4
l

I
l

w

I
rl

NH

|
h

w
W

 

A STUDY 2’3? THE RELATSO'NSHEP EETWEEN THE
MULTEF’LE SCALWERMu-t ANALYSIS ERROR $CGRE
AN!) CER‘I’ézfié MRSC‘NALET‘E VAREABLES

mm. (*3? fits: begun» a? M. A.
MiCHfifiAR WAT? UNL‘!§§€$!TY
James 3. Bahama
39522

W!
n. ’4

 
   

Michigan Sta tc
University

 

A Study of the Relationship between
the multiple Scalogram Analysis Error Score
and Certain Personality variables

by

James J. DeJonge

A THESIS
Submitted to the College of Science and Arts
Michigan State University of.Agriculture and
Applied Science in partial fulfillment of
the requirements for the degree of

MASTER OF.ARTS

Department of Psychology

1962

ACKNOWLEDGMENTS

The author wishes to eXpress his gratitude
to those who have assisted in the design and
execution of this investigation. He wishes
especially to thank Dr. Charles F. Wrigley for his
help in the design of this study, Dr. Alfred G.
Dietze for his patience and generous assistance in
the preparation of this manuscript, and Dr. John R.
Hurley and Dr. William T. Stellwagen for the time
and constructive criticism which they offered. In
addition, the author wishes to thank the Michigan
State University Computer Laboratory for the use
of their facilities.

11

A Study of the Relationship between
the Multiple Scalogram Analysis Error Score
and Certain Personality Variables

by
James J. DeJonge

AN ABSTRACT
Submitted to the College of Science and Arts
Michigan State University of Agriculture and
Applied Science in partial fulfillment of
the requirements for the degree of

MASTER OF.ARTS

Department of Psychology

1962

ABSTRACT

The problem with which the present research
was concerned resulted from a study of Lingoes in
which a multiple scalogram analysis was carried out
on the responses of 100 subjects to the 30 items on
the K scale of the Ennnesota Multiphasic Personality
Inventory. Although no difference was found to exist
between subjects who were patients in a mental hospital
and subjects who were not patients in terms of their
total scores on the K scale, there was a significant
difference in the number of errors contributed by the
two groups and subjects could be reliably separated on
the basis of error scores. The present investigation
sought to determine whether differentiation between
normal subjects with respect to personality variables
is possible on the basis of multiple scalogram analysis

GI‘I'OI' SCOI‘GB o

The subjects for this investigation were 126
students at Michigan State University. Two tests were
administered to all subjects in a group testing
situation. The first of these was presented to the
subjects as a public Opinion questionnaire dealing
with the foreign policy of the United States. The

second instrument was the Edwards Personal Preference

Schedule.

The responses of all subjects on the public
opinion questionnaire were analysed by the multiple
scalogram method and scored in terms of error scores.
Product moment correlations between error scores and
scores on the EPPS scales were computed for all

subjects.

Significant positive correlations were found
to exist between error scores and scores on EPPS
scales Succorance and Aggression. EPPS scales Change
and Intraception were found to be significantly
negatively correlated with error score. Speculation
was presented regarding the cause of this relationship
and it was suggested that response inconsistency was
due to a difference in meaning or significance given
to certain test items by high error producing subjects.

Some suggestions for future research were made.

Kafka/Z]. ZUca Ag)

Major Professor

Approved

 

 

Date [ZLL=%LL;,;{ I) I7@ '2
VJ

Chapter

TABLE OF CONTENTS

I. INTRODUCTION . . . . . . . . . . .

Scalogram Analysis
Multiple Scalogram Analysis
Multiple Scalogram Analysis
of Hypothetical Data
Scalogram and Multiple Scalogram
Analysis: A Comparison
Previous Research
Problem

II 0 MTHOD O O O O O O O O O O O O O 0

Subjects
Tests
Procedure

III. RESULTS . . . . . . . . . . . . .

Multiple Scalogram Analysis
Product Moment Correlation

IV. DISCUSSION . . . . . . . . . . . .

BIBLIWRAPM O O O O O O O O O O O O O O O

APPENDIX.A
APPENDIX B

111

Page

. 1

. 21
. 26
. 30
. 37
. 38
. 40

Table

9.

10.

LIST OF TABLES

Hypothetical Reaponse Matrix . . .
Hypothetical Response Matrix . . .
Partially Reflected Matrix . . . .
Item Agreement Matrix . . . . . . .
Dimension 1*. . . . . . . . . . . .
Dimension II . . . . . . . . . . .

Frequency Distribution of MSA
Error Scores . . . . . . . . . .

Correlations between Error and EPPS
Scale Scores . . . . . . . . . .

Manifest Needs Associated with EPPS
Variables...........

Correlations between Error and EPPS
scaleSCOrGSOOOOOOQOoo

iv

10
11
15
16

28

29

35

39

I. INTRODUCTION

There are two ways of approaching the study of
attitudes. The first might be called the descriptive
approach since the aim of this approach is primarily to
describe or to measure the phenomenon called attitude.
The emphasis here is on the distribution and the
subgroupings of attitudes within.a given population,
and the social scientist using this approach is
essentially seeking to determine Egg; people believe.
This approach is typicallyrepresented by the Opinion-
polling studies which are so popular around election
time. The nature of the second approach is more along
evaluative rather than descriptive lines. The second
approach is usually more interesting psychologically
for the scientist using this approach is seeking to
determine Eg1_peop1e believe the way they do, i.e.,
the scientist seeks a relationship between attitudes
and personality, demographic, or cognitive factors, in
addition to determining what peOple believe. While
the measurement of attitude has been carried out with
enthusiasm by investigators, perhaps due to the
interest of the public in such studies, surprisingly
little research has been done to determine those

I

 

 

 

   

   

 

. l -
v 7' ' ' I I
. s _ A .
e n a u - — , . ,-
. . . . .
. u 0 n , .
t - - r \ _
. v .
. -. A ~ .
. a . . - .
. . . ,
~ . . - e ‘ n o , . .
., I
O
' ‘ _ . . . . . o
_ 1 _ , . -
g e — o a a . - .
. . ‘ I ’ ' ’ w . '
. ,7 «a
. I. a u v ' i a r. .
.
. , ..
e I - _ ' c
r ’ ' ' -
.. . g - V v .
.
. H ‘- ' ." _ ‘ e 0-. . . - o
. _, . . . .
w 0 \. ' ‘ ' '
‘ . ,“ ' . 'l H
v , U u- l O ' . - u -
\ . - - ' - (a . ,
Q . . ..
- o . afifi .7- .. -- -
.. .- . .- . . . -
.A I i
. . . o - - v e A . .
. . V A i - - , ‘ u v . 4 . V
n . ' ’
.'. _ \ - . . _
I t '
‘ v_ 7 _ _ _t‘ . l, A . . 7 ° '
. ‘ .
C O - ' - . c
«as -- -
o I n - ‘ n o
I i . ,
, , 4 j r‘ ‘ 4 - . \ a .‘ "‘
A A _ _ . . . r
I l ’ _. ‘ . < ,

v _ u r - ‘ ' ' ' ' e \V" "‘ I ' ' ' '|
Q . , . ~ ' t - . v . .-
. c

— . o n - ‘ _ - c . J - . . .
.. . v . r -
e a - "’
’ . . v u p e
. - . ,- . ‘ ~
' ‘ ' - . . - e g , .
,. r ' -v t -
‘ ( V - ..
. , . . e - - e ‘
. ’ ' ' - I Y ' ‘ ' ' r’ \ . - L
- ‘ t ‘ . A L. A- ' .
. . ,
L - - a . I O ‘*
,, . .. o . -

2
factors which correlate with a subject's attitudes.
Those studies which have been done in this particular
area, e.g., Adorno, gt a; (1950), Smith, Bruner, and
White (1956), have been of such large scale that
those interested in attempting research of a similar
nature have been discouraged. The present investigation
is an attempt to penetrate this relatively forgotten
area of social psychology. This study was carried out
in order to examine some personality correlates of
attitude consistency. The subject of attitude
consistency was approached in a unique manner, the measure
of this variable being the multiple scalogram analysis
error score. However, before elaborating on the
problem with which the present research is concerned,
the method of analysis used in this study and the
rationale for using this method will be discussed.

There are numerous ways of measuring a person's
attitudes. One such method is Guttman's scalogram
analysis. This method, however, has two weaknesses:
it lacks clear meaning for the concept "universe of
content" and provides no objective methods for
choosing items relevant to this universe; and it
provides no adequate concept in its theoretical
structure which would account for deviations from pure

subject-types. Because of these apparent weaknesses

3
in what seemed to be an effective method of scaling

attitudes, Lingoes (1960) devised a method for scaling
items according to the criterion set by Guttman but

with the above-mentioned weaknesses eliminated. This
method is known as multiple scalogram analysis (MBA).
Before describing the multiple scalogram method, however,
a brief discussion of the scalogram method of Guttman
will be given in order to acquaint the reader with

some of the concepts to be developed later.

Scalogram Analysis

The idea basic to Guttman's theoretical model
is that of oumulativeness. It is assumed that a set
of items with a common "universe of content," i.e.,

a single idea or trait underlying all statements which
can be made in reference to a particular subject, can
be arranged in such an order that a subject who
responds in a particular way to any item in the set
responds in the same manner to all items of lower rank
order. It is likewise assumed that subjects can be
arranged in an order according to the highest
threshold of items to which the subject responds in a

particular way.

Before carrying out a scalogram analysis the

investigator must first select a set of items which

4
represent some attitude, Opinion, or trait that he
is interested in measuring. These items should be
constructed in such a way that they vary with respect
to either level of difficulty or level or probable
endorsement, i.e., the items should possess the
property of cumulativeness so that the endorsing of
an item which is more difficult or more popular would
be likely to imply the endorsing of items which are
less difficult or less popular. Such a series is
represented, for example, by the items: 1) I am at
least 5 feet tall, 2) I am at least 5 feet 6 inches
tall, 3) I am at least 6 feet tall. Similarly, the
following attitudinal items might form a cumulative
series: 1) I would not mind dealing with a colored
person in a business transaction, 2) I would not mind
having a colored person as a member of my church or
social club, 3) I would not mind having a colored
person as a close friend. Very little inconsistency of
response would be expected with the first set of items
given above, since a person responding positively to
item 3 would necessarily respond in the same way to
items 1 and 2. 0n the other hand, the degree to which
the items in the second set are endorsed would be

expected to vary.

The set of items is administered to a number of

5
peOple. A matrix is formed by putting the subjects
in rank order according to their scores, i.e., the
number of positive or negative responses. Items are
likewise ranked from highest to lowest. Table 1
gives a hypothetical response matrix so ordered. The

score is the number of ls in the subject's response

pattern.
Table 1
Item Score
1 2 3
Subject 1 1 l l 3
2 l l 0 2
3 1 O 0 1
4 0 O O 0

If the three items appearing in this table were those
given above concerning height, we can see that subjects
are ordered according to height and items are ordered

according to degree of endorsement.

The next step in the method of scalogram
analysis is that of evaluating an ordered matrix for
scalability. Scalability can be approached from the
point of reproducibility. If a subject's response
pattern can be reproduced on the basis of his score

and a knowledge of item ordering alone, then the set

6
of items is said to be perfectly scalable. The
individual response patterns in a perfectly scalable
matrix are said to be pure subject-types. The matrix
appearing in Table l is perfectly scalable. Knowing
that the items are placed in rank order, a score of 2,
for example, would indicate a response pattern of

( l l O ) for subject 2.

Suppose that the matrix in Table 1 were not
perfectly scalable, i.e., subject 2 had instead the
response pattern ( 1 0 1 ) or ( O 1 1 ). Error would
be introduced if one were to attempt reproducing
subject 2's response pattern from a score of 2 and
assuming that item ordering was cumulative. Guttman's
theoretical structure contains no concept which
adequately takes such deviations into consideration.

It would seem obvious that such a model is unrealistic.
In multiple scalogram analysis an attempt is made to
develop a model which does not place such stringent
restrictions on departures from the pure subject-types,
yet one which produces scales which have the same
properties as Guttman scales. Since the present
research makes use of the multiple scalogram method

a description of this technique will be given.

Multiple Scalogram Analysis

"Multiple scalogram analysis is an objective

7
and empirical technique for partitioning a binary
response matrix into a number of submatrices, such
that each submatrix tends to be maximally homogeneous"
(Lingoes, 1960). Submatrices are made maximally
homogeneous according to the following three formal
criteria:

1) Items which have the largest number of
elements in common are brought together in order to
minimize the distance between adjacent sets of items.
Agreement scores, rather than conventional measures of
correlation, are used for relating adjacent sets of
items.

2) Each item is allowed to contribute only
its proportional share of the error of the entire
submatrix of items and subjects. £239; is defined in
terms of deviations from pure subject-types which would
entail perfect reproducibility of the submatrix. The
method of counting errors is discussed below.

3) Each subject is similarly allowed to
contribute only a certain percentage of error to the
submatrix.

Items and subjects are brought together in accordance
with these three criteria. The result is a matrix
of subject-item responses which is homogeneous and

one in which the number of errors is kept at a minimum.

8
Since multiple scalogram analysis is a

relatively new technique and used primarily at Michigan
State University and the university of Michigan, an
example of the analysis of the hypothetical data presented
in Table 2 (Lingoes, 1960) by this method may be in

order. In this table a "one" represents a positive
response to any of the eight items in,a test. A ”zero"

represents a negative response.

 

Multiple Scalogram Analysis 2; gypothetical Data

The first step in the analysis is to sum over
columns, writing the totals at the foot of the columns.
If any column sum is less than one-half of the number of
subjects (n), then that column is reflected, i.e., the
scoring direction of each item is changed by making
every 0 into a l and every 1 into a O. A new response
matrix is then constructed which includes the items
which have been reflected, as in Table 3. Columns
which have been reflected are marked with a "minus"
sign above the column number. The sums of the reflected
columns are, of course, also reflected by subtracting
the original sum from n. Items 2, 3, 7, and 8 are
shown reflected in Table 3. The revised column sums

for these items are 20, 15, 20, and 15, respectively.

a.—

i
.
. .
.
O
.
«L
, .1 .
.
v . .
. .
.
.. .
.
,
.
o t
w ..
. .
.
o
..
a
.1 .
.
I,
~
9 u
. .
.
. .
, .
.. O
.
.
.
e . y
,.,
r. .
n
. I.
4 ‘l
. . .1
.
.
A A
. .y ,
u r.

1
. O O.
Q
t
.e . u
. i _.
Q I "Ti.
. . O\
O. _ .
G l
.. —
.
-
’I

9
Table 2

Hypothetical Response Matrix

 

 

Item 1 2 3 4 5 6 7 8 Score Errors
8 1 l O 0 1 O l 0 1 4 4
2 l 1 1 O 1 l O 0 5 2
3 l 0 0 O l O 0 1 3 4
4 l 1 O O l 1 0 1 5 4
5 O O 1 1 l 1 O O 4 4
6 l O 1 1 O 1 O 0 4 2
700101100 3 4
8 1 O O l l 0 0 l 4 4
9 1 O 1 1 O 1 l O 5 4
10 1 O O l l 1 O O 4 4
11 0 O l l 0 l l O 4 4
12 1 1 O O 1 l O 0 4 4
13 l O 1 O 1 l O O 4 4
14 l O 0 0 l 1 O O 3 4
15 1 0 O 1 0 l l O 4 4
16 l O 0 1 0 1 0 O 3 4
17 1 O O l 1 l O 1 5 4
18 1 O O 1 O O O 1 3 4
19 1 O O O 1 1 O l 4 6
2000110100 3 4
21 1 O 1 1 1 l O O 5 2
22 1 0 0 l 0 0 1 1 4 4
23 O l 1 O 1 1 O O 4 4
24 1 0 O 1 O l 1 1 5 6
25 l l O O l O O l 4 4
Sum 20 5 10 15 15 20 5 10 - 98

The second step in the analysis is to calculate
the inter-columnar agreement scores and to form a matrix
of these scores ( of. Table 4). Working from the
reflected response matrix (Table 3), the agreement
score between any two columns is equal to the number of
identical subject-responses in those columns. One
agreement occurs between two items each time a subject
makes the same response, i.e., both positive (1) or

both negative (0), to the two items. Items which have

. .
.
..., 7-H A . a a. _. 2 .. fi- . . . -7 w A. . _ -
. ‘ ya .o u
.. ,. - -
I 4
I 4 u a
‘ r
'.
i 'u ‘
\ -. r- v
I — ‘ 1 - .-
l , _ , m
,. . _ ‘ ‘ _ P
. - v
I - K . - -- .
.
. , ~ I ' _
. .
V r- I C I'
1 . . U .‘ .-
. ‘l ‘ " I. ,- ,. l.‘ ,- -
‘. ‘a . ..
.. ,. .- 1
v r T
w 'l _ .
_ ,. .

H. 2 . ,
.' o . ,. .. u
n I ,
' I ’D - - V
‘ D
- \ t '
‘ J ’- .
I . ’- a
\ . g 4 .
- . ‘ -§. .‘ p
u . ,—
H r .
. . - 4 .
. . _. .
‘ l
- . V J-
I
f" . ,-
. ' ’
. ~
- 1 5 A .< r " .
' - r 'e
|
I ‘ v I
- . ‘ .A. ‘
l t. I g _. r. ,.. .
- :
v_ '
. . ~. .
' u- .. e - ‘
‘ ’ t ' 4
.
- . I

-- - .
. . -1 .. -_..-._ - .. 1 --. "a - _, - .1 . -«- - — — ---
~ . on‘ - u .- _ --
0- .
l ‘
‘ l '. v — p.
v , '4 A ' . ' ’,.. — ' c . v .
t y . ‘ . I l‘.
‘ w
. ‘ ~ - .- k ." ,
. ‘ . O
. -- - . A. ‘ _ .
‘.r . , , ‘ ,
' a
o -
. , , ~
‘ . . -
C - O I
o . _. .
I - ‘ .
r O - :u e. \
. . 4
' — I IV
- . .. . . -
v e
I' ‘ .1
I .
‘ O . - , - ‘v . .
' v
n A - | .
‘ ,
\ - - 1
. ' . _
- s o o O— Q ~
.' . . - -
. . . v 1

.qvv— ..

10
been reflected should be so indicated by means of a

"minus" sign in the row and column headings.

Table 3
Partially Reflected Matrix

 

 

Item 1 2 3 4 5 6 '7' 3
S 1 1 1 1 1 0 1 1 O
2 1 O O 0 1 1 1 1

3 1 1 1 0 1 O 1 O

4 1 O 1 0 1 1 1 0

5 O 1 0 1 1 1 1 1

6 1 1 0 1 O 1 1 1

7 0 1 O 0 1 1 1 1

8 1 1 1 1 1 O 1 O

9 1 1 0 1 0 1 0 1
10 1 1 1 1 1 1 1 1
11 0 1 0 1 O 1 O 1
12 1 O 1 0 1 1 1 1
13 1 1 0 0 1 1 1 1
14 1 1 1 0 1 1 1 1
15 1 1 1 1 0 1 0 1
15 1 1 1 1 O 1 1 1
17 1 1 1 1 1 1 1 O
18 1 1 1 1 0 O 1 0
19 1 1 1 O 1 1 1 O
20 0 1 O 1 0 1 1 1
21 1 1 0 1 1 1 1 1
22 1 1 1 1 0 0 0 0
23 0 0 0 0 1 1 1 1
24 1 1 1 1 0 1 O O
25 1 0 1 O 1 O 1 0
'Sum- ‘20 20'15'15'15'20 20'15

The third step consists of finding the initial
item, ;, whi h.is to appear in a dimension. This is
done by selecting the item which has the largest column
sum from the reflected matrix of Table 3. The logic for
this step is the fact that the item set which contains

the largest number of positive elements (1s) after all

 

 

 

 

. . - , 7
, ‘ . - ‘
. ,_ . .
, .
.
. t . _. . ‘
-,._ . _ ‘ . -.
V
n n ,7 . - . - i -. ,.
. . . - '
.
,— - ,A .—
- ' u '
I
_ ,. . - ..
I . ,- ,
, - .. . 2
I
, , . . .
,
. In ,. . ,-
.
a A .. ._
_ - .. .. .- ,'
. .- '- '
,e . . . . .4
I
.
. e
- .. w . - .-
‘ , .
. a .7 - . .7
,.
, 4 .- .- ,. a-
. , . . - o n o
1 ‘ ‘
.
.. .4 — -, '
.
_ . - i i '.
. - - .. , , .. ,
.
. - - 1
.
, _. . .. . . -
. -— . n '
.
.
. .
2 _ , I .- . . A -
- .— ~ v "‘ - . i
.
I
I- ' o w - , . , u U ' l '
I , s . I .
. .
. -_ . '
. , ‘ r ., ' c t
. ‘ _ , , . _. ' ,-
, ‘ ' e I . ‘ .
, , » ‘ , , . .‘
e ' -
. . ‘ , n
. _ . , i
. t e ; ‘ I

 

11
item sets with pa less than .5 have been reflected is
the one which is most likely to include other sets as
preper subsets. In the case of ties among column sums,
item i,is selected arbitrarily. The effect of making
an arbitrary decision in this matter is merely that a
different ordering of both items and dimensions is
obtained. In the present example, item 1 was chosen
from among items 1, 2, 6, and 7 with tied column sums
of 20. One keeps track of the variables used by placing

a check mark over the item numbers in Table 2.

Table 4
Item.Agreement Matrix
Item'1'2'345678
' l x 17 20 l4 14 15 17 10
-2 17 x 14 2O 10 17 15 14
-3 20 14 x 13 13 10 14 5
4 14 2O 13 x 5 13 10 13
5 14 10 13 5 x 14 2O 13
6 15 17 10 13 14 x 17 20

-7 17 15 14 10 2O 17 x 14
-8 10 14 5 13 13 2O 14 x

 

 

The fourth step in the analysis consists of
selecting the next item, 1, to be linked with item A.
This is accomplished by finding the item which has
either the highest agreement score or the highest
disagreement score with item 1. Item ;_and item 4 are
then placed in a new matrix (Table 5), reflecting or
re-reflecting item 1, if the largest value in the

12
lth column is the disagreement score. A "plus" sign
is placed over the item number in Table 5 if a
re-reflection has occured. After item thas been
linked with item ;, the next item 1 is linked with the
previous item 1. If ties are encountered, then one
should select the item to pair with the previous one
which introduces the fewest errors, i.e., the 01

response pattern.

In our example, item 3 has the highest agreement
score with item 1. Therefore, item 1 and item 3 form

the first two columns of Table 5.

The final step is to determine the amount of
error introduced by bringing in item 1. Each
displacement of a O or a l, i.e., a 01 response
pattern, is counted as two errors. If the amount of
error, as determined by [ §°3m _ 21:19:21) where :1
equals the number of subjects and m is the number of
items in the dimension, exceeds some predetermined
value, e.g.. .10n or .15n, then a new dimension must
be begun and the initial item of this dimension found.
The latter is done by selecting the item with the
largest column among those items in Table 3 which have
not been used previously in any dimension. On the

other hand, if the amount of error is less than the

above criterion, then one must find what item has

 

e
u
.
I. .
u
.
C
e
c A
O. o. . .
L _ ,
.1
1’
_
-
t.
. ,
. I
I.
.
O
D
I
O

. Q .
e. . J
r
.
I A
.1
r» _ .
4‘
l
4' . r.
L:
.
y
. .
I
,
.
e
.
.1
.
r

13
the highest agreement score or disagreement score with
the last item included in the dimension. This new item
represents item 1, and the previous item 1, becomes item

1. The fifth step is then repeated as before.

In the present example, the links between items
1 and 3, 3 and 8, and 8 and 6 meet the above criterion.
The links between item 6 and either item 4 or item 7
exceed the criterion and therefore a new item ;_is
selected from those items in Table 3 not previously used
as the initial item of a second dimension. Beginning
with item 7, the links between items 7 and 5, 5 and 4,
and 4 and 2 result from the repetition of the third,
fourth, and fifth steps. This set of items, along with
items 1, 3, 8, and 6, produce the data appearing in
Tables 5 and 6.

Inspection of Tables 5 and 6 indicate that
any two subjects with equal scores have identical
response patterns. For example, subjects 4 and 19 both
received a score of 3 and both have answered the four
items in the same way, i.e., 1110. This example also
illustrates the concept of perfect reproducibility.
Knowing that items are placed in rank order and the
scores for each subject, all responses can be predicted
with 100 per cent accuracy. Fbr example, subject l4s

scores are 2 and 3 on the two sets of items, yielding

 

.e

(1

.‘

14
a predicted response pattern for the 8 ordered items

of: 11001110.

It should be pointed out that in actual practice
perfect reproducibility is quite rare. In addition, as
the error criterion is lowered, more error appears in
each dimension and the reproducibility is consequently
decreased. A lower criterion, however, does allow more
accurate assessment of subjects with respect to error.
Resolution of this dilemma is only a matter of
experience and purpose. In the present investigation an
error criterion of .15n was used. The criterion was not
lowered further since the resulting dimensions would

then tend to be meaningless.

6 Score

15

Table 5
Dimension 3
1 3 8

 

Item

31-43010412021222343414034.

0010000100000000010001001
1011000100000000111001011
1011000101010111111001011
1111010111011111111011011
1123456789 R
S

 

16

Table 6
Dimension II

7 5 4 2 Same

Item

 

1.434213202043301213120404

0101000000010000000000101
0111001000011100001000101
01111011010111.00101010101

1111111101011101111110101

123456789012 I
I11
.5

 

17
Scalogram and Multiple Scalogram Analysis: 15 Comparison

It would be well to point out some similarities
and differences between Guttman's scalogram method of
analysis and multiple scalogram analysis. First, while
MSA results in scales which have the same property as
Guttman's scales, i.e., cumulativeness, there is no
real need for the concept "universe of content" in the
MSA model. The investigator lets the responses of the
subjects themselves group the items instead of
subjectively choosing items which best fit the model.
While it is not recommended that the scientist using
MSA abandon all criteria for item selection, it is
pointed out that he is not bound to selecting items
which have the greatest probability of fitting the
model. MSA would seem to be the more objective of

the two approaches.

A second difference is that of approach to
the data. Scalogram analysis takes a set of items and
tests the homogeneity of these items. MBA, on the
other hand, both finds and tests the homogeneity of
a group of items. MSA would appear to be the more

efficient of the two methods in handling the same data.

A third difference between scalogram analysis
and the multiple scalogram method is that the latter

l7
Scalogram and Multiple Scalogram Analysis: A’Comparison

It would be well to point out some similarities
and differences between Guttman's scalogram method of
analysis and multiple scalogram analysis. First, while
MBA results in scales which have the same pr0perty as
Guttman's scales, i.e., cumulativeness, there is no
real need for the concept "universe of content” in the
MSA model. The investigator lets the responses of the
subjects themselves group the items instead of
subjectively choosing items which best fit the model.
While it is not recommended that the scientist using
MSA abandon all criteria for item selection, it is
pointed out that he is not bound to selecting items
which have the greatest probability of fitting the
model. MSA would seem to be the more objective of

the two approaches.

A second difference is that of approach to
the data. Scalogram analysis takes a set of items and
tests the homogeneity of these items. NBA, on the
other hand, both finds and tests the homogeneity of
a group of items. MSA would appear to be the more

efficient of the two methods in handling the same data.

A third difference between scalogram analysis
and the multiple scalogram method is that the latter

18
is presently limited to handling dichotomous data
while Guttman's method can be used on items with any

number of categories of response.

Since multiple scalogram analysis is a
relatively new technique and used primarily at Michigan
State University and at the University of Michigan, a
few illustrative studies making use of this method will
be described.

Previous Research

The use of MBA has been primarily in connection
with voting behavior. Lingoes (1960) analysed the
voting responses of the 83rd United States Senate on
256 issues. The MSAs yielded 15 dimensions, accounting
for 105 of the 128 Sample A issues, and twelve dimensions,
accounting for 106 of the Sample B issues. Dimension
sizes varied between three and 25 items. From the
results of the analysis it was possible to select eight
key senators whose voting patterns could be used as a
basis for predicting the voting patterns of the other
senators. The purpose of this study was primarily to
illustrate the properties of a new procedure of
analysis and to compare the results of the analysis

with other standardized procedures.

19

An MSA of the responses of 100 solders who had
been in combat on nine fear symptoms (Stouffer, §p_§l,
1950) yielded two dimensions of seven items dealing with
gastro-intestinal symptoms and two items related to
musculo-skeletal symptoms (Lingoes, 1960). The purpose
of this study was again to illustrate the properties
of the multiple scalogram method.

An analysis of 22 items relating to driver, car,
and environmental characteristics occuring in 955 fatal
car accidents resulted in scales ranging from two items
to six items. Item content of the scales corresponded

with the above-three characteristics (Allen, 1962).

Problem

The problem with which the present research is
concerned resulted from a study of Lingoes (1960) in
which a multiple scalogram analysis was carried out on
the responses of 100 subjects to the 30 items on the
K scale of the Ndnnesota NMltiphasic Personality
Inventory. The analysis yielded only one dimension of
five items for the sample of subjects. Similar results
were obtained using the scalogram method of Guttman.

No difference was found to exist between subjects who
were patients in a mental hospital and those who were

not patients in terms of their total scores on the K

20
scale. However, there was a significant difference
in the number of errors contributed by the two groups
and subjects could be reliably separated on the basis

of error scores.

The question has arisen whether further
differentiation between subjects with respect to
personality variables is possible on the basis of MSA
error scores. Assuming that the MBA error score is
a measure of response consistency, a low error score
indicating a more consistent response pattern than a
high error score, one would expect that groups of
normal subjects could be differentiated with respect
to these scores. This line of thought follows that
of Berg (1957), i.e., deviant response patterns in
one area are associated with deviant responses in

other areas.

The present investigation sought to determine
whether differentiation between normal subjects with
respect to personality variables as measured by the
Edwards Personal Preference Schedule is possible on
the basis of multiple scalogram analysis error scores.
A discussion of the method used in attacking this

problem is given in the next chapter.

II. METHOD

The design of the present research called
for the administration of an attitude scale and a
personality test to a number of subjects. The attitude
scale was analysed by the multiple scalogram method and
scored in terms of error scores. Error scores were
correlated with scale scores on the personality test.
In this chapter the subjects, tests, and procedure used

in the present investigation are described.

Subjects

The subjects for this investigation were 126
students at Michigan State University who were enrolled
in introductory psychology courses during the 1962
Winter term. The subjects were approximately equally
distributed with respect to sex. Although the subjects
were not differentiated with respect to class level,
the majority of the subjects were assumed to be

sophomores or juniors.

Tests

Two tests were administered to all subjects.

The first of these was presented as a public opinion

21

22
questionnaire dealing with the foreign policy of the
United States.1 It consisted of 44 items from the
S3312 p£.Attitudes toward war and Peace (Droba, 1931)
and 16 items selected at random from the Public Opinion
Questionnaire about war and Foreign Policy (Wrigley,
1961). Certain items on the Scale of Attitudes were
modified in keeping with the issues to which they
referred. Preliminary research indicated that an
attitude scale of the equal appearing interval type
produced more meaningful dimensions than other types
of attitude scales when analysed by the multiple
scalogram method. Attitudes dealing with foreign
policy was chosen for measurement since it is a subject
about which peOple have more or less crystallized views
and yet one in which there is a substantial division

of Opinion.

The second instrument was the Edwards Personal
Preference Schedule (Edwards, 1954). This particular
personality test was felt to suit best the purpose and
sOOpe of this research. The Edwards Personal Preference
Schedule (EPPS) not only can be administered quickly
to a large number of subjects, but it also provides
measures of fifteen relatively independent personality
variables based on Murray's list of manifest needs

(1938). In addition, in the EPPS an attempt is made

1See Appendix B below.

23
to minimize the influence of social desirability in
test responsiveness. This is accomplished by the
forced-choice technique in which items of equal weight
with respect to social desirability are paired. The
names of the variables which the EPPS measures are:
(1) achievement, (2)? deference, (3) order,
(4) exhibition, (5) autonomy, (6) affiliation,
(7) intraception, (8) succorance, (9) dominance
(10) abasement, (ll). nurturance, (12) change,
(13) endurance, (l4) heterosexuality, (15) aggression.

The EPPS also provides means of measuring
response consistency. By comparing the number of
identical choices made on two sets of the same 15 items,
it is possible to determine whether the subject is
responding to the items by chance alone or in
accordance with his personality structure. In the
present research a consistency score of eleven or more
identical choices out of the possible 15 was taken as
the criterion for keeping or discarding a subject's
tests, the probability of eleven or more identical
choices occuring on the basis of chance alone being

approximately .06.

The public Opinion questionnaire and the
Edwards Personal Preference Schedule were administered

in a group testing situation. There were three groups

A r
r . .
e
. . a
e :’ ‘ .
e a
, ‘ ,. .
i
|
. . 3 ‘7
. u
3' .‘ 1

. ‘
" - v e Y
A , Q

. e
. , .
~ -—~ 1

\ v

o a 1
‘ l
I \
u
p ' _. ‘
1
q . . ~
- f. , , .
n
, . . . . .‘_
x.
. I

- . v 3‘

~ a

1'

24
of approximately 45 subjects each. Before the tests
were given, the following statement was read to the

subjects:

"A group Of us in Psychology have been working
during the past months on methods Of determining
the response patterns that occur when peOple take
certain kinds Of psychological tests. But, in
order to perfect these methods, we have to try
them out on actual data -- and that is where you
come in. Each of you will be given two Of these
tests. One is a survey Of public Opinion about
the United States' foreign policy. The second is
a standardized personality test. There are
instructions on each of these which you shouldn't
have any trouble understanding. However, if you
do have any questions, feel free to ask them.

"There are three things I would like to ask
Of you. First, we would like tO keep your
responses as anonymous as possible so please do
not put your name or student number on these
tests. Second, you are asked tO work as quickly
as possible and £93 to omit any items. The tests
should take between an hour and an hour and one-
half to complete. Third, after completing both
tests, place the IBM answer sheet inside the
Opinion survey and hand them in. Thank you."

Procedure

The responses Of the subjects on the public
Opinion questionnaire were analysed by the multiple
scalogram method as previously described. The
Michigan State University computer laboratory library
program K9M was used for this purpose. The error score
for each subject was calculated from the result Of

this analysis. Product moment correlations between

 

 

 

 

"

 

25
error scores and scores on the EPPS scales were computed
for all subjects. Michigan State University computer
laboratory library program K5M was used.

III. RESULTS

multiple Scalogram Analysis

Using a criterion Of .15n for the amount Of
error allowable in any one dimension, the multiple
scalogram analysis yielded ten dimensions which
accounted for 46 of the 60 public Opinion items.

Dimension sizes varied between three and 16 items.

Error scores for the 126 subjects on all ten
dimensions varied between zero and 16. Table 7 gives
the frequency distribution Of the error scores. The
distribution was skewed to the right, the modal error

score being 6, and the mean error score 6.61.

Product Moment Correlation

Product moment correlations between error
scores and EPPS scale scores are given in Table 8.
Correlations significant at the .05 level are indicated
by an asterisk. Significant positive correlations were
found to exist between error scores and scores on EPPS
scales Succorance and Aggression. EPPS scales Change
and Intraception were found to be significantly

26

27
negatively correlated with the MSA error score.
Differences with respect to sex are discussed in

Appendix A.

28
Table 7

Frequency Distribution Of MSA Error Scores

 

 

 

 

 

 

 

 

O 2 ‘# £5 8 11) 12 It 16

29
Table 8

Correlations between Error and EPPS Scale Scores

 

Scale Correlation
l. ach .151
2. def -.l47
3. ord -.093
4. exh -.058
5. aut .001
6. eff .170
7. int -.186 e
8. suc .225 *
9. dom -.054

10. she .008
ll. nur .073
12. chg -.257 *
13. end -.112
14. het .036
15. agg .213 *

 

Correlations significant at the .05 level
are indicated by an asterisk.

I
. .
- - --_... __ l ,,
.. ,r.
.
I
.
,.
_ .
..
. .
. .
.
I
.
U.
.
I
’w
I l
t. v' '
.
. ,_
.....
0 g
' . t 1.. 5 x

,7 _e
I
. - '—

 

IV. DISCUSSION

The results Of the present investigation
suggest that differentiation between normal subjects
with respect to personality variables is possible on
the basis of multiple scalogram analysis error scores.
In this chapter a brief discussion Of the results Of
this study is presented and suggestions made for

future research.

The aim of the present research was to
determine whether normal subjects could be
differentiated with respect to personality variables
on the basis Of multiple scalogram analysis error
scores. It was found that in a specific situation a
differentiation between subjects could be made. The
results of this investigation verify those Of Lingoes
(1960), i.e., error scores seem to be indicative of
irratic personality. Lingoes suggests that such
response inconsistency is perhaps due tO a difference
in the meaning or significance given to certain test
items by high error producing subjects. This
eXplanation would seem to be particularly true in the

case Of attitudinal items, since the items appearing

30

31
on the various dimensions resulting from a multiple
scalogram analysis are not items which would tend to
be ambigious, i.e., items in the middle Of the
attitude scale, but items with seemingly clear-cut
meanings. The results of this study also suggest
that response inconsistency may be due to an
unsuccessful attempt on the part of the subject tO
compensate for some undesirable element present in
his personality structure. The rationale for this

assumption is discussed below.

It should be emphasized at this point that
the type Of consistency to which is referred in this
study is not the same as that tO which is referred in
present-day cognitive theory, e.g., Festinger (1957).
Whereas the latter type Of consistency usually refers
to a more or less internal process which occurs,
consistency, used in the present context, deals only
with an aspect Of the relationship of a single
individual with respect to others in his group, i.e.,

a more or less external relationship.

In an attempt to account for the results Of
the present investigation it was speculated that
subjects with high error scores are typified by strong
dependency needs and subjects with low error scores

have low dependency needs. The following rationale

32
for this assumption is given: A number of factors
characterize the person who is overly dependent. 0n
the one hand, he has an exaggerated desire for
sympathy, encouragement, and social approval usually
stemming from a situation in which these needs were
satisfied by over-protecting parents. At the same
time, he does not want to eXperience new situations
which would tend to place him in positions where he
would feel helpless and insecure. He also has very
little sense of responsibility and concern for other
people. On the other hand, the overly dependent
person is frustrated because of his dependent nature.
This gives rise to a vigorous desire for self-
assertion and other aggressive behavior (Cameron and
Magaret, 1951, Baldwin, 1955). A person with such a
need pattern would tend to score high on EPPS scales
Succorance and Aggression and low on the Change and
Intraception scales. The reader is directed to Table
9 below for a list Of the manifest needs associated

with the four EPPS scales mentioned above (Edwards,
1954).

The present investigation found that persons
with such a need pattern tended to have high multiple
scalogram analysis error scores, i.e., tended to be

inconsistent with reapect to their pattern of

33
attitudes. Conversely, persons scoring low on EPPS
scales Succorance and Aggression and high on the
Change and Intraception scales tended to be more
consistent with respect to their attitude patterns,

i.e., tended to have low error scores.

The question is then raised why the overly
dependent person is more apt to be inconsistent when
it comes to eXpressing Opinions regarding some social
issue. The following eXplanation is suggested: The
overly dependent person does not want to appear so to
other peOple. Therefore, in situations in which he
feels such a disclosure might be made, he may attempt
to conceal his nature by compensating in some way,
i.e., by an exaggerated trend toward independent self-
sufficiency. In the present case, when the overly
dependent person was asked to eXpress his Opinions
regarding war and foreign policy, he responded in the
way he thought a person far less dependent would
respond. Unfortunately, however, he was "caught" by
virtue of a high error score which reflected his

inconsistent response pattern.

Before assuming any relationship between error
scores and personality variables to be universal,

further research is imperative. Such research would

34
necessarily involve a replication of the present
study. In addition, a more comprehensive test, such
as the MMPI, should prove valuable in finding other
variables which may be related to response consistency.
It is also suggested that an investigation be made
to determine whether similar results are obtained
using different types of attitude scales. The equal
appearing interval type of attitude scale was used in
the present study because previous research indicated
that this type of scale seemed to lend itself more
readily to analysis by the multiple scalogram method,
i.e., it produced more meaningful dimensions as well as
a better distribution of error scores. It is finally
suggested that an investigation be made dealing with
the effect of varying the error criterion. Although
the relationship seems to be similar, there is
evidence that error scores resulting from multiple
scalogram analyses under different criterions for error
correlate with somewhat different personality variables.
A discussion of this subject was felt to be beyond the
scope of this paper, but such research is essential
for a complete evaluation of the multiple scalogram
analysis error score as a tool for differentiating

subjects.

35
Table 9

Manifest Needs Associated with EPPS variables

Aggression

tO
to
to
to
to
to
to
to
to

Change

to
to
to
to
to
to
to
to

to

attack contrary points of view

tell others what one thinks about them
criticize others publicly

make fun Of others

tell others Off when disagreeing with them
get revenge for insults

become angry

blame others when things go wrong

read newspaper accounts of violence

do new and different things

travel

meet new peOple

eXperience novelty and change in daily routine
experiment and try new things

eat in new and different places

try new and different jobs

move about the country and live in different
places

participate in new fads and fashions

Intraception

to
to
to
to
to

to

to
to

analyse one's motives and feelings
Observe others

understand how others feel about problems
put one's self in another's place

judge peOple by why they do things rather
than what they do

analyse the behavior Of others

analyse the motives of others

predict how others will act

36
Table 9--Continued

 

Succorance

to
to
to
to

to
to
to
to
to

have others provide help when in trouble

seek encouragement from others

have others be kindly

have others be sympathetic and understanding
about personal problems

receive a great deal Of affection from others
have others do favors cheerfully

be helped by others when depressed

have others feel sorry when one is sick

have a fuss made over one when hurt

 

37
BIBLIOGRAPHY

Adorno, T. W., Frenkel-Brunswik, E., Levinson, D. J.,
and Sanford, R. N. The Authoritarian Person-
ality. New York: Harper, 1950.

Allen, T. M. Unpublished Research, Department Of
Psychology, Michigan State University, 1961.

Baldwin, A. L. Behavior and Development_in Childhood.
New York: The Dryden Press, 1955.

Berg, I. A. "Deviant Responses and Deviant PeOple:
the Formulation of the Deviation Hypothesis,"
JQurnal Of Counseling_§sychology, 1957. 4,
154-161.

Cameron, N., and Magaret, A. Behavior Pathology.
Boston: Houghton Mifflin Company, 1951.

Droba, D. D. "A Scale of Militarism-Pacifism,"
Jgurnal of Educational Psychology, 1931, 22,
9 "J-lle

Edwards, A. L. Edwards Personal Preference Schedule.
New York: The Psychological Corporation, 1954.

Festinger, L. A Theory of Cognitive Dissonance.
Evanston, Illinois: Row, Peterson, 1957.

Lingoes, J. C. "Multiple Scalogram Analysis: a
Generalization Of Guttman's Scale Analysis."
Unpublished Ph.D. dissertation, Nuchigan
State University, 1960.

Murray, H. A., 23 El. Explorations in Personality.
New York: Oxford University Press, 1938.

smith, W. B. , Bruner, J. S. ’ and. White, R. W. 921n-
ions and Personality. New York: John Wiley
and Sons, Inc., 1956.

Stouffer, S. A., g3 g1. Measurement and Prediction.
Princeton, New Jersey: Princeton University
Press, 1950.

Wrigley, C. F. "Public Opinion Questionnaire about
War and Foreign Policy," Unpublished Re-
search, Department of Psychology, Michigan
State University, 1961.

39
Table 10

Correlations between Error and EPPS Scale Scores

 

Scale Males Females
1. ach " "041 .361 *
2. def -.161 -.118
3. ord -.187 .079
h. exh -.023 -.141
5. aut -.068 .114
6. aff .237 ~.O39
7. int -.323 -.078
8. suc .332 .017
9. dom -.l27 .263

10. aba -.256 -.O38
ll. nur .083 -.206
12. chg -.047 -.383
13. end .007 .077
14. het .221 .020
15. ass .282 * .306 *

 

Correlations significant at the .05 level
are indicated by an asterisk.

Appendix B
‘Bshlis,9pinion.fiunxey pp War and Foreign Policy

Below-are 60 statements which represent widely—held
Opinions about war and foreign policy. These statements have
been selected from speeches, newspapers, books and other
sources. They have been chosen in such a way as to represent
a variety of viewpoints. As a result, some people are likely
to agree with some of the statements, and other peOple with
other statements. ?“\

After reading each statement, you are requested to
record your personal opinion regarding it. Do so by placing
a check mark before every statement with which you agree.
Interprets the statements in accordance with your own
eXperience.

Please answer frankly. This is not a test, and there
are no "right" or "wrong" answers. We are only interested in
finding out how people feel about war and foreign policy, and
our interest is in the percentages who agree and disagree with
each statement. Your responses will be treated as completely
confidential.

1. War in the modern world is as needless as it is suicidal.

2. Many peOple benefit by learning the lesson of war-time
discipline.

3. Only those military units which afford training to the
body and mind should be retained.

4. Army discipline injures self-respect and individuality.

5. The losses of human life and property, great as they are,
are small evils compared to the undermining of morals
and the lowering of standards Of culture and civilization
caused by war.

6. Wars are justifiable only when waged in defense of weaker
nations.

7. war is ennobling and stimulative of the highest and best
qualities Of man.

8. It is the moral duty of the individual to refuse to
participate in any way in any war, no matter what the
cause.

9. Might makes right.

10.

11.

12.
13.

14.

15.

16.

17.

18.

19.

20.
21.

22.

23.

24.

25.

26.

27.
28.

29.

The most we can hOpe to accomplish is the partial
elimination of war.

Economic aid to foreign countries is more useful in
furthering the cause of democracy than is military aid.

The evils Of war are slightly better than its benefits.

The image of the United States abroad has suffered since
World war II.

Until an equivalent discipline is organized, war must have
its own way.

Because right may be more important than peace, war may
be the lesser of two evils.

Along with patriotism, world citizenship should be
taught in all secondary schools.

Persons should pledge themselves never to aid in any
future war.

It is almost impossible to have a large military force
without being tempted to use it.

The Russian peOple basically want to live happily and
in peace.

war is the tonic of the races.

There is little chance of our being involved in a
nuclear war.

When the next war breaks out we should tell the diplomats
who lead us into it that we will not follow them.

Militarism is necessary for the prOper defense and
protection of the individuals of a country.

SO long as any peOple, white, black, brown or yellow,
hold weapons in their hands, we must not commit the folly
of disarming.

The best way of preventing war involving the United States
is to keep our nation militarily stronger than our enemies.

Under the scourge Of war a nation has no Opportunity for
cultural develOpment.

The soldier suffers tremendously and gains very little.
The evils that war brings far outweigh any possible benefits.

There is no progress without war.

30.

31.

32.

33.

34.

35.

36.

37. '

38.

39.

40.

41.

42.

43.

44.
45.

46.

47.
48.

We should have a moderate amount of military training
in our schools.

No scheme of aggression or conquest can be pursued for
any length of time without enfeebling victor as well as
vanquished.

When war threatens we should refuse the call to service
and increase our anti-war activity.

We shall never get rid of war because humans are incurably
aggressive.

Universal disarmament is the only way we can be sure of
eliminating war.

Militarism should be abolished from the curriculum of the
schools.

It is not in war but in peace and prosperity that our
worst vices develOp and grow.

We cannot hOpe to do away with war, because it is part of
the unending struggle for survival in a crowded world.

If there is a Third World war, the United.8tates foreign
policy will have been largely to blame.

If armed conflict between individuals can be :Ltlawed,
it is possible to outlaw armed conflict between nations.

Every war shows cowardice, murder, arson, graft, and
leaves a trail of personal and national demoralization.

The United States can be depended upon never to start a
war.

The most frequent cause Of war is the rivalry Of nations
for possession of territory, markets, and spheres Of
influence.

There is no conceivable justification for war.

Military training is imperative, but it should be voluntary.

We should not try to make any disarmament agreement with
the Russians because they are not tO be trusted.

Nations should agree not to intervene with military force
in purely commercial or financial disputes.

The United Nations is a waste of time and money.

Peace and war are both essential to progress.

49.

so.
51.
'52.
53.

54-.

55.

56.

57.

58.
59.

60.

The abolition of war would mean effeminacy, softness,
and degeneracy.

wars should be fought in order to free Oppressed nations.
The American people have become afraid of war.
Communist China should be admitted to the United Nations.

Election of a Democratic president has reduced our chances
of staying at peace.

A host Of young men entered the war in a spirit of idealism
and unselfish devotion to a great cause, only to return
disillusioned and cynical as to the value of ideals.

Compulsory military training should be established in
all countries.

Pugnacity, rivalry and self-interest are natural, but
need not result in war any more than human desire for
dominance need result in slavery.

We should stand behind the present administration, no
matter what decisions are made or how we feel toward
those decisions.

It would be better to have Communism take over the world
than to have a nuclear war.

If we are to survive, every American family must have
a bomb shelter and know how to live in it.

The United States is the most powerful nation on earth.

MICHIGAN STATE UNIVERSITY LIBRARIES

3 1293 03070 9996