade 8.

3. Attended high school (grades 9 through 12) but did not
graduate.

4. Graduated from high school.

5. Attended college but did not graduate.

6. Graduated from college.

7. Attended graduate school or professional school but did
not attain a graduate or professional degree.

8. Attained graduate or professional degree.

The educational levels of the father and mother were quantified
simply by assigning point values of one through eight corresponding to
each of the numbered alternatives listed above.

Data were collected regarding the father's occupation by asking

students to "describe in a line or two....' what their father did for

a living. The occupational descriptions were then classified according

62

to the North-Matt classification system as set forth by Bendix and
Lipset.2#

Ratings of occupations ranged.from a point value of 96 given
to U.S. Supreme Court Justice to 33 points given to a shoe shiner.25
The usefulness of the North-Hatt occupational ratings was considerably
increased.by additional research at Ohio State and.by the work of
A. o. Haller of Michigan State University.26 Both the staff at Ohio
State university and Dr. Heller have provided interpolations to the
original scale which have resulted in ratings for over 500 different
occupations.

The first step in the quantification of the social prestige of
the father's occupation was to examine the NortheHatt listings to de-
termine if the occupation had been rated in the original work. If the

occupation was not contained in the North-Matt work, the ratings provided

 

2“Reinhard Bendix and.Seymour Martin Lipset, Class, Status gag
Power, Glencoe, Illinois: The Free Press, 1953, pp. 110-115.

25The National Opinion Research Center conducted a survey which
‘was based on the plans developed by Paul K. Matt and C. C. North. Ran-
domly selected subjects rated 90 occupations as (1) excellent standing;
(2) good standing; (3) average standing; (4) somewhat below average
standing; (5) poor standing; (6) I don't know where to place that one.
‘When ”don't know“ answers were excluded and other reaponses were
weighted, the scoring theoretically allowed a maximum of 100 points
for any job receiving only "excellent” ratings, and.a minimum of 20
points for work that was unanimously rated as "poor.“

26Both the interpolations derived at Ohio State University and
the interpolations derived by Dr. Haller were supplied.by Dr. A. O.
Haller, Department of Sociology and AnthrOpology, Michigan State
University; The North-Hatt interpolations are recorded in an unpublished,
nileographed paper by Russel Dynes entitled “The Horth.Hatt Scale.”

63

by The Ohio State University were used. In the event that the occupa-
tion was not rated in the original North-Matt work, nor in the Ohio
State University interpolations, Dr. A. a. Haller's interpolations were
used as a third source.

If the father's occupation was not given by the student, the
following preference order was utilized: (1) the step-father's occupa-
tion, if any: (2) the mother's occupation, if any; (3) the mean rating
of all occupations as established in the North—Hatt study. In over 95
per cent of the cases, the father's occupation was listed.

After the data regarding the educational level of parents and
father's occupation had.been quantified, the three variables were then
combined into a single social status index score. Factor analysis was
used to determine the weights for combining the variables into the index
score which would be most representative of social status or social
background. The method for the construction of the index was based on
the preperty of the first factor. ”This factor represents the dimension
along which the items can best discriminate."27 It was assumed that the
one phenomenon these items could best measure, in combination, was the
social status position of the family;

The intercorrelations among the three biographical variables
used in the index are presented in Table 6. The Pearson product-moment
correlation coefficients were computed using the scores of every member

in the study sample. The correlation matrix was factor analyzed and

 

27Margaret J. Hagood and Daniel 0. Price, Statistics :31; Sociolo-
ists, New York: Henry Holt and Company, 1952, p. .

64

TABLE 6

INTERCORRELATIONS OF BIOGRAPHICAL VARIABLES RELATED
TO THE SOCIO-ECONOMIC POSITION OF'THE FAMILY

 

 

Biographical Father's Mether's
Variables Educational Educational
Level Level

 

Mether's Educational
Level .228

Father's Occupational
Prestige Rating .350 .150

 

the first factor was identified by a set of factor loadings. Correla-
tion coefficients were obtained between the first factor and each of
the biographical variables.

The correlation coefficients of each of biographical variables
with the first factor are presented in Table 7. To form an index it
was necessary to weight the variables in prOportion to their correla-
tions with the first factor.28

The raw scores were transformed into standard scores. Each
standard score was then multiplied by the correlation coefficient of
the respective variable with the first factor. The following formula

29

was used:

 

28S. S.‘Wilks, "Weighting Systems for Linear Functions of Cor-

related Variables'When There Is No Independent Variable,“ Psychometrika,
3: 24-143, March, 1938.

29Hagood and Price, gpp‘gi§., p. 528.

TABLE 7

CORRELATION COEFFICIENTS OF THE BIOGRAPHICAL
VARIABLES WITH THE FIRST FACTOR

 

 

Biographical Variables

Correlation Coefficients

 

Father's educational level
Mother's educational level

Father's occupational

prestige rating

 

Index Score

where: X1, N1, 51;

X2: 362: 523

X3,

 

are the raw score, mean,
deviation, reapectively,
educational level.

are the raw score, mean,
deviation, respectively,
educational level.

are the raw score, mean,
deviation, respectively,
occupational prestige rat

.786
.584
.731
x .3? X .35 x .55
#8603533) + .584( g 2) + Jam-£33)

and standard
of father's

and standard
of mother's

and standard
of father's
ing 0

The means and standard deviations for the three biographical

variables are presented in Table 8.

Using the known values presented in Table 8, it was possible to

simplify the formula to the following:

Index Score = (.u35x1 + .3151:2 + .080X3) - 8.474

where:

X1: raw score of father's educational level.

raw score of mother's educational level.

X3: raw score of father's occupational prestige rating.

66

TABLES

MEANS AND STANDARD DEVIATIONS OF FATHER'S EDUCATION,
MOTHER'S EDUCATION, AND FATHER'S OCCUPATION
FOR THE SAMPLE OF THE STUDY

 

 

 

 

 

 

Biographical Variable Mean Standard Deviation
Educational level of fathers 4.277 1.808
Educational level of mothers 3.381 1.854

Occupational prestige rating
of fathers 69.363 9.116

 

Using the International Business Machine's Electronic Punch
machine the first factor index was calculated for all members of the
sample. The above formula yields a mean of zero with both plus and
minus values in the score range. A constant of six was added to each
index score which caused all scores to be positive. Decimals were re-
moved by multiplying each score by 10, thus providing index scores
which ranged from 22 to 106.

Collection of Data

During the registration periods, September 21-26, 1958, the fol-
lowing instruments were administered to entering freshmen at Michigan
State University: 2113 Inventogy _<_>_i_‘. Beliefs, Egg; 3;, Th_e_ '_I‘_e_§_t_ p_f_ Critical
Thinking, M Q, The; Differential Em Inventozy, Rokeach's Doggatism
§_<_:_a_l_e_, The. Michigan M University Reading Leg, The. College anlifi-
32.33911 Te_s_t, ing _a_ Biographical Questionnaire. Two subsequent testing
sessions were arranged for students who did not attend the first session.

Out of 2,973 entering freshmen who met the requirements of the restricted

6?

pOpulation, complete and usable test data were obtained for 2,746 stu-
dents, or 92.4 per cent of the total restricted pepulation.

The names and student numbers of students who withdrew from
college during any of the three terms of the freshman year were ob-
tained from the Office of the Registrar. .An additional check was made
during the winter and Spring registration periods to identify students
who did not register for study.

First term grade point averages were obtained for all students
in the restricted population at the end of Fall term. The scores of
students on the various test instruments, the status of students with
respect to college persistence, and the Fall term grade point average
of students were recorded on International Business Machine Cards, using
a separate card for each student. The student number, and a code denot-

ing sex were also included on the card.
The Statistical Model and Computation Procedures

After a review of several possible statistical models, multiple
discriminant analysis was selected as the most appropriate technique.
Multiple discriminant analysis is a statistical method of combining test
scores or other available data so as to maximize the difference among
groups and minimize the differences within each group. Through the
separation of individuals who are known to belong to mutually exclusive
groups, it is possible to determine the combinations of variables which
will maximally discriminate among the different groups. It is also
possible to observe the magnitude of the group differences and to classify

future individuals into one of these groups on the basis of similar data.

68

In this study, individuals had been classified according to collegiate
persistence, first term grade point average, and sex.

A series of observations or measurements for each member of ten
defined and mutually exclusive groups was collected.29 The problem re-
quired a statistical methodology which would maximally discriminate
among the ten groups of the basis of the information available. It was
desirable to use a statistical tool which would indicate the intensity
and direction of the difference. Because many of the variables used in
the study were interrelated, it was also advisable to use a technique
designed to identify basic, independent factors which accounted for

possible group differences.

Description of Multiple Discriminant Analysis

 

The computational procedures followed were based on a method
provided by Bryan in his doctoral dissertation.30

An illustration of computation procedures, including worksheets,
is available in an Air Force research report by Bryan, Rulon, and

Tiedeman.31 A discussion of the development and perfection of the

 

2
9See Table 1.

3"Joseph G. Bryan, _A_ Method 395 gig Exact Determination 2;; pp
Characteristic Euation ppd Latent Vectors pg _a_ Matrix with Applications
t2 Eh; Discriminant Function ill}; More Than _Typ Groups, Cambridge: Harvard
University Graduate School of Education (Unpublished.Doctoral Disserta-

tion), 1950.

31David V. Tiedeman, Joseph G. Bryan, and Phillip J. Rulon, Tpp,
Utility p}; 3.113 Airman Classification Batteg F9}; Assignment 9}; Airmen
pp_Eight pip,Force S ecialties, Cambridge, Mass.: Educational Research
Corporation, June, 1951.

69

multiple discriminant analysis technique has been published by Tatsuoka
and Tiedeman.32
The linear combinations of variables which maximize the differ-
ences between groups and minimize the differences within groups are
derived from the solution of the determinantal equation,
'A - )w Iv = o
where: A = The Among Matrix
W = The Within Matrix
)\= The Latent Roots of the System

V = The Latent Vectors, or Discriminant
Coefficients

The first step in the computational procedure was to compute the
intercorrelation matrix, means, standard deviations, variances and co-
variances for the seven variables in each of the ten groups and for the
total group. The Michigan State Integral Computer, MIS__T_T_C_:_, was used to
compute these values.33

The means computed for each group and for the total group were
used to compute the among groups, or A matrix. The element in the ith
row and the jth column of the A matrix was computed according to the

following formula.

1'] g=1 Ng E Ng

 

32Maurice Tatsuoka and David V. Tiedeman, "Discriminant Analysis,“
Review pf Educational Research, 21+:402-420, December, 1951+.

33Computer Laboratory, Michigan State University, "K5 .. M, Cor-
relation, Means, Standard Deviation, Variance, Card Input,” MISTIC
Library Index, East Lansing, April, 1959 (Mimeographed .

H

70

where: aij is the element in the ith row and the jth column
of matrix A.

p is the subscript denoting individuals.

g is the subscript denoting groups.

Xi is an individual's score on the ith variable.
Xj is an individual's score on the jth variable.
N8 is the number of individuals in each group.

The variance and covariance matrices were used to compute the
within groups, or‘W matrix. The formula for the computation of the‘w
matrix is as follows:

0
wij = t 1%?!in - ZXiZX.

g=1
NE

where: wij is the element in the ith row and the jth column
of matrix W.

g is the subscript denoting groups.
is the subscript denoting individuals.

P
X. is an individual's score on the ith variable.
X

i
j is an individual's score on the jth variable.
Hg is the number of individuals in each group.

The element in the ith row and the jth column of the variance-
covariance matrix was multiplied by the number of students in the group
and the corre5ponding products from the ith row and.the jth column for
each of the ten groups were added. The resultant sum was equal to the

element in the ith row and jth column of the'W or within matrix.

71

After the computation of the among and within matrices, the
determinantal equation, ’A - )W'V = 0 was solved through the use of
the Michigan State Integral Computer.3z+ Several linear combinations
of variables are possible in the solution of the determinantal equation..
The first linear combination of the determinantal equation maximizes the
discriminant criterion; the second linear combination maximizes the ratio
of the residual dispersion among groups to the residual dispersion with-
in groups after the effect of the first linear combination has been re-
moved; the third linear combination maximizes the ratio of the among
groups dispersion to the within groups dispersion after the effects of
the first two have been removed. Subsequent linear combinations con-
tinue to maximize the ratio of the among groups dispersion to the within
groups dispersion after the effects of the preceding linear combinations

have been removed.
Assumptions of the Statistical Model

The assumption which is made in the use of multiple discriminant
analysis is that the test scores of the pepulations under study are
multivariate normal with equal variance and covariance matrices. A
review of the literature was made to determine the availability of
methods to test the assumption of multivariate normality and equality

of variance and covariance matrices. The review of the literature

 

3“"M5 - 139, Solution of Determinantal Equation A .. Ma = o,"
MISTIC Librapy Index, East Lansing, Michigan: Computer Laboratory,
Michigan State University, April, 1959.

72

revealed no complete test of this assumption. In the review of previous
research which had used the discriminant analysis statistical technique,
no studies were found which reported a test of the assumption.

David V. Tiedeman, an authority on the use of discriminant
analysis in the fields of education and psychology, was presented with
the problem of (1) the availability of a method to test the assumption;
and (2) the advisability of testing the assumption if a method were
available. Tiedeman stated that,

I don't know of a simple, direct test of the multivariate
normal assumption. There are no tables of which I am.aware
which specify the expected frequencies under the assumed
multivariate normality. Ybu might be able to compute these
under existing electronic equipment by Mente Carlo methods
but it would take time.... we usually test the normality of
the discriminant scores rather than that of the original
scores. In addition, we do it for the discriminants
separately rather than simultaneously. It's not just that
it's easier; it's also because I feel that the expected
distributions are not affected too markedly by considerable
departures from.the assumption of multivariate normaagty
necessary for the derivation of those distributions.

The decision was therefore made not to make a formal test of the
assumption of multivariate normality and equality of variance-covariance
matrices of the populations. If the data of this study were to be used
to classify future freshmen students into the existing groups defined in
this study, a test of the normality of the discriminant scores should be

made.

 

35A letter received from.Dr. David.Tiedeman, Professor of Educa-
tion, Harvard University.

73

Summary

The population of the present study consisted of the Fall, 1958,
entering freshman class of Michigan State University with the exclu-
sion of foreign students, part time students, transfer students, and
students who had incomplete test records.

The academic status of all members of the freshman class, includ-
ing number of credits carried and course grades during the first term
of study were collected.

All students who withdrew from college between September and
June of the first academic year were classified into six separate groups
on the basis of sex and first term grade point average. A random sam.
ple of 250 freshmen students who remained enrolled through the end of
the freshman year was drawn and subsequently classified on the basis of
sex and first term grade point average. Thus, with the samples of en-
rollees and withdrawals, classified on the basis of collegiate persist-
ence, first term grade point average, and sex, a total of ten groups
was available for study.

The instruments used in the study included.Tpp,College Qpalifi-
9.9.29.2 Leif” _tpg Le_s_t_ p_f_ Critical Thinking, Egg}; 9, up Michigan _S_1._a_t_e_
University Reading _T_e§t_, 1113 Differential _V_a_l_1_1§_§_ Invento , T112 Irm-
m pf Beliefs, Rokeach's Dogpatism §ppa_l_g, and g Biographical pppg
§Qgg§, A social status index score was computed through the combination
of the biographical variables of father's educational level,.mother's
educational level, and father's occupational prestige rating into a

single index score. All instruments were assumed, on the basis of prior

7#

research, to have sufficient validity and reliability to be included as
criterion measures in this study.

Multiple discriminant analysis was used to analyze the data.
The solution of a determinantal equation yields the linear combinations
of variables which maximize the differences among groups and minimize
the differences within groups. The first linear combination maximizes
the ratio of the dispersion among groups to the dispersion within groups.
Subsequent linear combinations continue to maximize the ratio of the
among groups to within groups dispersion after the effects of preceding
linear combinations have been removed.

The results of the analysis are presented in Chapter Four.

CHAPTER IV

THE ANALXSIS OF DATA

The data were initially analyzed to determine group status on
each of the variables and interrelationships among the variables. As
a matter of record, group means on each of the variables used in the
study are presented in Appendix A. The within group intercorrelation
matrix of the seven variables used in the study is presented in Table
9. A clear understanding of the basic interrelationships among the in-
struments used in the study is helpful in the interpretation of the
multivariate linear combinations of variables in the discriminant
functions.

TABLE 9

PEARSON PRODUCT-MOMENT WITHIN GROUP CORRELATION COEFFICIENTS
AMONG THE VARIABLES USED IN THE STUDY

 

VARIABLES 1 2 3 a 5 6 7

 

1. Inventory of Beliefs
2. Critical Thinking .2h
3. Differential Values -.03 .03

h. Rokeach's Dogmatism -.58 -.17 .12

5. Reading .22 .53 .03 -.18
6. Coll. Qual. Test .22 .51 .03 -.13 .61
7. Social Status .08 .02 ' -.16 -.05 .05 .11

 

It is evident from Table 9 that the cognitive measures are highly

interrelated. The highest correlation coefficient was .61 between the

76

College Qualification Tp§p_§pg_ppg.Michigan SippglUniversity‘Reading
Egg}, The lowest coefficient of correlation was .51 between the igpp
Q Critical Thinking and the College @alification Iggy.

A low order but statistically significant relationship was found
between the two attitudinal measures and the three cognitive instru-
ments.1 The Inventopy p£.Beliefs, for example, correlated .22, .2#,
and .22 with the Michigan Sgt; University Reading kg, 1.32 T_e_§_t p_f_
Critical Thinkin , and the College Qpalification Tppp, respectively.
Similarly, Rokeach's Dogpatism _S_9_g_lp correlated -.18, -.17, and -.13,
respectively, with the same three cognitive instruments. In contrast,
however, a high correlation coefficient of -.58 was found between the
Inventopy p£.Beliefs and Rokeach's Dogmatism §pplg,

The Differential Valpgg Inventopy was not found to be signifi-
cantly related to the three cognitive measures. The correlation co-
efficient between the Differential Vplppg’lnventopy and the Inventopy
p§,Beliefs did not reach the magnitude required for statistical signifi-
cance. A low but significant correlation was found, however, between
the value instrument and the social status index. The correlation co-
efficient of -.16 would indicate a slight tendency for individuals with
high traditional value scores to have a lower score on the social status

index.

 

nu.

1Based on the appropriate F test, a within groups correlation
coefficient of .12 is significantly different from zero at the .01 level
of confidence. A coefficient of .09 is significant at the .05 level of
confidence.

77
Results of the Multiple Discriminant Analysis

The Test of the Hypothesis

Following the procedures outlined in Chapter Three, the within
groups and among groups matrices were computed. These matrices are
presented in Appendix B of the study.

The solution of the determinantal equation 'A - AWIV = 0, where
A is the among matrix,‘W'is the within matrix, v represents the discrim-
inant coefficients, and 7t represents the latent root of the system, was
necessary for the test of the hypothesis of the study. This hypothesis,
stated in null form, was as follows:

There is no difference in intellectual ability, social back-
ground, attitudes, and values among groups of students
classified by collegiate persistence, first term grade point
average, and sex.

A test of the statistical significance of the latent roots, or
discriminant functions, has been presented by Rao to test multivariate
discrimination among several groups.2 The formula used to test the sta-

tistical significance of the discriminant functions was as follows.

)éEN - i- (p + k)]loge (1 +)\)
where: the total sample of 553 individuals.
the total number of variates, or 7.
the total number of groups, or 10.
the discriminant function, or latent

root of the system.

gyrrmstz

Values of chi square computed by the above formula can be referred

to the tabled distribution of chi square values with the appropriate

 

20. Radhakrishna Rao, Advanced Statistical Methods in Biometric
Research, New York: John‘Wiley and Sons, Inc., 1952, pp. 372-73.

78

degrees of freedom. The latent root, the chi square value, the degrees
of freedom, and the significance level obtained for each of the discrimi-
nant functions are presented in Table 10.
TABLE 10
LATENT ROOTS , CHI SQUARE VALUES , DEGREES 0F FREEDOM

AND STATISTICAL SIGNIFICANCE LEVELS FOR
EACH OF THE SIX DISCRIMINANT FUNCTIONS

 

_
——_“

 

 

 

 

maxim a x2 5min”
V1 .5120 225.113 ‘ 15 41.001
V2 .2022 100.270 13 <:.001
v3 .0t20 22.395 11 41.05
V4 .0196 10.569 9 .40
V5 .01h8 8.113 7 .50
v 6 . 0054 2. 921» 5 .95

 

Two discriminant functions were significant beyond the .001 level
of confidence. The third function was significant at the .05 level of
confidence. The magnitude of the three remaining latent roots, or dis-
criminant functions, was not sufficient to achieve statistical signifi-
cance. The latter three functions were therefore concluded to represent
chance variation.

If the sum of the latent roots were to be considered an estimate
Of the total variance or dispersion among groups as defined by the in-
struments used in the study, the first discriminant function would ac-
count for approximately 6h.3 per cent of the total dispersion among

groups; the second discriminant function would account for 25.4 per cent

79

of the total dispersion among groups; the third discriminant function
would account for 5.3 per cent of the total dispersion. The remaining
three functions, combined, would account for less than 5 per cent of

the total dispersion among groups.

Interpretation of the Significant

Discriminant Functions

It is possible to interpret the discriminant functions by an
examination of the conventionalized coefficients or by an examination
of the conventionalized coefficients weighted by the standard deviation
of the corresponding variate. Tiedeman and Bryan make the following
comment on the interpretation of discriminant functions:

It can be shown that the individual values of the discriminant
function are independent of the units of measurement, and
origin of coordinates of the initial variates, since the co-
efficients automatically adjust themselves (linearly) to the
scales employed. 0n the other hand, the interpretation of
separate coefficients does depend on the units of the initial
variates.3

Tiedeman and Bryan concluded that in cases in which the units
of measurement are sufficiently comparable (i.e., instrwments have
similar or identical range, means, and variance) that interpretation

of functions may be made directly from the conventionalized coefficients.4

 

3David V. Tiedeman and Joseph.G. Bryan, "Predictions of College
Field of Concentration,“ Harvard Educational Review, 2h:122-39, Spring,

1951+, p. 132.

“The term "conventionalized coefficient" indicates that dis-
criminant coefficients were divided by the value of the largest co-
efficient which yielded a value of 1.00 for the largest coefficient,
and lesser values for the remaining coefficients.

80

If the variates, however, doxot have similar units of measurement,
and if the standard deviation of some variates is two or three times
that of other variates, for purposes of interpretation, the discriminant
coefficients should be weighted by the standard deviation of the in-
strument. Such a procedure would adjust the coefficients in terms of
the initial units of measurement.

The variables used in this study did not have similar or com-
parable units of measurement. For this reason the weighting procedure
was followed for purposes of interpretation. The standard deviation of
each of the instruments is presented in Appendix C.

Because only the first three discriminant functions were statis-
tically significant, the remaining three discriminant functions are not
included in subsequent discussions of the findings. The conventionalized
discriminant coefficients for all six functions are included in Appendix

D of the study.

 

Inteppretation of the First Discriminant Function

The weighted conventionalized discriminant coefficients of the
first and.most powerful discriminant function are presented in Table 11.
Examination of the data in Table 11 revealed that the first function
may be considered intellective or cognitive in nature. The College
Qualification 22.5.2: the Michigan §_t_a_l_t_e_ University Readipg 23.1., and the
Tppp.p£_Critical Thinking were all weighted heavily and positively.
The dominance of the three cognitive measures supports the generaliza-

tion that the first function was primarily an intellective one.

81

TABLE 11

WEIGHTED CONVENT IONALIZED DIS CRDIINAIH COEFFICIENTS
OF THE FIRST DISCRDIINANT FUNCTION

 

 

 

Variable Weighted Coefficient
Inventopy p_f_'_ Beliefs - .319
Legit; 2;: Critical ThinkingI 6.290
Differential Lalpgg Inventogy 1.590
Rokeach's Domatism M - .959
Michigan Spats University Reading T_e_§_t 6.858
College C‘Malification Legit. 11.096
Social Status Index 5.011

 

The social status index was weighted positively in the first
discriminant function, along with the cognitive variates. Apparently
some element measured by the social status index is positively related
to the dimension differentiated by the cognitive instruments. This
finding would agree with the previous research findings concerning the
relationship between social background and measures of academic ability.

The contribution of the Differential _V_a_l_u_e_§ Inventopy in the
first function was minor when compared to the weightings of the cogni-
tive instruments and the social status index. The values instrument
was, however, weighted positively in the first function, in the expected
direction of traditional values of achievement, morality, individualism,
and orientation toward the future.

The attitudes of stereotypy as measured by the Inventopy pf

Beliefs and dogmatism as measured by Rokeagh's Doggtism Scale did not

82

appreciably influence the nature of the first function. This conclu-
sion is based first upon the slight weightings received.by the two atti-
tldinal measures in the first function. Secondly, because the two tests
are negatively correlated it was enigmatic that both instruments were
weighted in the same direction in the first function. It is likely that
the minimal influence of one instrument would be at least partially re-
moved by the other. Of course, the discriminant coefficients are so
small that chance alone could account for the seemingly contradictory
weightings.

Thus, to summarize the discussion of the first discriminant func-
tion, the complex of discriminant coefficients indicated that the function

was primarily an intellective function.
Interpretation of the Second Discriminant Function

The weighted conventionalized discriminant coefficients of the
second discriminant function are presented in Table 12. The College
Qualification Tppp,received the heaviest weighting in the second func-
tion and was weighted negatively. The Tp§p_p£|Critical Thinking received
a small and negligible weighting. The Michigan.§pppg,Universitleeading
Zpgpfwas heavily weighted in a positive direction.

Examination of the sub-tests of the College Qualification T_ea_sp
revealed.that the total score was strongly influenced.by numerical and
science items. The numerical sub-test, of course, contributed to the
numerical factor of the total score. The information sub-test also con-
tributed to the numerical-science influence because half of the items

‘were taken from the natural science subject matter areas.

83

TABLE 12

WEIGHTED CONVENTIONALIZED DISCRD-IINANT COEFFICIENTS
OF THE SECOND DISCRIMINANT FUNCTION

 

 

 

 

 

 

 

Variable Weighted Coefficient
Inventopy pi: Beliefs 1.236
2231': pf Critical Thinking ”605
Differential M Inventopy -1.’+08
Rokeach's Dogmatism §_c;a_._l_e_ - .119
Michigan _S_t_a_t_._e_ University Reading IEEE 6.858
College Qualification 1gp -9.539
Social Status Index 5.495

 

The positive weighting received by the highly verbal Michigan
M University Reading 23st; in the second function can be assumed to
have neutralized the negative weight received by verbal elements in the
College Qualification 3.332- It is probable, however, that a negative
weighting on some of the numerical-scientific elements of the College
Qpalification 2.2% remained. It may thus be concluded that the inter-
action of the cognitive instruments resulted in a slight negative weight
on a numerical-scientific ability factor.

The meaning of the second function, however, does not become
clear until the non-cognitive measures are examined. Examination of
the weighted discriminant coefficients of the non-cognitive measures
suggests the influence of various social-cultural elements related to

socio-economic background and to sex. The social status index, for

84

example, received a heavy positive weighting. The traditional values
of the Differential Vglpgg Inventopy were weighted negatively in the
second function, favoring the emergent values of sociability, group
conformity, and moral relativism, previously associated.with students
from high social strata and with females.

The slight contributions of the two attitudinal measures are also
in the direction of attitudes previously associated with individuals
from the middle class culture and with females.

The negative weighting of the College Qualification Tgpp, and
the positive weighting of the Michigan Spgpg’University Reading Eggp,
when considered in relation to the non-cognitive variables, appear to
result in the same social-cultural influence. The negative influence
of numerical ability is in agreement with previous investigations con-
cerning sex and social-cultural differences on cognitive or ability
measures. College females, for example, have been shown to receive
higher scores than college males on tests of verbal ability. Similarly,
males are known to typically score higher than females on tests of nu-
merical ability.5 Previous research also indicates that there are
social-cultural differences on measures of verbal ability.6

The complex of discriminant coefficients in the second discrimi-

nant function would indicate that the function may be interpreted

 

5Dcrothy H. Eichorn and Harold.E. Jones, “Development of Mental
Functions," Review'pf,Educational Research, 22:#21-h38, December, 1952.

6Samuel R. Pinneau and Harold E. Jones, ”DeveIOpment of Mental
Abilities,” Review'pf'Educational Research, 28:392-400, December, 1958,
p- 394.

85

primarily as a social-cultural function, maximizing differences previ-
ously associated with different socio-economic and sex groups. Such a
generalization is supported by the positive weight of the social status
index, the negative weight of the Differential M Invento , and
the negative weight on numerical-scientific abilities resulting from

the interaction of the cognitive instruments.

The Interpretation of the Third Discriminant Function

The weighted conventionalized discriminant coefficients of the
third discriminant function are presented in Table 13. The most promi-
nant variable in the third discriminant function was the positively

TABLE 13

WEIGHTED CONVENTIONALIZEE) DISCRDIINANT COEFFICIENTS
OF THE THIRD DISCRDHNAM FUNCTION

 

 

 

 

Variable Weighted Coefficient
Inventom p_f_ Beliefs 1.87
1‘38; _o_f_ Critical Thinking -2. 54
Differential la_lp_e_s_ Inventog .04
Rokeach's Domatism §pa_l_e_ 2.17
Michigan m University Reading 21511 6.86
College Qualification T3333 -2.69
Social Status Index -3.91

 

weighted Michigan State University Reading Test. The two remaining

cognitive measures, however, received negative weights. The social

86

status index was negatively weighted in the third function and was second
in magnitude only to the reading test. The weightings received by the
Inventopy‘p£,Beliefs, Rokeach's Dogmatism Spplp, and the Differential
Valppp Inventopy were small and were considered negligible. In addi-
tion, the fact that both attitudinal measures were positively weighted,
and highly negatively correlated suggests that each instrument may serve
to counter balance or negate the other in the final discriminant score.

Examination of group differences when the raw score group means
have been weighted by the correSponding discriminant coefficients of
the third function is helpful in the interpretation of the function.

In Table 14, the raw score means of certain variables, weighted by the
correSponding discriminant coefficients, are presented for each group
and for the total group. In the second column of Table 14 the values
for groups resultant from the combination of cognitive instruments are
listed. In the third column the contributions of the social status in-
dex are presented. Contributions of the attitude and value measures are
listed in column four and the mean discriminant scores for groups and
for the total group on the third discriminant function are presented in
the final column of Table 14.

A careful examination of the data contained in Table 14 revealed
that when an above average discriminant value on the cognitive measures
was combined with a below average discriminant value on the social status
index, a high score on the third function resulted. This pattern is
illustrated in groups one, two, five and six. The converse condition,

a below average discriminant value for the cognitive instruments and an

above average discriminant value on the social status index, resulted

87

TABLE 14

DISCRIMINANT FUNCTION VALUES FOR SELECTED COMBINATIONS
OF VARIABLES AND FOR THE FINAL SCORE ON THE THIRD
DISCRDiINANT FUNCTION FOR EACH GROUP AND
FOR THE TOTAL GROUP

 

 

 

 

 

Group Description Cognitive Social Attitudes Mean Score on
Instruments Index and Values Third Function

1. Withdraw, No GPA,

Males 4.13 42.85 22.57 13.85
2. Withdrew, No GPA,

Females 6.03 ~15.44 22.41 13.00
3. Withdrew, Below

2.0, ”3193 2044‘ -13089 22.74 11029
4. Withdrew, Below 2.0,

Females 3.97 -15.39 22.80 11.38
5. Withdraw, Above

200, mes 4054' -15021 22.87 12020
6. Withdrew, Above

2.0, Females 7.17 -15.40 23.41 15.18
7. Enrolled, Below

2.0, Males 3.63 -15.87 22.69 10.45
8. Enrolled, Below

2.0, Females 4.64 -17.03 22.99 10.60
9.. Enrolled, Above

2.0, Males 4.52 -16.20 23.06 11.38
10. Enrolled, Above

2.0, Females 6.35 -17.76 22.49 11.08

Total Group 4.43 -15.55 22.80 11.68

in a low score on the third function. This pattern is illustrated best

by groups seven and eight. When the cognitive and social status dis-
criminant values are approximately proportionate in magnitude, an average

score on the third function resulted.

88

The minimal contribution of the attitudinal and value measures
in the third function is also illustrated in Table 14. Thezange between
the highest and lowest group scores is only one point. It may be con-
cluded that the differences which are apparent on the third function
are resultant from an interaction between the cognitive measures and
the social status index.

Focusing attention on the cognitive measures and the social
status index, the data in Table 14 reveal that the discriminant con-
tributions resultant from.the combination of the cognitive instruments
and also the discriminant contribution of the social status index are
higher for female groups than for male groups. Because group differ-
ences on the cognitive combination are resultant from interaction be-
tween a highly verbal instrument, (reading) with instruments more heavily
loaded with numerical and scientific factors, group differences on the
cognitive combinations likely represent sex differences rather than
differences in general intellectual ability.

It is therefore concluded that the interaction between the cogni-
tive instruments and the social status index results in a removal of the
sex factor from the social status index. If the first function were
considered an intellective function with sex and cultural differences
included, and the second function a social-cultural function including
sex differences, the third function may logically be considered a social

background function, with sex and intellective differences removed.

89

Group Differences on the Significant

Discriminant Functions

For purposes of discussion and analysis, group differences on
the first two discriminant functions will be presented first and group
differences on the third discriminant function will be discussed separ-
ately. The mean discriminant scores for the first and second discrimi-
nant functions are presented in Table 15.

TABLE 15

MEAN DISCRIMINANT SCORES ON THE FIRST AND
SECOND DISCRIMINANT FUNCTIONS

 

 

 

Group Description First Function Second Function
Withdraw, No GPA, Males 104.40 3.42
‘Withdrew, No GPA, Females 123.22 8.86
‘Withdrew, Below 2.0, Males 107.04 3.38
Withdrew, Below 2.0, Females 100.02 9.94
Withdrew, Above 2.0, Males 135.03 3.14
‘Withdrew, Above 2.0, Females 122.03 10.54
Enrolled, Below 2.0, Males 120. 5'3 6. 5a
Enrolled, Below 2.0, Females 110.94 12.89
Enrolled, Above 2.0, Males 137.83 4.08
Enrolled, Above 2.0, Females 130.77 12.22

 

The position of each of the ten groups in the two dimensional
space defined by the first two discriminant functions has been plotted

in Figure 1. It is apparent that the intellective discriminant function,

90

Figure 1

Differences Among Group Means Plotted in the
Two Dimensional Space of the First Two
Significant Discriminant Functions

142

138 'i' Enrolled, Above 2.0, Males

134 i Withdrew, Above 2.0, Males

130 if Enrolled, Above 2.0,
First Discriminant Females
Function
126
"The Intellective if _Withdrew, No GPA, Females
Function" 122 _. X ‘Withdrew, Above 2.0, Females
X Enrolled, Below 2.0, Males
118
114

110 i Enrolled, Below 2.0,
Females

X ‘Withdrew, Below 2.0, Males

 

106 .
32' Withdrew, No GPA, Males

102

100 LWithdrew, Below .2.0, Females

 

O 2 6 10 14 18 22

The Second Discriminant Function, or
"Social-Cultural Background"

91

plotted vertically, differentiates most effectively between above and
below average achievement groups. 0f the tOp five groups on the first
or intellective function, four groups received above a 2.0 grade point
average, while the remaining group received no grade point average.
These group differences support the previous interpretation of the first
function as an intellective function.

The second discriminant function, the social-cultural background
function, effectively discriminated between males and females. Plotted
horizontally, in Figure 1, the social-cultural function placed the five
female groups toward the right of the graph, and the five male groups
toward the left, illustrating that females received higher scores than
males<n the second function. Group differences on the second function
also support the previous interpretation of the second function as a

social-cultural function maximizing sex differences.

Discrimination Between‘Withdrawal Groups and Enrolled Groups

Groups of students who withdrew are contrasted to groups of stu-
dents who remained in college through the end of the freshman year in
Figure 2. All ten groups are plotted in the two dimensional space, the
intellective discriminant function scaled vertically and the social-
cultural discriminant function scaled horizontally. Lines have been
drawn to connect similar sex and achievement groups which withdrew or
remained in college. Of course, no lines are drawn from the withdrawal
groups having no grade point average as they have no counter-part in

the enrollee category.

92

Figure 2

Differences Between the Withdrawal and.Enrollee Groups,
‘Within Similar Sex and Academic Achievement Categories,
Plotted In the Two Dimensional Space of the First
and Second Discriminant Functions

 

 

142
138 -
X Enrolled, Above 2.0, Males
134 X ‘Withdrew, Above 2.0, Males
130 " Enrolled, Above 2.0,
Females
The 126
First Discriminant A
“mm“ 22 x Wathdrew, No GPA, Females
"The Intellective 1 X‘Withdrew, Above 2.0, Females
118
114
110 Enrolled, Below 2.0,
Females
106 X Withdrew, B ow 2.0, Males
I Vfithdrew, GPA, Males
102 _-
100 W 200: Females

 

O 2 6 10 14 18 22

The Second Discriminant Function
"Social-Cultural Background"

93

In all instances it is apparent that when similar sex and grade
point average groups are compared, freshmen who remained enrolled in
college were higher on the intellective function than freshmen who with-
drew. This generalization holds true for both males and females and
for above and below average achievement groups. The dominance of the
intellective function in the discrimination among groups is illustrated
by the fact that lines connecting groups tend to run vertically rather
than horizontally.

Further examination of Figure 2 indicates that within the achieve-
ment and sex categories, the social-cultural function discriminated be-
tween enrolled and withdrawn groups. 'Within sex and achievement groups,
students who withdrew tended to be lower on social-cultural function
than students who remained in college. This tendency is illustrated by
the consistent slant of the lines connecting groups. This difference
would suggest that students who remained in college came from higher
social strata homes and had more feminine characteristics than students
who withdrew.

Examination of group differences on the first two functions
clearly revealed that students who withdrew from college were not a
homogeneous group, and when classified by first term grade point aver-
age and sex, were found to be diverse in social-cultural background and
intellective ability. For example, the second highest group on the in-
tellective function was a group of students who withdrew from college.
Moreover, two out of the bottom five groups in intellectual ability

were groups of students who remained in college.

94

Discrimination Between Above andggplpy
Average Achievement Groups

Differences between above and below average achievement groups
are illustrated in Figure 3. As previously indicated, above average
achievement groups were consistently higher on the intellective dis-
criminant function than the below average achievement groups. All
four above average achievement groups were higher on the intellective
function than the four below average achievement groups, regardless of
collegiate persistence or sex.

The discrimination among the high and low achievement groups
was accomplished almost entirely by the intellective function. There
was little or no difference between above and below average achieve-
ment groups on the social-cultural background function. This, of course,
is as expected, recognizing that the group dispersion defined by the

first function is removed from subsequent functions.
Discrimination Between Males and Females

Differences between males and females in similar achievement
and collegiate persistence groups are illustrated in Figure 4. As pre-
viously noted, all five of the female groups had higher discriminant
scores on the social-cultural function than any of the male groups.
Within a given achievement level and collegiate persistence group, males
tended to be higher than females on the intellective function. An ex-
ception to this generalization was the group of females who withdrew
from college prior to receiving grades. These females were considerably
higher on the intellective function than would be expected from the

pattern of the remaining nine groups on the function.

95

Figure 3

Differences Between Above and Below Average Achievement Groups,
Within Similar Collegiate Persistence and Sex Categories,
Plotted in The Two Dimensional Space of the First

and Second Discriminant Functions

142

138

134

130

The 126
First Discriminant

Function
122

”The Intellective

Function"
118

114

110

106

102
100

 

 

 

 

 

 

X WithdrewJ Below 2.0, Females

 

X Enrolled, Above 2.0, Males
'i' thdrew, Above 2.0, Males
SE Enrolled, Above 2.0,
Females
if Wit drew, No GPA, Females
E. ithdrew, Above 2.0, Females
Enro l , Below 2.0, Males
H Enrolled, Below 2.0,
Females
if Withdre. Below 2.0, Males
i Withdraw, o GPA, Males
0 2 6 10 1L» 18 22

The Second Discriminant Function
"Social-Cultural Background"

96

Figure 4

Differences Between Males and Females, Within Similar
Collegiate Persistence and Achievement Categories,
Plotted In The Two Dimensional Space of the
First and Second Discriminant Functions

   
   
  
  
  
    

 

 

142
138 _.
Fhlrolled, Above 2.0, Males
131+ " With ew, Above 2.0, Males
130 if Enrolled, Above 2.0
Females
The 126
First Discriminant
Function 122 Withdrew, No GPA, Females
"The Intellective Withdrew, Above 2.0, Females
Function” Enrolled, Below 2.0, Males
118
114
110 " Enrolled, Below 2.0,
Females
106 Withdrew, Below 2.0, Males
W1 drew, No GPA, Males
102 _
100 Withdrew, Below 2.0, Females

 

0 2 6 10 14 18 22

The Second Discriminant Function
”Social-Cultural Background"

97

Group Differences on the Third Discriminant Function

While the magnitude of group differences on the third discrimi-
nant function was less than on the first and second functions, the dif-
ferences among groups on the third function were statistically significant
and theoretically interesting. In previous discussion it was suggested
that the third function was a social background function, with the usual
correlates of intelligence and sex removed. The differences among groups
on the third function would lend theoretical support to such a generali-
zation. The means of each of the ten groups on the third discriminant

function are presented in Table 16 in order of magnitude of the group

 

 

 

mean.
TABLE 16
’ GROUP MEANS ON THE THIRD DISCRIMINANT FUNCTION
warm at

‘Withdrew, Above 2.0, Females 15.18
Withdrew, No G.P.A., Males 13.85
Withdrew, No G.P.A., Females 13.00
‘Withdrew, Above 2.0, Males 12.20
‘Withdrew, Below 2.0, Females 11.38
Enrolled, Above 2.0, 1ales 11.38
Withdrew, Below 2.0, Males 11.29
Enrolled, Above 2.0, Females 11.08
Enrolled, Below 2.0, Females 10.60

Enrolled, Below 2.0, Males 10.45

 

98

The highest five groups on the third discriminant function all
‘withdrew from college during the freshman year. Four out of the bottom
five groups remained enrolled in college through the end of the fresh-
man year. It is also of interest to note that among the withdrawal
groups, the four top groups on the third function were males and females
who withdrew from college with above average achievement records, or
students with no reported grades. Similarly, the two lowest groups on
the third function were those who remained enrolled in college, despite
a below average achievement record during the first term.

The groups of students nearest the median on the third function
‘were those whose collegiate persistence was more nearly in accord'with
achievement records. For example, withdrawal groups nearest the median
on the third function had below average first term achievement records.
The enrollee groups nearest the median were above average in first term
achievement.

These differences between students who withdrew from college and
students who remained theoretically support the interpretation of the
third function as a social background function, with sex and intellectual
differences removed.

There were no apparent sex differences on the third function.
‘Within collegiate persistence and sex categories, however, the below
average achievement groups tended to be lower on the third function
than the above average achievement groups. It is thus apparent that
the differences defined by the third function are independent of in-
telligence, inasmuch as the above average achievement groups have a

tendency to fall in the same direction as the withdrawal groups.

99

It may be concluded that the third discriminant function repre-
sents the linear combination of the instruments which best discriminated
between students who remained enrolled in college and students who with-
drew, regardless of achievement level or sex. These differences would
lend theoretical support to the conclusion that the third function was
a social background function with the influences of intelligence and

sex removed.

Summary

The findings of the study failed to support the null hypothesis
that there was no difference in intellectual ability, social background,
attitudes and values among groups of students classified by collegiate
persistence, first term grade point average, and sex. It may therefore
be concluded that such differences did exist among groups of students
so classified.

Three statistically significant discriminant functions were
found. The strongest discriminant function, accounting for more than
half of the total variance among groups, was heavily loaded with cogni-
tive variates. In addition, the social status index contributed posi-
tively to the first function. Although the contribution of the values
inventory was slight in the first function, it was in the direction of
the more traditional "Protestant ethic" values of achievement, morality,
individualism, and orientation toward the future. Both the nature of
the discrimination among groups and the weights accorded the variables
indicated that the first discriminant function was primarily an intel-

lective function.

100

The complex of variates forming the second significant discrimi-
nant function indicated that the function primarily discriminated along
a social-cultural continuum. It was suggested that the interaction be-
tween the Michigan §tgtg‘University Reading Tegt'and the College Qgglir
fication Eggtlresulted in a slight negative weight on numerical science
ability factors, known to be related to cultural background and sex
differences. The social—cultural content of the second function was
further influenced by the strong positive weight given the social status
index. In addition, the Differential Eglggg Inventogy received a nega-
tive weight, in favor of the more emergent values associated with per-
sons from higher social strata and with females.

All of these conditions pointed toward the probability that the
second discriminant function differentiated along social-cultural lines,
maximizing sex differences measured by the variables. Group differences
on the second discriminant function support this conclusion.

The third significant discriminant function was described as an
interaction between a sex related difference on cognitive instruments
and he social status index. The formation of the discriminant score
suggested that an average score on the third function resulted.when the
discriminant value of the cognitive instruments was prOportionate to
the magnitude of the discriminant value of the social status index. A
high discriminant value for the cognitive combination and a low social
status value resulted in a high score on the third function. The inter-
action among the cognitive measures and‘between the cognitive measures
and the social status index was interpreted as resulting in the removal

of the sex factor from the social background element. Thus, the function

101

might be thought of as a social background function with sex and in-
tellectual differences removed.

Group status on each of the functions was discussed. The first,
or intellective discriminant function, differentiated.most effectively
between high and low achievement groups. ‘When similar achievement and
sex groups were compared on the intellective function, students who re-
mained enrolled in college were higher on the function than students who
‘withdrew. In addition, within similar collegiate persistence and achieve-
ment groups, males tended to be higher on the intellective function
than females.

Examination of group status on the second discriminant function
revealed that it differentiated most effectively between males and fe-
males. All five of the female groups were above the five male groups
on the second function. 'Within sex and achievement groups, students who
remained enrolled in college through the end of the freshman year were
higher on the socio-cultural function than students who withdrew during
the same period. There was no apparent difference in the status of high
and low achievement groups on the social-cultural background function.

The third significant discriminant function discriminated be-
tween those students who withdrew from college and those who remained
through the end of the freshman year. Highest on the third function
were the groups of students who withdrew from college with no first term
grade point average. Lowest on the third discriminant function were the
males and females who remained enrolled in college despite a below aver-
age achievement record for the first term. No discrimination between

males and females on the third function was evident. ‘Within collegiate

102

persistence and sex groups, the above average achievement groups tended
to be higher on the third function than the below average achievement
groups. As withdrawal groups were also highest on the third function,
differences among achievement groups were considered to be independent
of the intellective factor.

In the chapter to follow the implications and conclusions of
the study will be discussed. In addition, attention will be directed
toward those questions either left unanswered by this study, or sug-

gested by the findings of the study.

CHAPTER V

SUMMARY AND CONCLUSIONS

Summary

The purpose of this investigation was to study the nature of the
student who withdraws from college prior to the completion of his first
academic year. The study was designed to differentiate between students
who withdraw from college and students who remain in college, and also
to differentiate among different types of students who withdraw from
college such as withdrawals of above and below average achievement and
male and female withdrawals.

The variables of intellectual ability, social background, atti-
tudes, and values were shown to be theoretically related to collegiate
persistence and attendance. Accordingly, the following hypothesis was
formulated and subsequently tested.

It is possible to differentiate among groups of students

classified by collegiate persistence, first term grade

point average, and sex, on the basis of intellectual
ability, social background, attitudes, and values.

The Methodology

The population under study consisted of the 1958 entering fresh-
man class of Michigan State University. Only those students who had
never before attended a college or university, enrolled for full-time
study of 12 credits or more, and.American citizens, were included in

the study. All test data were collected during Orientation‘week.

10#

Students for whom full test records were not available were excluded from
the study. These qualifications observed, the total population numbered
2,746 Michigan State University freshmen students.

The first term grade point average of all students was obtained
from the Office of the Registrar, Michigan State University. The re-
cords of the Registrar's Office were also used to identify those fresh-
man students who withdrew from college, or failed to re-enroll.

At the conclusion of the 1958-1959 academic year, 303 students
from the total group of 2,746 students had withdrawn from the University.
All of these 303 students were included in the withdrawal sample of the
study.

A random sample of 250 students was drawn from the remaining
number of 2,4h3 students who were enrolled at the University through
the end of the academic year.

Both the "withdrawal" and "enrollee" groups were further classi-
fied into sub-groups on the basis of sex and first term achievement. A
grade point average of 2.0 was selected as an appropriate criterion for
the classification of students into above and below average academic
achievement groups. Classification of withdrawals and enrollees by
achievement and sex yielded ten groups, six of which.withdrew from col-
lege and four of which remained enrolled.

To test the hypothesis of the study, it was necessary to collect
data relevant to the intellectual ability, social backgrounds, attitudes,
and values of the members of the sample. Intellective characteristics
were measured by three cognitive instruments: (1) Th2.ggllggg_gg§li£ip

cation Test, a general academic ability measure; (2) Thngest 2£_Critical

105

Thinkin , £252.93 a measure of logical reasoning and problem solving
aspects of critical thinking ability; and, (3) Thg_Michigan Stgtghggig
versity Reading Eggt, a measure of the ability to comprehend thoughts
expressed in reading passages representative of textual materials found
in the several academic areas in college.

The measurement or quantification of social background was ap-
proached.by combining the biographical variables of father's occupa_
tional prestige rating, level of father's education, and level of
mother's education into a single index. First factor loadings from a
factor analysis of the correlation coefficients among the three variables
'were used to combine the three variables into a single index score.

The index score was assumed to measure the general social status of
the family, and hence, was called the social status index.

Attitudes were measured by two instruments, the Inventogg o_f_‘_ Eg-
lj._e_f_§, form I, and Rokeach's Domtism §_c_:_a_l_e_, £3133 E. Both instruments
'were designed to measure related phenomena, the former measuring
stereotypy, or rationality of belief systems, and the latter measuring
dogmatism, or general authoritarianism.

A measure of values was obtained through the use of the Qiffggr
gatigl_!§lgg§ Invento , an instrument designed to measure the degree
of commitment to the ”Protestant ethic" values of achievement, indi-
vidualism, puritan morality, and orientation toward the future.

Multiple discriminant analysis was selected as an appropriate
statistical technique. Through the use of multiple discriminant analy—
sis, it was possible to define the linear combinations of variates which

.maximized the differences between groups and minimized the differences

106

within groups. Through the examination of the discriminant coefficients
which determine the linear combinations, it was possible to gain insight
into the nature of the basic phenomena which characterized the student
who withdrew from college.

The first linear combination of the multiple discriminant analy-
sis maximized the discriminant criterion; the second combination maxi-
mized the discriminant criterion after the effects of the first combination
had been removed; the third linear combination maximized the discriminant

criterion after the effects of the first two had been removed.

The Test of the Hypgthesis

The analysis of data yielded three significant discriminant
functions. The null hypothesis was therefore rejected. A test of sig—
nificance of the latent roots of the determinental equation indicated
that the first two functions were statistically significant at the .001
level of confidence, and the third significant at the .05 level of con-
fidence. Three additional discriminant functions were obtained, but

‘were not found to be significantly different from chance.
The Nature of the Three Significant Discriminant Functions

The first discriminant function accounted for over 60 per cent
of the total dispersion among groups as defined by the variables, and
was considered to be an intellective function. The College Qualifica-

tion Test, the Test g£_Critical Thinkin , and the Michigan State

107

University Reading Tg§t_were all weighted positively in the first func-
tion, and were the three most heavily weighted variates. The social
status index was weighted positively in the first function and was the
fourth heaviest weighted variable. The Differential Mglgg§_lnvento§z
was weighted positively in the first discriminant function, favoring

the traditional values of achievement, individualism, puritan morality,
and orientation toward the future. The weights given the two attitudinal
measures were so small as to suggest that the instruments did not mater-
ially affect the nature of the first function. Examination of the dis-
criminant coefficients in the first discriminant function revealed that
the function was a linear combination of intellective elements buttressed
by social background and value elements which complemented the cognitive
complex.

Careful examination of the discriminant coefficients of the second
discriminant function revealed that the function could be considered to
be social-cultural in nature. Such a conclusion was supported by the
positive contribution of the social status index, and the negative weight
given the Differential Eglggg Inventogz. The Michigan Stgtg University
Reading Tg§t_had a strong positive weight in the second discriminant
function. The College anlification.Tg§t, however, had an even stronger
negative weighting. It was concluded that the interaction among the
cognitive measures in the second discriminant function resulted in a
slight negative weight on numerical-scientific abilities, associated
with cultural and social differences.

The third discriminant function was formed through an interac-

tion between the cognitive instruments and the social status index.

108

This interaction resulted in a removal of sex differences from the
social status index, especially all those sex characteristics defined
by second discriminant functions. It was concluded that the third
function represented a.measure of social background factors with the

usual correlates of intelligence and sex removed.

Group Differences on the Three Siggificant Discriminant Functions

The intellective discriminant function, differentiated most ef-
fectively between the above average achievement groups and the below
average achievement groups. All of the above average achievement groups
were higher than any of the below average achievement groups on the in—
tellective function, regardless of collegiate persistence or sex. ‘When
similar sex and achievement groups were compared on the intellective
function, it was apparent that, on the whole, students who remained en-
rolled in college were higher on the intellective function than students
who withdrew. This finding is, of course, in accord.with the findings
of previous investigations. It was also evident, however, that hetero-
geneity existed among the different withdrawal groups, equally as great,
if not greater, than the heterogeneity among students who remained en-
rolled. Finally, when similar collegiate persistence and achievement
level groups were compared, males tended to be higher on the intellective
function than females.

‘ Examination of group differences on the second, or social-cultural
background function, indicated that the most marked discrimination of
the second function was between males and females. All five female

groups were higher than the five male groups on the social-cultural

109

background function, regardless of collegiate persistence and achieve-
ment. A slight difference was apparent between the withdrawal and
enrollee groups when similar sex and achievement level groups were com-
pared. There was no apparent difference, however, between the above

and below average achievement groups on the second discriminant function.

The third significant discriminant function was not as powerful
in differentiating among groups as were the first two. The patterns of
groups differences, however, were especially interesting. The third
discriminant function differentiated primarily between groups of stu-
dents who withdrew from college and groups of students who remained in
college through the end of the freshman year. Especially interesting
was the fact that various, apparently atypical groups, appeared on the
extremes of the distribution of mean scores on the third function. For
example, the highest groups on the third function were composed of stu-
dents who withdrew from college with no first term grade point average
or an above average achievement record. Conversely, the lowest groups
on the third discriminant function were males and females who remained
enrolled in college despite a below average achievement record for the
first term.

Although the third discriminant function discriminated most
clearly between withdrawals and enrollees, a slight difference was
apparent between above and.below average achievement groups. The dif—
ference, however, was other than one would expect if the discrimination
between withdrawals and enrollees were primarily due to a cognitive ele-
ment. ‘While those who withdrew from college were higher on the third

function than those who remained enrolled, there was a slight tendency

110

for the above average achievement groups to be higher on the function

than the below average achievement groups.
Conclusions and Generalizations

Differences Between‘Withdrawals and Enrollees,
Eithin Sex and Achievement Categorieg

The findings of this investigation tend to support the findings
of previous research on the problem of collegiate withdrawal. ‘Within sex
and achievement categories, students who remained enrolled in college
had more general academic ability than those students who withdreW.
Moreover, within sex and achievement groups, those students who with-
drew from college tended to be lower on the social-cultural function
than students who remained in college. It is concluded, therefore, that
intellectual ability must continue to be considered as related to with-
drawal from college. Students who find it difficult or impossible to
achieve at the required levels of performance in college will naturally
have a higher attrition rate than the more capable students.
Differences Between‘Withdrawals and Enrollees,
Regardless of Sex and échievement

When students who withdrew from college were separated into
groups on the basis of first term achievement and sex, and compared to
similarly separated groups of students who remained enrolled in college,
the first two significant discriminant functions did g93.differentiate
gll_withdrawal groups from gllIgroups of students enrolled in college.

Such a finding clearly illustrates the fact that students who withdraw

111

from college are not a homogeneous group, but conversely, must be con-
sidered a diverse, heterogeneous group in terms of academic ability.
The data also suggest the possibility that the proportion of males and
females in a withdrawal sample, or the proportion of above average and
below average achievement students in a given sample of withdrawals,
may be more instrumental in the determination of differences in intel-
lectual ability or differences in social background than the collegiate
persistence status of the group.

The third discriminant function most effectively'separated'with_
drawal groups from groups of enrollees. The nature of the third function
is certainly less clear than the nature of the first two functions. It
was suggested, however, that the third discriminant function was a rather
pure social background function, with the usual correlates of intelli-
gence and sex related social-cultural elements removed. ‘With the removal
of the intellectual factors most closely related to achievement, and
the removal of the socialécultural factors related to sex, the element
most closely related to the collegiate persistence then remained. It
may be that the third function is an indirect measure of the social-
psychological motivation to remain in college. That is, the linear com-
bination of elements which removed the sex and intellective bias from
the social status index, thus leaving a ”pure” measure of social back-
ground, could theoretically reflect the environmental influences on the
student prior to admission to college.

If the above Speculations were correct, the nature of the third
discriminant function would agree closely with current social-psychologi-

cal theory. Such theory would suggest that the pressures from the

112

family, peers, and self directed toward attending and continuing college
are greater in a high social status home than in a low social status

home regardless of ability or sex. In fact, the pressures in a lower
social status home may be directed toward non-attendance or non-persistence.

A student with less than average ability, but from a high social
status home, may be unable to achieve at a satisfactory level in college,
but may continue to remain enrolled in college due to the pressures and
social needs created by his previous and current social environment.
Accordingly, a student of high ability may leave college because the
pressures of his prior and current social environment motivate him to
do so.

No claim is made that all of the pressures are in the form of
aspirations and expectations of self and significant others. Admittedly,
pressures may be largely financial, or may be a combination of financial
pressures and social-psychological factors.

The Research Division of the National Education Association,
after reviewing the research on high school drop-outs, reached the
following conclusions:

The typical drop-out's parents are unimpressed with the value

of education; often they openly scorn "book learning.” In

addition, the family is likely to regard school as a financial

burden; not only does it cost something to keep a child in

school, but the family is deprived of the money which the

boy or girl could be contributing to the budget.... Probably

the most important home connected factor is the attitudes of

the parents toward the child's education and toward school

in general. A child's attitudes are shaped largely by the

feelings of his parents...and are reinforced by the attitude

of the community. The value systems pf those around him
inevitably color the child's outlook.

 

1Research Division ofihe National Education Association, "High
School Drop-Outs,” Research Bulletin, 38:11-1#, February, 1960, pp. 12-13.

113

Although no definite conclusions can be made as to the nature of
the third discriminant function, the evidence suggests the possibility
that the function is a measure of the social-psychological motivation

to remain in college, independent of achievement or sex differences.

Differences Among the Various‘Withdrawal Groups

 

The findings of this study suggest that there is apparently an
ability difference between male and female Michigan State University
freshman students. Such a difference, in favor of males, was apparent
both within the withdrawal groups and.within the enrolled groups of
students.

It is difficult to conclude why males, within collegiate per-
sistence and achievement categories, were consistently higher on the
intellective discriminant function than females. There are four pos-
sible conclusions as to why the apparent discrepancy in ability between
males and females was found: (1) there are differences between the
Michigan State University males and females in general academic ability;
(2) the instruments which purportedly measured the cognitive elements
had a sex bias; (3) the multiple discriminant analysis technique maxi-
mized any existing sex differences in differentiating between the dif-
ferent groups; and (h) the possibility of sample bias. Any one, or
combination of the four, could provide an answer to why sex differences
were found on the first discriminant function. Further research is
necessary before any definite conclusions may be formulated.

The females who withdrew before receiving a grade point average

deviated sharply from the pattern established by the remaining nine

111+

groups examined on the first or intellective function. This group of
females scored unusually high on the first discriminant function, indi-
cating above average academic ability.

It is interesting to note, however, that while this group of
females was placed in an unusual status on the intellective function,
the position of the group on the third discriminant function would indi-
cate that the group was similar to other withdrawal groups. It is pos-
sible that Special circumstances such as marriage, pregnancy, or home-
sickness may have combined with low social status to precipitate
withdrawal regardless of above average ability. Of course, it may also
be that the social status of females is more potent in the determination
of collegiate persistence than in males.

Discussion of the Contribution of the Variggg
Instruments In Differentiating Among Groups

Although no probability test was made to test the statistical
significance of the contributions of each of the variates in the dis-
crimination among groups, it was possible to observe the relative size
of the discriminant coefficients of each function, and to suggest the
dominance of particular instruments in the functions. Recognizing the
lack of probability statements, the following remarks must be made with
reservation and caution.

The three cognitive measures, including the Michigan.§tgtgwygir
versity Reading Iggt, the College Qualification Eggt, and the Eggtflgf
Critical Thinkin , were most prominent in the differentiation between
different achievement level groups. The cognitive instruments were

heavily weighted in each of the three significant discriminant functions,

115

indicating that both intellective and social factors were measured by
the instruments.

The social status index consistently contributed to the formation
of all three functions, suggesting that the index, constructed from the
variables of father's occupational prestige, mother's educational level
and father's educational level contained both cognitive and non-cognitive
factors.

The Differential Eglggg Inventogy contributed modestly, but in
the theoretically expected directions, suggesting that while the values
inventory was not a primary factor in any of the three functions, the
“Protestant ethic” values are related to achievement and socio-cultural
sex differences. The lack of influence of the Differential E2;22§.l2r
ventogy in the third discriminant function suggested that the measured
values were not related to collegiate persistence.

The contributions of the attitude measures of Rokeach's Doggatiam
Sgglg,and the Invento§1‘2§,Beliefs suggested that the attitudes measured
by these two instruments were not important in differentiating among
groups of Withdrawals and enrollees, classified by achievement level
and sex. Although the instruments had the heaviest weighting in the
third discriminant function, each tended to negate the influence of

the other.
Implications For Further Research

The study of students who withdrew from college during the fresh-
man year has revealed several unanswered questions and possibilities

for further research. The more important of these suggestions are listed

below.

116

1 . Of first priority is the need for further study of the nature
of the third discriminant function. The third function differentiated
between groups of students who were enrolled in college and students who
withdrew from college, regardless of the sex or prior collegiate achieve-
ment. Equally interesting was the observation that the enrollee groups
whichwouldnornallyhavebeenpredicted, onthebasisoffirstterl
achievaent, to withdraw from college, were lowest on the third function.
Moreover, attheotherextreneonthethirdfunctionwere groups of stu-
dents who withdrew fro- college prior to receiving any notice of achieve-
neut, or subsequent to achieving at an above average level.

Further exploration of the nature of the third function would
have the possibilities of achieving a more definite understanding of the
nature of the function, and also of eventually constructing a more
precise and stronger measure of the phenomenon.

2. Furtherresesrch shouldbedirectedtowardanore exact defini-
tion of differences in collegiate persistence between males and false.
Attention should especially be directed toward the possibility that the
social-psychological activation to rush in college, or to withdraw,
ney be more influential in the case of females than with sales.

3. Itwouldbedesirabletcusethebasicdesignofthisstudy
to investigate collegiate withdrawal during the ”mo-on, Junior, and
senior years in college. Each year, a separate study could be conducted,
and a comprehensive stucbv including all withdrawal groups at all points
intine couldbeconductedattheendofthefouryearperiod. These
investigations would serve to validate the findings of this studv, and
would also add a in. dimension.

117

4. Further research is needed concerning the relationship between
collegiate persistence and student attitudes and values. The findings
of this study would suggest that attention should be directed to attitudes
and values related to college attendance and schooling in general. Also,
attention should be given to the relationship of values or standards of

personal achievement, the self concept, and collegiate persistence.

1 1 8
BIBLIOGRAPHY

Bendix, Reinhard, and Lipsit, Seymour Martin. ClassLStatus, and Power.
Glencoe, Illinois: The Free Press, 1953.

Bennett, George K., Bennett, Marjorie G., Wallace, Winburn L., and

Wesman, Alexander G. College CEalification Tests, ManualI 1252.
New York: The Psychological Corporation, 1957.

Berry, C. 1., and Jones, A. L. "Factors Involved in the Withdrawal of
Students from Grambling College at or Before the End of Their
Freshman Year." Journal 2; Negro Education. 25:1145-47, 1956.

Bragg, Emma W. "Study of Student Withdrawal at 'W.U."' The Journal
at; Educational Psycholog. #7:199-203, April, 1956.

Bryan, Joseph G. _A_ Method f9}; thg___ Exact Determination _o_f th_e _C____harac-
teristic Euation and Latent Vectors o__f_ a Matrix wi___th— Applications
t_9_ the Discriminant Function £93; More_ Than 2133 Groups. Cambridge,

Mass.: Harvard University Graduate School of Education, Un-
published Dissertation, 1950.

Computer Laboratory, Michigan State University. ”K5-M, Correlation,
Means Standard Deviation, Variance, Card Input.” MISTIC Library
Index. East Lansing, Michigan, 1959.

Computer Laboratory, Michigan State University. "MS—139, Solution of
Determinantal Equation A -,\B = o." MISTIC Libragx Index.
East Lansing, Michigan, 1959.

Cooperative Study of. Evaluation in General Education, Paul L. Dressel,
Director. Instructor's Manual 33.1.: 3.32 Inventor“)! 9f Beliefs.
The American Council on Education, Committee on *easurement and
Evaluation, 1953. (Mimeographed).

Cooperative Study of Evaluation in General Education, Paul L. Dressel,
Director. W mg; .1311; 3922 __Test 2£ __Critical Thinkin
m G. The American Council on Education, Committee on Measure-
ment and Evaluation, 1953. (Mimeographed).

Dressel, Paul L. and Mayhew, Lewis B. General Education: _E_xplorations
in Draluation. Washington, D.C.: American Council on Education,

19

Eichorn, Dorothy H. and Jones, Harold E. "Development of Mental Func-
tions.” Review 9;. Educational Research. 22:421-438, December,

1952a

119

Farquhar,‘WilliamTW. and Krumboltz, John D. "A Check List for Evaluat-
ing Experimental Research in Psychology and Education." Journal
2£_Educational Research. 52:353-35h, May, 1959.

 

Gable, Robert Irvin. A Stqu_ of the Studentm Wp:Qg Problem at Miami
University. Bloomington: —Indiana University. Unpublished
Dissertation, 1956.

Grace, Harry A. "Personality Factors and College Attrition." Peabody
Journal 2; Education. 35:36-40, July, 1957.

Hagood, Margaret and Price, Daniel. Statistics for Sociologists. New
York: Henry Holt and Company, 1952.

Hanks, Charles J. A Comparative Stugy_ of Factors Related to Retention
and Withdrawal of Freshmen Students at the UniversityZof Arkansas.
Fayetteville: University of Arkansas, _Unpublished Dissertation,

1954.

Halladay, D.‘Whitney, and.Andrew, Dean C. “Drop-Outs from Arkansas
Colleges.” Personnel gpd’Guidance'Journal. 37:212-13, November,
1958.

Hollinshead, Bryon S. Who Should Go to College. New York: Columbia
University Press, 1952.

Holmes, Charles H. "Why They Left College: A Study of Voluntary Fresh-
man‘Withdrawal from the College of Liberal Arts at Syracuse Uni-
versity." College gpd'University. 34:295-300, September, 1959.

Hood, Albert Bullard. Certain Non-Intellectual Factors Related 32
Student Attrition g§|Cornell University. Ithaca: Cornell University,
Unpublished Dissertation, 1957.

Iffert, Robert E. Retention gpngithdrawal 2§,College Students. U.S.
Department of Health, Education, and‘Welfare, Office of Education
Bulletin 1958, No. 1. 'Washington, D.C.: United States Government
Printing Office, 1957.

Lehmann, Irvin J. and Ikenberry, Stanley 0. Critical Thinkin , Atti-
tudes, and Values in Higher Education: A Preliminary Repgrt.
(Paul L. Dressel, Principal Investigator) East Lansing, Michigan:
Office of Evaluation Services, Michigan State University, 1959.

Lehr, Milton‘W’. A Statistical Description of Factors Related to Dro op-
Outs and Non-DrOp:ths_ at Northwestern State College. Norman:
University of Oklahoma, Unpublished Dissertation, 1956.

Lindquist, Charles B. College_ and University Facilities: Recent Per-
sonnel Egg;;pstructional Practices. United States Department of
Health, Education, and welfare, Office of Education Bulletin No.
27. 'Washington, D.C.: U.S. Government Printing Office, 1959.

120

Line, L. J., and Pitt, Hy. "The 'Staying Power' and Rate of Progress
of University of Wisconsin Freshmen.” College £151, Universit .

29:86-99.

McGrath, Earl J. and Russell, Charles H. Age Liberal Arts Colleges _B_e'-
coming Professional Schools? New York: Institute of Higher Educa-
tion, Columbia University, 1959.

Hunger, Paul F. ”Student Persistence in College." Personnel and
Guidance Journal. 35:2141-14-3, December, 1956.

Pinneau, Samuel R. and Jones, Harold E. "Development of Mental
Abilities.” Review 9; Educational Research. 28:392-h00,
December, 195 .

Plant, Walter T., Minium, Edward W., and Myers, Celestine. "An Analy-
sis of the Rokeach Dogmatism Scale Used with a Sample of American
College Students.“ (A paper read at the Annual Meeting of the
Western Psychological Association, San Diego, California, April 16-
18, 1959) (Mimeographed).

Prince, Richard. A Study o_i; 213 Relationshi; Between Individual Values
.3129 Administrative Effectiveness 1.31 3112 School Situation. Chicago:

University of Chicago, Unpublished Dissertation, 1957.

Rao, C. Radhakrishna. Advanced Statistical Methods in Biometric Re-

search. New York: John Wiley and Sons, Inc., 1952.

Research Division of the National Education Association. ”High School
Drop—Outs." Research Bulletin. 38:11—11+, February, 1960.

Rokeach, Milton. "The Nature and Meaning of Dogmatism." Psychological
Review. 61,3:1911—20lk, 1951+.

Rokeach, Milton. "Political and Religious Dogmatism: An Alternative
to the Authoritarian Personality." Psychological Monographs.
NO. #25, 1956.

Slocum, W. L. "Social Factors Involved in Academic Mortality."
College and. University. 32:53—64, 1956.

Smith, Margaret Ruth. A Study o_i: First Year Drogflts At Wme State
University. Detroit, Michigan: Wayne State University, 1957
(Mimeographed).

Smoke, E. Eileen. A Comparison 9; the Graduates £29; Non-Graduates 9};
th_e_ Class 9_f 1251 £39. Indiana Universit . Bloomington: Indiana
University, Unpublished Dissertation, 1955.

Spindler, George. "Education in a Transforming American Culture."
Harvard Educational Review. 25:158-163, September, 1953.

121

Tatsuoka, Maurice and Tiedeman, David V. "Discriminant Analysis."
Review 2;: Educational Research. 2h:1+02J+20, December, 1954.

Tiedeman, David V. and Bryan, Joseph G. "Predictions of College Field
of Concentration.” Harvard Educational Review. 2h:122—39,
Spring, 1954.

Tiedeman, David V., Bryan, Joseph G., and Rulon, Phillip J. 2113 Utility
_e_f, 3113 Airman Classification Batteg Fo_r Assiment 93 Airmen 1.2
Eight A}; Force Sgcialties. Cambridge, Mass.: Educational Re-
Search Corporation, June, 1951.

Harrington, Willard G. and Saupe, Joe L. "Development and Applications
of Test of General Mental Ability." Review 9; Educational Research.

Wilks, S. S. "Weighting Systems for Linear Functions of Correlated
Variables When There is No Independent Variable.” Psychometrika.
332MB, MErCh’ 19380

Wooster, G. F., and Stover, W. W. “Lost: Students.” Educational
W Mo 37:85-90. April. 1958.

Yoshino, Roger. "College Drop-Outs at the End of the Freshman Year."
Journal 9; Educational Sociology. 32:l+2-J+8, September, 1958.

APPENDIX A

GROUP I'IEANS FOR ALL TEN GROUPS AND FOR THE TOTAL GROUP

ON THE MEASURES EMPIDYED IN THE STUDY

123

 

No.me sm.mna om.om «a.mefi mw.nm :a.em we.se mm amass .o.m
o>op< .eoaaoacm .a
ma.me mm.ooH mm.sm aa.oafi em.mm mm.mm mm.ae em moHASma .o.~
:oaom .emHHOtcm .w
ma.oe mm.eaa ow.mm mm.aea mm.mm mm.om Hm.oe mm mwaes .o.m
:onm .eoaaoacm .a
ma.wm es.:aa ma.mm ma.oaa ma.sm ma.om w~.se an amassma .o.m
o>on< .Hezategpax..e
mm.mm mm.wma Hm.mm om.oeH mm.mm “a.mm ae.ee mm amass .o.m
o>on< .Hmzeeegpnz..m
ma.wn os.mm me.fim He.oaa an.mn «3.3m wa.mn we «masses .o.m
:oawm .Hazaeegpaz..s
wa.nm as.e0a mm.mm mo.maa ma.sm Ho.am mm.wm man mode: .o.~
:oamm .1553: .m
ca.mm 0a.aaa me.wm mm.mea om.mm om.am mm.ae om mmaesoa ..<.m.o
oz .Hezaeeaswz .N
Hm.aa ma.mo~ Hm.m~ mm.aaa mm.mm an.mm so.am mm ”was: ..¢.m.e
oz .Hazaeegpaz..a
K0105” amok coapmo meadow 533600 Showcasing” mafixcfla mmowaom @595 mcofifiwuowon 95.5
seesaw naeaflasa .s.m.zc m.eoaoxom sun assesses no we
deacon oonHoo bopcanH ouwm

 

 

Baum “a; 2H ESE 35mg mm“. 20 Abomo

.289 ES. mg 8,2. 950mm zma Add mom mzé nSomU

4 xaecmaaa

12h

 

Hm.mn sm.mfia ma.em ea.aea mo.sm «a.mm om.He mun
mm.ae ae.mma ma.am mm.ana Ha.mm oa.mn mm.me an

Amsomo Hmpoa pomv

monsom .o.m
o>on¢ .eeaaoecm .Oa

 

NoocH umoe coapmo maficmom Emfipmsmon huovco>cH mcwxcflna mmowaom macaw

mapwpm lamaaesc .D.m.z m.:omoxom >19 Heowpflao no mo
Hwfloom omoHHoo huopco>cH beam

ncoapmauowon acouo.

 

Aco::flpcoov 4 xfiocomd<

APPENDIX B

MATRIX A AND MATRIX w OF THE DEPERMINAN’I‘AL EQUATION
'A - Aw, v = o

A = The Among Groups Matrix
W = The Within Groups Matrix
V = The Latent Vectors of the System

A = The Latent Roots of the System

126

APPENDIX B

THE AMONG GROUPS MATRIX

 

 

 

 

 

+ooso7o +005112 -ooozso -007018 +005u31 +0185u3 +006669

+005112 +006912 +ooo752 -007267 +0065h1 +027028 +006157

-ooozso +ooo752 +000866 +00083u +ooou75 +oou582 -001297

-007018 -007267 +ooos3u +01u075 -007633 +025735 -010921

+oosu31 +0065h1 +ooou75 -007633 +006725 +027678 +006h66

+0185u3 +027028 +004582 -025735 +02u678 +116785 +016044

+006669 +006157 -001297 -010921 +006466 +016044 +015078
THE‘WITHIN GROUPS MATRIX

+0999»?

+011858 +024850

-001661 +000848 +027982

-107755 -015852 +012204 +348583

+009816 +011672 +ooo7oz -014658 +019284

+035539 +041898 +002398 -039110 +044227 +273923

+008059 +001129 -008326 -009436 +002081 +018575 +108602

 

APPENDIX C

STANDARD DEVIATIONS OF THE PEASURES EMPLOYED

IN THE STUDY FDR THE TOTAL GROUP

128

APPENDIX C

STANDARD DEVIAT IONS OF THE MEAS URES FA‘IPIOYED

IN THE STUDY FOR THE TOTAL GROUP

 

 

 

 

 

 

Variable Standard Deviation
Inventog p_i_'_ Beliefs 13.75
Test 32 Critical Thinking 7.58
Differential Mg Inventog 7.22
Rokeach's Dogmatism M 25.61
Michigan §_t_a_._t_g University Reading leg; 6.86
College Qualification Egg}; 26.58

Social Status Index 11+. 95

 

APPENDIX D

CONVEN‘T IONALIZED DIS CRDIINAMT COEFFICIENTS CORRESPOI‘IDING

TO THE SIX DISCRIMINANI' FUNCTIONS

130

 

 

 

 

 

 

 

 

 

 

 

 

sauna. omneo.- eases. mmaem.u semen. momma. xoecH seesaw Hmaoom
asmma. maesa.s womea.a emaofi.u mmmmm.u eases. smog compaoaeaaeed omoaaoo
Homam. mm . aaaeo.- ooooo.H ooooo.H 00000.“ . peas mcaemom

9% H3222: 33m 3mg
ammao.a anaem.a mamsm. «memo. meeoo.a mammo.- oaeom amapaamom m.gomoxom
«meme. anode. ooooo.a anemoo. mesma.u aoomm. unopcm>cH mozae> Heapcosmemam
00000.“- «enema. mamas. wasmn.- mamas. omamm. unmanaga Heoasapo_mm smog
emon. ooooo.H- emomo.n omens. awamo. ammmo.u meoaaompmm_umopco>cH

o> m> +5 m> N> a> 033mg?

 

 

D NHszmmd

mZOHEE bgHEomHQ NHm BE. CH
GZHBHOmmmMmOU mEEHOHEOU “ES/“Hagan" EH38” 928200

> a
1‘- .‘u

“4%.,

 

 

 

 

"iti-

 

 

 

 

 

"IMINIMAL?!NI