The Influence Of The Control Of Personal Set Upon Prediction By Factored Tests Of Temperament And Interest

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content T his d isserta tio n has been 64— 2601
m icrofilm ed exactly as r eceiv ed
PLITTM AN, Jack C ., 1921-
THE INFLUENCE OF THE CONTROL OF
PERSONAL SET UPON PREDICTION BY
FACTORED TESTS OF TEMPERAMENT
AND INTEREST.
U n iversity of Southern C alifornia, P h.D ., 1963
P sychology, general
University Microfilms, Inc., Ann Arbor, Michigan
THE INFLUENCE OF THE CONTROL OF PERSONAL SET
UPON PREDICTION BY FACTORED TESTS OF
TEMPERAMENT AND INTEREST
by •
Jack"Plittman
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(Psychology)
June 1963
UNIVERSITY O F S O U T H E R N CALIFORNIA
GRADUATE SC HOOL
UNIVERSITY PARK
LOS A N G ELES 7. C ALIFORNIA
This dissertation, written by
JACK C. PLITTMAN
under the direction of h%P.....Dissertation Com
mittee, and approved by all its members, has
been presented to and accepted by the Graduate
School, in partial fulfillment of requirements
for the degree of
D O C T O R O F P H I L O S O P H Y
Dean
D ate June . . .1963
ACKNOWLEDGMENT
I express my gratitude and appreciation to
Dr. Floyd L. Ruch for encouragement of the broadening of
my specialization in psychology and for his enthusiasm and
support during the course of my graduate study and this
dissertation.
X am also indebted to the students and their in
structors from the schools of nursing mentioned below, who
served as subjects in this study. I especially thank the
following individuals: Mrs. Rebecca C. Bosworth, Chairman,
Nursing Department, Los Angeles City College; Mrs. Glennis
Burke, Chairman, Department of Nursing, Riverside City
College; Mrs. Gladys Golz, Head, Department of Professional
Nursing, Mount San Antonio College; Mrs. Elizabeth Jacobsen,
Chairman, Department of Nursing, San Bernardino Valley
College; Mrs. Phyllis Linden, Chairman, Nursing Department,
East Los Angeles Junior College and Mrs. Margaret Slotkin,
Director of Nursing, Chaffey College.
TABLE OF CONTENTS
Page
LIST OF TABLES....................................... iv
Chapter
I. INTRODUCTION AND HYPOTHESES ................ 1
Introduction
Hypotheses
II. BACKGROUND OF THEORY AND RESEARCH......... 8
Problems in Personality Measurement
Studies of Nursing Personnel with
the Guilford Inventories
III. METHOD ..... ............................. 24
Subjects
Procedure
Design
Measures
The tests
Special test scores
The criteria
Statistical Procedures
IV. RESULTS............. 46
V. DISCUSSION.................................... 95
VI. SUMMARY...................................... 108
REFERENCES............................................ 112
APPENDICES..................... 116
iii
LIST OF TABLES
Table Page
1. Age for Subject Groups in Y e a r s ............. 27
2. Comparison of Means of GZTS Factors under
Experimental and Control Conditions for
Total Groups................................. 47
3. GZTS Means and Standard Deviations for
Control Condition for Pass, Fail and
Dropout Groups.......... 50
4. GZTS Means and Standard Deviations for
Experimental Condition for Pass, Fail
and Dropout Groups.......................... 51
5. DFOS Means and Standard Deviations for
Control Condition for Pass, Fail and
Dropout Groups .............................. 5 3
6. DFOS Means and Standard Deviations for
Experimental Condition for Pass, Fail
and Dropout Groups.......................... 54
7. Comparison of Validity Coefficients for
GZTS Factors under Experimental and
Control Conditions for Fail Criterion,
Using Fisher's Transformation ........... 55
8. Comparison of Validity Coefficients for
GZTS Factors under Experimental and
Control Conditions for Dropout Criterion,
Using Fisher's z_ Transformation.......... 57
9. Comparison of Multiple Correlations, under
Experimental and Control Conditions, for
the 10 GZTS Factors with the Fail and
Dropout Criterion Groups .................... 59
10. Comparison of Validity Coefficients for
DFOS Factors under Experimental and
Control Conditions for Fail Criterion,
Using Fisher's j s Transformation........... 61
iv
j
I Table Page
11. Comparison of Validity Coefficients for
DFOS Factors under Experimental and
Control Conditions for Dropout Criterion,
Using Fisher's £ Transformation ............. 62
12. Comparison of Multiple Correlations, under
Experimental and Control Conditions, for
the 10 DFOS Factors with the Fail and
Dropout Criterion Groups ...................... 63
13. GZTS Means and Standard Deviations
for Instructors............................... 66
14. DFOS Means and Standard Deviations
for Instructors............................... 66
15. Significance of the Difference Between
Means of Instructors and Pass Groups;
and Instructors and Fail Groups for the
Control Condition on the GZTS Factors . . . 67
16. Significance of the Difference Between
Means of Instructors and Pass Groups;
and Instructors and Fail Groups for the
Experimental Condition on the GZTS Factors . 68
17. Profiles of Mean T-Scores, Based on GZTS
Norms (1949), for Instructors and
Criterion Groups ............................ 69
18. Comparison of Profiles of T-Scores for
the 10 GZTS Factors Between Instructors
Group and Each of the Criterion Groups . . . 73
19. GZTS Popular Score Means and Standard
Deviations............................. 75
20. DFOS Popular Score Means and Standard
Deviations............................. 76
21. GZTS Pppular Score Validities (r bis) .... 78
22. DFOS Popular Score Validities (r bis) .... 79
23. Summary of Beaver R.N. Key Statistics for
Condition and Criterion Groups and
t tests of Differences Between Means .... 81
v
Table
Page
24. Falsification Scale Comparisons Between
Control and Experimental Conditions and
Standardization Norm Group for Gross
Falsification (GF), Subtle Falsifi
cation (SF), and Carelessness-Deviancy (CD). 83
25. Items per Factor, Answered by Significantly
Different Proportions of the Control and
Experimental Groups, by Criterion Groups;
Pass, Non-Pass and Total and by any of the
Groups....................................... 87
26. Items per Factor, Answered by Significantly
Different Proportions of the Control and
Experimental Groups and Answered by over
50 Per Cent of the Experimental Group in
the Non-Keyed Direction ................... 90
27. Items of the Beaver R.N. Key Found to
Differentiate Pass from Non-Pass Groups
under Control Condition ................... 93
vi
CHAPTER I
INTRODUCTION AND HYPOTHESES
Introduction
Personality and interest questionnaires have been
criticized regarding the fact that the subject may inten
tionally alter his responses in order to create a good
impression. Tliere has been some hesitancy to use these
tests in employment situations and in other selection situa
tions because of the fear that the tests are relatively easy
to "fake." Many of these tests are excellent instruments
and it is important that an effort be made to make them more
useful. It is the purpose of this investigation to attempt
to increase the utility of these tests through altering the
test administration without changing the test in any funda
mental way.
In usual test practice, the instructions may leave
too much opportunity for individual personal response
tendencies to operate. Different motivating conditions are
elicited in the classroom situation than are elicited in the
employment situation. The present investigation deals with
increased structuring of the test situation. Guilford
(1954), discussing response sets, contends that the less a
test is structured the more opportunity exists for personal
response sets to be effective. The problem of structure has
been raised even in regard to the Rorschach Test. Benton
(1950) offers the suggestion that the Rorschach Test have
specific instructions so that certain types of response sets
may be induced by the examiner. The purpose of this would
be to reduce the operation of adventitious external factors
which may have unknown influence on the subject's response.
Guilford (1954) points out that although the response sets
which are personal to the subject may contribute to the true
variance of a test because they are inherent in the subject,
they do this at the expense of common factor variance which
the test intends to measure. He holds that the focus of
this bias is the problem because the direction of the focus
is not known. Structuring the test in such a way as to
focus the bias may be a fruitful approach.
In the present study the instructions to the subject
were altered in such a way as to focus the bias or response
set and thus to structure the test. In one way, both honest
and dishonest subjects were told to simulate when they
answer the test questions. In this manner it is expected
that the honest-dishonest variance may be ruled out. The
instructions to the subjects were also designed to motivate
them to conscientiously deal with the test questions, and by
the use of reference to the occupational goal-ideal it is
hoped that the test was made somewhat projective.
The ability of the test scores to operate as
effective predictors of a criterion will be the crucial test
of these changes in the structure of the test situation.
In this study, the experimental subjects, student
nurses, were instructed to respond to the Guilford-Zimmerman
Temperament Survey (GZTS) and the Dynamic Factors Opinion
Survey (DFOS) in the manner that a good graduate nurse would
respond. It was assumed that the responses under the new
testing conditions would still yield personality and inter
est scores. Because the instructions were oriented toward
an acceptable and highly valued occupational ideal it was
anticipated that they would be readily adopted and that the
test scores would have improved predictive ability. The
improvement in prediction should result from the reduction
of the influence of adventitious personal sets. The orien
tation of the instructions in the direction of the occupa
tion should bring out in the test protocols of the students
the effect of some type of insight into the job personality
requisites. A distinction between those who succeed or fail
on a criterion may depend on the assumption of a role-
concept, so that those who are successful "understand" the
supervisor's motivations and needs, and consequently their
behavior is such as to be found most acceptable. It is
recognized that a new variable might have been introduced as
a result of the change in the test instructions, such as one
relating to social intelligence, or role-playing ability.
The criterion to be predicted was that of academic success
in the training towards that occupational goal.
Hypotheses
It is assumed that predictive relationships exist
between the factors on the two tests used in this study and
the criterion of academic success. It is further assumed
that due to the random response sets which operate under
standard instructions these relationships are less than
maximum possible. It is also assumed that the experimental
instructions relating to the occupational ideal may create a
bias or response set which is highly correlated with the
criterion of academic success.
To answer the question, "What would be the effect of
controlling the personal set of subjects upon the predictive
ability of the personality and interest inventories?", the
following hypothesis will be tested in this investigation.
Hypothesis I. The instructions to adopt the role of
the occupational ideal (graduate nurse) will influence
responses to personality and interest inventories (GZTS and
DFOS) so that the scores which result will better differen
tiate good from poor subjects, relative to that role
(nursing students), than the inventories taken under stan
dard conditions. More of the factor scores will be more
predictive under experimental than under the control
conditions.
The criterion of pass-fail is dependent on the
instructors of the student nurses. Since in the major
hypothesis the various traits are assumed to be related to
the criterion, information is desired regarding the rela
tionship of the traits or factor scores to the instructors.
It is desirable to know which of the factors are seen as
important by the instructors. The instructors have taken
the test under the experimental instructions and their
results will be compared with the students. It is assumed
that a relationship exists between the factors that are im
portant for the instructors and the factors which are most
predictive of the criterion.
Hypothesis II. A positive relationship exists
between the factors which are found to be important to the
instructors and those factors found most predictive of the
criterion. This effect will be most pronounced under the
experimental conditions.
A popular score based on the majority response to
each item, which is a measure of a homogeneity of the group
should be more effective in predicting the criterion under
the experimental conditions than under the control
conditions.
Hypothesis III. The popular score will be more
predictive of the criterion under the experimental condi
tions than under the control conditions.
An additional response key derived for the GZTS
specifically for nurses (Beaver, 1955) should be more
effective under the experimental conditions than the control
conditions.
Hypothesis IV. The Beaver R. N. key will be more
predictive under the experimental conditions than under the
control conditions.
It is desirable to analyze the effect on the items
of changing the instructions. It may readily be assumed
that one effect will be that the subjects will attempt to
falsify their responses in the direction of social
acceptability. The Falsification Scales of Jacobs and
Schlaff (1955) is a measure of test-taking attitude on the
GZTS.
Hypothesis V. The Falsification Scale scores will
be higher under the experimental condition than the control
condition.
The 300 items of the GZTS were analyzed to study the
effect of the experimental instructions upon item responses.
The significance of the difference in the proportion under
each condition responding to each item as keyed in the test
manual was computed.
Hypothesis VI. Significant differences between
experimental and control conditions will occur in the pro
portion under each condition answering each item in the
direction as keyed in the test manual at greater than chance
frequency or expectancy.
If these hypotheses can be substantiated, it will be
accepted as support for the proposal that test-taking
attitude variance can be effectively channeled to aid
prediction.
CHAPTER II
BACKGROUND OF THEORY AND RESEARCH
Problems in Personality Measurement
Several concepts which have been discussed in the
literature bear upon this study. They are falsification,
social favorability, bias, response set and projection.
The problem of falsification in self-report inven
tories was identified a number of years ago. Steinmetz
(1932) found that college students could distort their
scores on the Strong Vocational Interest Blank in the
direction of a desired occupation regardless of their basic
score on the occupation. Kelly, Miles and Terman (1936),
found that on their own M-F test, college men and women
could drastically influence their scores on masculinity and
femininity under instructions to do so. Attention was
turned by Ruch (1942) to devising "detection" scores, where
by to identify faking. He developed "honesty" weights and
scores for the Introversion items of the Bernreuter
Personality Inventory, which accurately identified faking
introverts, based on the size of the critical ratios of the
differences in the frequencies of reply under honest versus
"salesman applicant" conditions. This effort was followed
by attempts to develop correction factors which would reveal
the extent of faking influence on specific scores. Ruch
(1945) worked on a "lie detector" key approach and the MMPI
developers (Meehl and Hathaway, 1946) introduced the K scale
for use as a suppressor variable with other MMPI scales.
Another approach has been to write test items and present
them in such a way that the respondent has to make a "forced
choice" between items of equal social favorability (Gordon,
1951; Edwards, 1954). Green (1951) and Heron (1956)
examined the effect upon inventory scores obtained under the
real-life motivation of an employment selection situation
compared with non-motivated circumstances. Their studies
revealed that adjustment inventories are responded to dif
ferently under employment motivation, and yield scores in
the direction of better adjustment. The approach of the
present study is to capitalize on the motivation of the
respondent in a selection situation, encouraging all sub
jects to simulate the responses of the successful selectee.
It is hoped that prediction will be improved when all the
respondents operate under the same set.
Guilford and Zimmerman (1949) in the manual of in
structions of the GZTS, discuss the problem of biasing
influences. They state that many of these influences may
operate upon a subject who takes the personality inventory.
There is no doubt that intelligent subjects can inten
tionally alter their scores but there is little clear
10
evidence of how much they actually do modify their scores
under ordinary testing conditions. There is evidence that
the change in their scores due to a bias will frequently be
in the unfavorable direction, as well as in the favorable
direction, even when the subject is trying to present him
self most favorably. Such changes are just as disturbing to
correct classification as are situations where the person
appears in a more favorable light. They also state that it
is known that some persons are set, consciously or uncon
sciously, toward unfavorable scores and many of them succeed
in influencing their test scores this way.
As a potential index of a probable invalid score,
Guilford and Zimmerman (1949) suggest that one of these is
the number of question mark answers given by a subject.
They point out that 1 1 ?" responses are always given a weight
of zero, and thus, they always contribute to an unfavorable
score. Many question marks on the answer sheet may mean
that the examinee does not know himself very well or that he
does not have much confidence in what he knows about
himself. It may also mean that he knows about himself but
he is evasive or secretive. Then again, it may be related
to the particular trait in question in that the person may
not be well organized in regard to that trait and does not
adequately respond to questions or statements in that area.
Since, under the experimental conditions of the
present study, the subjects respond to the test items not as
themselves but as some ideal person, the tendency might in
crease to feel that "question mark" is the most appropriate
answer. The subjects were discouraged from using the cate
gory of "?" and encouraged to respond, whenever possible,
with either a "Yes" or "No." Guilford and Zimmerman (1949)
report an unpublished study which dealt with the attitude of
students regarding the question mark category. About 60 per
cent of the students stated that they could not very well do
without the question mark category. Forty-seven per cent of
the group who were administered a test without the question
mark category present, said they resented being forced to
reply in one direction or the other and 9 per cent said they
resented it very much. Although the subjects may have re
sented being forced to reply, no information was presented
pertinent to whether the absence of the question mark cate
gory interfered with the efficiency of the test items.
Perhaps the resentment about being forced might be worth
considering in a situation where public relations is impor
tant, but the predictive validity of the test might be
improved by limiting the subjects to answer "Yes" or "No."
Cronbach (1946) discusses the various kinds of
response sets which effect test scores. He defines response
set as "any tendency causing a person consistently to give
different responses to test items than he would when the
same content is presented in a different form." He explains
"form" as form of the statements, choice of the responses
available and directions to the respondent.
Of the types of response sets which may be of im
portance to consider in this study, Cronbach (1950) lists
"gambling" and "bias of acquiescence." "Gambling" is the
tendency to answer an item when the appropriate answer is
"?" or "don't know," and "bias of acquiescence" is a
tendency to agree or answer "yes" to questions to a greater
frequency than to answer "no." Gambling set tends to in
crease the spread of individual differences and thus
increase test reliability; but it may attenuate the corre
lation between the true measure of the variable and the
criterion and thus reduce validity. Acquiescence bias has
been demonstrated in personality and attitude tests.
Phrasing test items such that "yes" represented a socially
undesirable answer increased item validity over items for
which "yes" was a more favorable answer, according to
Rundquist (1940). Individual differences in acquiescence
tendency has been found to be reliable by both split-half
and test-retest with equated forms methods, according to
Cronbach (1946)- He also reports that this type of bias may
reduce reliability of a test, because it reduces the range;
and that predictive validity may be increased or decreased
depending on the correlation between the bias and the
criterion. In general, individual differences in response
set have been found to be reliable. Response set has its
greatest influence on score in ambiguous or unstructured
13
situations. It causes systematic errors in measurement.
Sherif and Cantril (1946) describe response sets in
term of frames of reference. They review studies which con
firm the view that internal conditions of the organism
determine response in any unstructured situation. They
state :
In the absence of an objective scale or frame
and objective standard, each individual builds up
a scale of his own and a standard within that scale.
The range and reference point established with each
individual is peculiar to himself when facing the
situation alone. Once the scale is established,
there is a tendency for the individual to preserve
this scale in subsequent sessions. . . . These
frames or points of reference are by no means
always confined to consciously accepted instruc
tions or imposed norms but can become established
without an individual’s realization of it. (Sherif
& Cantril, 1946)
Cronbach (1946) states that essentially the notion
of response set that he develops is an application to the
field of testing of the findings reviewed by Sherif and
Cantril, the crucial question being whether response sets
are transient or fixed. Cronbach suggests that variations
in response set can be reduced by increasing the structure
of the test; such as by changing the directions or elimi
nating the "?" or failure to respond category (1946).
Berg (1955; 1959) contends that the response bias
may be very strongly related to personality. He proposes a
method of studying personality which concerns itself with
those persons who do not display a response set when the
majority of the subjects do display it, that is, in
14
situations where a response set can be identified. An
example of this situation is one where a large percentage of
the persons behave consistently in a certain way in
responding to the test items. He contends that those who
respond in the non-set method or pattern, such as saying
"False" when most people say "True," have more strongly en
trenched personality characteristics. He calls these
"deviant responses" and tends to show that these deviant
responses indicate that the person is deviant from the norm
in other ways on other measures.
Berg’s position, then, is that the subject's mode of
response is not transient, but may properly be considered an
enduring characteristic of the person. However, Cronbach
holds that personal response set may be reduced or elimi
nated by changing the instructions. He further states that
the predictive validity of a test may be increased when the
bias and the criterion are correlated. Sherif & Cantril
hold that in the absence of structure the personal set which
may develop is not always confined to consciously accepted
or outwardly imposed norms, but may be unconsciously
established.
In the manual for the GZTS, Guilford and Zimmerman
(1949) discuss the operation of the "projective principle."
In their discussion of the construction of items, they state
that avoidance of using personal pronouns tends to allay
resistance and increase the operation of the projective
principle. By projective principle, it appears that they
refer to operations which tend to make the statements in the
inventory less objectionable, less personal and less like an
inquisition or confession. They also refer to the operation
of devices which tend to throw the subject off guard and
cause him to forget that he is reporting about himself.
They suggest that using affirmatively worded statements,
which are not direct questions but with which the respondent
must signify some agreement, tend's to facilitate the pro
jective device or principle over those questions which are
directly related to the person. In the publication des
cribing the factor analysis of human interests, out of which
the DFOS was developed, Guilford, Christensen, Bond and
Sutton (1954) discuss projective items. They contend that
these items probably depend for their success on the same
psychological mechanisms that make projective tests so
effective, without the faults of a completely unstructured
test situation. One example is, if you ask someone "Are you
painstaking about your work?", he is likely to say "Yes"
whether he actually is or not, if he wants to make a good
impression. If, however, you ask him "Do you admire people
who are painstaking about their work?" he can readily say
"No" without necessarily realizing what he has revealed.
In such a way, one can make more attractive certain re
sponses that would be otherwise unattractive or make less
attractive responses that are socially popular. However,
16
the final proof of whether these projective items discrimi
nate as they are intended must depend upon how well they
correlate with other measures which are found to make a
similar discrimination between subjects.
. This projective principle perhaps would be more
effective under the circumstance where the subjects do not
take the test as themselves but take it as some hypothetical
person, as some ideal, or as some person who occupies their
concept of occupational success. This may somehow appear to
contradict the conception of projective tests as less struc
tured than non-projective ones, since in this instance there
is structuring of the situation. However, the aim of this
structuring is to de-emphasize the personal reference of the
individual who is taking the tests.
In the discussions on response set Cronbach, as well
as Sherif and Cantril, make a strong point of the fact that
response sets are most likely to be influential when the
situation is least structured. In this study, an attempt
is made to reduce the idiosyncratic response sets and im
pose a uniform response set in all the experimental
subjects. This structuring of the situation, it is hoped,
will reduce the effect of individual response sets.
Jarrett (1957), in a study which used the GZTS and
the DFOS inventories, hypothesized that when the tests were
given under special instructions, calculated to encourage
projection on the part of the individual, the tests would
17
yield personality characteristics which could be utilized
for predicting overall proficiency (of, in his case, munici
pal fire fighters). He further stated that the modification
of instructions to create a projective situation would not
alter the assumption that if personality is treated as a
pattern of traits or a way of reacting to external stimuli,
it is then both analyzable and unitary. Jarrett instructed
his subjects to respond in a manner which would make them
look like their image of a good fireman, even though this
might not be a true characterization of themselves. The
middle response category, was eliminated to prevent an
excessive number of responses. He does not state any basis
for justifying his elimination of the question mark
category. Jarrett also does not present any specific argu
ments toward establishing that his test instructions made
the test projective.
A comparison of Jarrett's data, under "occupational
ideal" response instructions, with the GZTS published norms
(Table 4 of the GZTS manual) is worth considering.
The means for Jarrett's subjects are higher than the
norm group on all of the traits except T. A comparison of
the standard deviations reveal that Jarrett's are slightly
smaller than the norm group, except on trait F. A compari
son of the DFOS factors data for Jarrett's two groups and
the norm group (male high school subjects) reveals that
Jarrett's mean scores are higher in all factors except
18
"Need for Attention" and "Need for Freedom." In Jarrett's
"sample two," one factor is lower than the norm group, it is
"Adventure versus Security."
Attention is called to these differences in con
sideration of what may be the effect on the means of in
structing subjects to answer the test as an occupational
ideal person would answer. What appears to have happened is
that this imposed set has elevated all of the scores. It is
not known whether the differences between Jarrett's mean
scores and the norm groups are significant. However, some
of them are large and appear to be significant. Since high
scores on the GZTS are in the positive direction, in the
direction of better adjustment, it seems that the imposed
set directed the subjects to present a "better adjustment"
personality than they would otherwise do. The reduced
standard deviations indicate reduction in variability has
occurred. The same interpretation of the DFOS mean scores
cannot be made since the factors are in the nature of
interests and on an a priori basis, it is not obvious
whether high or low scores are considered more socially
desirable.
A pilot study was done on psychiatric hospital ward
personnel by the present writer, in which 150 items of the
GZTS were taken under standard conditions and then 150 items
of the test were taken under special instructions. The
instructions were: "Take the test as you would assume the
19
best possible kind of' ’psychiatric technician’ would answer
these questions. Keep the role of this person in your mind
as you take the test." The test was divided in such a way
that one-half or 15 test items for each factor score
occurred under control conditions, and the other 15 occurred
under the experimental conditions. The tentative findings
for a small sample on the pilot study are in agreement with
Jarrett's data in that the imposed set tends to elevate the
score in the direction of good adjustment. A chi square
significant beyond .01 was found.
Thus, we find that there is a general tendency
toward socially favorable responses when the subject is told
to fake good, is looking for a job, or when he is asked to
assume an "occupational ideal" role. However, we do not
know about individual differences to respond this way.
Heineman (1953) in a study using the Manifest Anxiety Scale
had his subjects rate the items for social favorability„ He
found that anxious subjects rated the items more favorably
and non-anxious subjects rated them less favorably, signifi
cant at the .001 level. Each group thought the items they
endorsed were the socially favorable ones and yet they were
different items. In the present study, the endorsement of
socially favorable items, if that is what the new instruc
tions produces, was tested against the criterion of success
as nursing students. Test behavior under standard instruc
tions was evaluated against the same criterion.
20
Jacobs and Schlaff (1955) derived Falsification
Scales for the GZTS which measure test-taking attitudes.
Their Gross-Falsification (GF) scale is made up of items
usually not answered in the keyed direction, but easily
fakable by subjects under instructions to "give the most
favorable impression" of themselves. Their Subtle-
Falsification (SF) scale contains items answered in the
keyed direction approximately half the time, but which under
"fake good" instructions are even more likely to be answered
as keyed, indicating a mild tendency to impress favorably.
What effect will changing the test instructions have
on these items which have been scaled for frequency of
response? The falsification keys, when applied to the
experimental and control data may show if the effect of
imposing a set increases the falsification score. The fre
quency of certain item responses is perhaps related to their
acceptance by the general public, their social favorability.
To what extent will the students assume that answering as a
good nurse would answer is the same as answering the items
in the most socially acceptable way?
Studies of Nursing Personnel with the
Guilford Inventories
Healy and Borg (1951; 1952) gave the Guilford-Martin
GAMIN and STDCR tests and also the Personnel Inventory to
student nurses, graduate nurses and a norm group of female
university students. The nursing students were in the early
weeks of their training. The 187 student nurses were fol
lowed through the first year and the 140 who succeeded were
compared with the 47 who dropped out. They were compared on
the above mentioned tests and on the Kuder Personal Prefer
ence, Form CM. In reporting the results here, the factors
are reported in terms of the 10 factors of the GZTS. The
correspondence of the GAMIN, STDCR, and Personnel Inventory
factors to the GZTS factors follows the definitions in the
GZTS manual (1949). All scores are stated such that the
higher score pertains to more favorable adjustment. Healy
and Borg (1951) found that the graduate nurses had higher
scores than both the student nurses and the norm group on
the emotional stability factor (higher scores on traits D,
C and N); factor R - restraint (lower on R-rhathymia in the
STDCR test); factor F - agreeability and factor P - coopera
tiveness, all significant at the .01 level. The graduate
nurses were found to be higher on 0 - objective than both
the other two groups but not significantly. The authors
consider the following factors as very important in being a
good nurse; E, R, F and P (Healy and Borg, 1951, 1952).
Comparing the student nurses with the norm group it
was found that the student nurses were less "ascendant" and
"cooperative," and more "introverted" and "depressed," all
significant at the .01 level. In GZTS terminology this
means that the student nurses were lower on factors A, S, P
and E, than the norm group of women students.
22
In the comparison of the 47 students who did not
pass (academic failure - 33, dislike of nursing - 8, got
married - 6) with those who remained, those who did not pass
were significantly lower in factor P - cooperativeness
(Healy and Borg, 1952). The analysis of the data in terms
of "C-scores" reveals that more of the dropouts had very low
C-scores, in the zero-to-three category, indicating poor
adjustment on the factors E - emotional stability,
R - restraint, C - objectivity and P - cooperativeness.
The dropouts had fewer persons in the zero-to-three category
on the factors of A - ascendence and G - general activity.
On the Kuder, they found no significant differences
between the dropouts and the non-dropouts but they did con
clude that the dropouts were higher on persuasive and
clerical interests and lower on outdoor, scientific and
social service interests.
Beaver (1955), comparing 108 student nurses and 91
education majors in a liberal arts college, using the GZTS,
found that the two groups of subjects approximate the
average of the standardization group in five of the traits,
G, R, A, S, and T, but vary on the positive side in four.
The student nurses show greater deviation from the norm than
do the controls in traits E, 0 and F. In trait P, the con
trol group has a higher mean than the nurses, although both
groups have high mean scores, in comparison with the
standardization groups. The trait differences found to be
23
significant at the .05 level are E and S higher for the
nurses, and M higher for the control group. Trait F was
higher for the nurses, at the .01 level.
On further examination of the GZTS responses through
item analysis, Beaver (1955) found a number of choices which
differentiate the two groups. The item analysis identified
97 "Yes" or "No" choices (occurring on 67 of the 300 items)
in which there is a significant difference between the two
groups at the .05 level of confidence. Sixty-six of these
97 item choices are significant at the .01 level or better.
The author then attempts to describe the two groups in the
study on the basis of these items which significantly dif
ferentiated them. Content analysis of these responses gives
indications of personality patterns characteristic of the
two groups. The student nurse gives evidence of basic
emotional stability, interest in science and detail work,
relatively submissive tendencies and serious and idealistic
attitudes. The education major is less stable, has a more
tolerant attitude toward other people. She prefers litera
ture and music to science and mathematics. She is more
concerned with having a good time than is the nurse. She
is ascendant and courageous.
CHAPTER III
METHOD
Subjects
The subjects of the study consist of 200 female
student nurses and graduate nurse instructors from six
junior colleges. Nursing students were used for subjects
for several reasons. They are a group which is selected
from a wide geographical area using fairly uniform standards.
There are many student nurses in training at any one time
and their training period is relatively short (2 years).
From the point of view of a criterion measure, it is widely
known that student nurses have a high rate of attrition
during training. This high rate has proven very costly of
facilities and time in a profession with great personnel
shortages, so that improvement of selection of candidates
for training could be quite valuable. It is relevant to use
personality and interest measures in conjunction with
selection since personality factors are known to be impor
tant variables in work which is so closely involved with
people.
The six schools from which subjects were drawn are
all in Southern California. They all present the R.N.
degree after a two-year program of study. All of the
24
25
schools integrate the nursing students into the regular
junior college program, involving selection tests and
academic standards on a uniform basis. All six of the
schools are sufficiently similar in entrance requirements
and in course requirements, and all of the students must
pass the same state examination at the end of a two-year
program, so that it was felt there would be no error in
grouping the subjects as if all were combined in only one
school. To account for any differences that might have
existed in the schools, the subjects were randomly divided
so that each school is equally represented in the control
group and the experimental group. The fourteen instructors,
who are all graduate nurses and also teachers, were dis
tributed across the six schools.
The subjects were randomized in the following
manner. The alphabetical student list for each school was
numbered consecutively, from one. A table of random numbers
was entered and the list was re-ordered according to the
order that each subject's number was drawn. The first half
of the list was designated as one group and the other half
the other group. The list was scanned and the number of
subjects counted who were 24 years of age and over. Those
over age 24 were then evenly divided into the two separate
groups according to order of random selection. This pro
cedure resulted in two groups, each having equal numbers of
younger and older subjects, randomly assigned to the groups.
26
Then a coin toss was used to assign one group to the control
condition and the other to the experimental. What resulted
was two groups each containing 93 students, equally repre
senting the six schools, equated for sex and age. Table 1
presents the age for the two groups of subjects. The range
is 17 to 47 for the control group and 17 to 5 3 years for the
experimental group. The mean ages of the two groups were
found to be not significantly different from each other,
with the control group mean age of 24.46 and the experi
mental group, 24.31 years.
At the end of the first academic year the 93 sub
jects in each group fell into three criterion groups; pass,
fail, and dropout. Their distribution over the three
criterion groups was fairly uniform under both conditions.
Under the experimental condition 61 passed and 5 9 passed
under the control condition. For the fail group, 17 experi
mental failed and 18 control failed; for the dropouts, 15 of
the experimental group dropped out and 16 of the control
group.
Procedure
Design
The design of this experiment is a straightforward
method of difference design. The subjects are randomly
divided into two experimental groups. One group, the con
trol, responds in a normal way; the other group, the
27
TABLE 1
AGE FOR SUBJECT GROUPS IN YEARS
Subject groups N
Age
Range Mean S .D.
Control group 93 17-47 24.46 7.87
Experimental group 93 17-53 24.31 7.96
28
experimental, has some experimental intervention occurring.
The following experimental controls were exercised.
Regarding subject errors, a strictly random procedure was
utilized, assigning the subjects to treatment groups.
Regarding treatment group errors, the assignments of groups
to experimental treatments was done by strictly random
procedures. To deal with errors which may have occurred due
to selection of schools, all junior colleges who teach
nursing in this specific geographical area were included and
all female freshman students in each department of nursing
were included in the population. Also all of the instruc
tors who dealt with these freshman students were included.
The subjects were tested in the early weeks of their
first semester in nursing school. They were given two
tests : The Guilford-Zimmerman Temperament Survey and the
Dynamic Factors Opinion Survey. The control group took the
tests under standard conditions and the experimental group
took the tests under special instructions. The instructors
also took the tests at the same time under the experimental
conditions. In each school the experimental and control
groups were tested simultaneously in separate rooms'.
The experimental groups were orally instructed as
follows : "Answer the items of this survey the way you think
a competent graduate R.N. who likes her work would answer
them. We are not trying to find out how you answer, but how
you think a good R.N. would answer." The above was also
29
typed in red capital letters on the cover of the test book
let adjacent to the instructions on using the answer sheet.
The control groups were given the standard instructions and
also informed that this test would be important in the
school. It would not affect their grades but the test would
be available to the Nursing Department. All subjects placed
their names on the answer sheets. All subjects used a
special electrographic pencil.
At the end of the first year of school those sub
jects who failed were assigned to the fail group and those
who dropped out were assigned to the dropout group. These
criteria were then used to analyze the differences under the
two experimental conditions. The three criteria were pass,
fail, and dropout.
Measures
The tests
The tests used in this investigation were derived
through factor analysis. Each of the two test inventories
yields scores on ten factors or traits which are said to
have proven factor validity. Both inventories are, in a
sense, measures of personality variables. The Guilford-
Zimmerman Temperament Survey (GZTS) (1949) yields 10 scores
on temperament traits. The DF Opinion Survey, An Inventory
of Dynamic Factors (DFOS) (Guilford, Christensen and Bond,
1956) yields scores on 10 interest variables. The GZTS was
standardized on Southern California college men and women,
similar to those used in this study. The inventory is
widely used and well accepted. The temperament traits of
this inventory were also measured by several previous forms
of the test for a number of years. The previous forms are
superseded by the GZTS. The DFOS measures interests which
are defined basically as dimensions of motivation.
According to the test manual (Guilford, Christensen and
Bond, 1956), in addition to the more familiar vocational-
interest factors now generally covered in published inven
tories, there were found many factors that could be regarded
as broader and more basically motivational variables. This
inventory covering these motivational variables was an out
growth of a very comprehensive factor-analytic investigation
of interests (Guilford, Christensen, Bond, and Sutton,
1954). Both of these inventories are comprised of 300
items to which the respondent answers "yes," "?" or "no."
For each of the factors, 30 out of the 300 items make up
the score. A description of the factors is as follows:
The Guilford-Zimmerman Temperament Survey
(1949)
This survey follows the factor-analytic approach in
deriving ten factor measures of personality. The traits
are: General Activity (G) , Restraint (R) , Ascendance (A),
Sociability (S), Emotional Stability (E) , Objectivity (0),
Friendliness (F), Thoughtfulness (T), Personal Relations (P)
and Masculinity (M). The trait descriptions are as follows
High Low
General Activity
Has strong drive, energy. Lacks drive, energy, and
and activity. A very .high activity. A very low G
G score may indicate manic score may represent a
behavior, in which there is hypothyroid condition,
usually much random be- anemia, or other physical
havior and wasted effort. cause of inactivity. The
G factor is a catalyzer
for the other traits.
Restraint
Restrained and serious Happy-go-lucky, carefree,
and impulsive
Ascendance
Ascendant Submissive
Bold socially
High Low
Sociability
At ease with others, enjoys Withdrawn and reserved
their company and readily
establishes intimate rapport
Emotional Stability
Optimistic and cheerful Emotionally unstable
Emotionally stable Depressed
Objective
Objectivity
Touchy
Hypersensitive
Friendly
Agreeable
Introverted
Reflective
Friendliness
Hostile
Belligerent
Thoughtfulness
Extraverted
Thoughtless
Personal Relations
Gets along with others
Cooperative
Critical
Intolerant
Masculinity
Masculine in emotions Feminine in emotions
and interests and interests
33
The DF Opinion Survey
(Guilford, Christensen, and Bond, 1956)
This inventory is an outgrowth of a comprehensive
factor-analytic investigation of interests (Guilford,
Christensen, Bond, and Sutton, 1954). Interests are defined
basically as dimensions of motivation. Few, if any, of the
variables included in this inventory are now covered as such
in either interest or temperament inventories. The DF in
the title stands for "dynamic factors."
Ten traits are included, each assessed by means of
30 items. The traits are: Need for Attention (NA), Liking
for Thinking (LT), Adventure vs. Security (AS), Self-
Reliance vs. Dependence (SR), Aesthetic Appreciation (AA),
Cultural Conformity (CC), Need for Freedom (NF), Realistic
Thinking (RT), Need for Precision (NP), and Need for
Diversion (ND).
A brief description of the factor measures follows:
High Low
Need for Attention
Craves recognition Recognition means little
Enjoys status Status is unimportant
Is exhibitionistic
Liking for Thinking
Enjoys mathematical Dislikes mathematical
thinking, logical thinking, logical
problems, planning problems, planning
34
High Low
Adventure vs. Security
Likes to explore, to Seeks security
take personal risks Avoids danger
Is bold Is timorous
Self-Reliance vs. Dependence
Self-reliant Dependent
Responsible Seeks support
Dependable Subservient
Aesthetic Appreciation
Enjoys art in all forms— Not inclined to enjoy art
drama, music, literature,
or graphic arts
Cultural Conformity
Fully accepts social customs Rejects social customs
Highly developed conscience Little bothered by
conscience
Need for Freedom
Likes freedom Likes order, system, and
Nonconformist organized life
Dislikes system and order Accepts controls
35
High Low
Realistic Thinking
Takes realistic view of self Prone to wishful thinking
Matter-of-fact attitude Appreciates humor
Forthright and direct Expresses hostility
indirectly
Need for Precision
Likes exactness, precision, Dislikes exactness,
and detail precision and detail '
Need for Diversion
Craves amusement Little need for recreation
Likes to play Not playful or interested
Inclined to romanticism in-amusements
Special test scores
Popular scores.— Popular scores were derived on
both inventories in the following manner. For each group of
subjects (control group, experimental group, instructors
group) the number of subjects selecting each item response
(yes, ?, no) was counted. The item response that was
chosen by more than 5 0 per cent of the particular group was
then designated the popular response for that group. A
scoring key was constructed of only those popular responses
and each subject's answer sheet was scored, yielding a
single popular score. In addition, the popular scoring key
36
derived from the responses of the instructor's group was
applied to the student's answer sheets and this score was
obtained for each student. So, for each student there were
two popular scores based on (1) the student's own group key
and (2) instructor's key. Six popular scoring keys were
derived; GZTS control group students, experimental group
students, and instructors; and for the DFOS, control group
students, experimental group students, and instructors. The
number of items in each popular score key (that is, the
maximum possible score, is as follows: GZTS, control 283,
experimental 296, instructors 280; DFOS, control 2 77,
experimental 289, instructors 263. The popular score is
then the typical way of response of the particular group it
is based on. It is a measure of homogeneity of group
response.
R.N. key.— As part of a study of nurses using the
GZTS, Beaver (1955) did an item analysis differentiating the
student nurses from education majors in college. On 42 of
the items there were significant differences between propor
tions, separating the nursing students from the education
majors. These 42 items comprised the R.N. key for this
study. All of the subjects in the experiment, including the
instructors, were scored for the R.N. key score as an addi
tional score of their GZTS protocols. This additional score
was obtained for the purpose of seeing if better prediction
would occur under the experimental conditions than under the
37
control conditions on this score which was derived specifi
cally as a differentiation of nursing students.
Falsification scales.— Test-taking attitude scales
have been derived for use with the GZTS by Jacobs and
Schlaff (1955). Two of the three scales which were de
veloped refer to the tendency to create a favorable impres
sion when taking the test. The third scale is one to
indicate carelessness in answering the questions or extreme
deviation from the population tendency. The population upon
which the falsification scale was derived is similar to the
college student population of the present study. The simi
larity between the populations makes the falsification
scales readily applicable to the subjects in the present
research. The Gross-Falsification (GF) Scale was made up of
items usually not answered in the keyed direction but easily
fakable under instructions. These are items which are
answered in the direction as keyed in the test manual by
less than 50 per cent of the group taking the test under
standard instructions but by more than 5 0 per cent of a
group taking the test under instructions to 1 1 fake good."
In addition, the difference between the "fake good" and the
"standard" group was at least 40 per cent of the group.
Thirty-four items of the GZTS are part of the GF scale. The
other attitude scale. Subtle Falsification (SF), was made
up of fakable items which were answered in the keyed direc
tion half the time under standard instructions. The SF is
38
a moderately popular response under standard instructions,
which becomes a popular response under "fake good"
instructions. The scale authors considered that this scale
will pick up those examinees who tend to give themselves the
benefit of the doubt on items often answered in an unfavor
able direction. There are 17 items on the SF scale. The
third scale, Carelessness-Deviancy (CD), consists of 28
items which are answered by less than 10 per cent of the
population under the standard conditions. The ..scale manual
reports odd-even reliabilities for females of .73 for GF and
.70 for SF. The falsification scales are included in this
research as a measure of what happens to the item responses
under the experimental instructions. The questions may be
raised, "Are the subjects faking good?",* "Are they merely
faking good?"; and "Are they faking good but also changing
their response in some additional way?".
The criteria
In a preliminary survey of several nursing schools
and also a survey of the literature (Healy and Borg, 1952),
it has been learned that approximately one-third of a
beginning class will not be in class at the end of the
first year. This large amount of dropout has been a problan
for the schools. In using dropout as a criterion, it is
recognized that persons may drop out or be dropped for
various reasons. Some of the reasons may seem unrelated to
39
the variables which are measured in this study. However,
the manifest reasons for dropping a course in training might
be quite different from the real reason which motivates a
person to change his course of action after taking definite
steps to follow a certain profession. Dropping out for
reasons that seem to be unrelated to the temperament and
attitude measures might actually be so related. Inability
to do the course work may be one such reason since usually
this is not due to low intelligence, as all nursing schools
are intellectual level screening devices, such as the A.C.E.
All schools also use evaluation of the student's high school
achievement.
The criterion of success or failure in training used
in this study is whether or not the student has dropped out
or has failed at the end of the first academic year. This
is a dichotomous criterion, pass versus fail. It is assumed
that the criterion is artificially dichotomized and that the
characteristics the criterion is based upon are normally
distributed in the student nurse population.
The "simple" pass-fail criterion is somewhat diffi
cult to evaluate. Both aspects of the dropout criterion,
academic failure, and voluntary dropping out are complex.
The academic failure is based on a combination of academic
grades, evaluations of hospital ward work and evaluations by
the instructors. Although the academic grades are fairly
uniform from school to school, there is no uniformity in the
40
evaluations of the ward work. The students are placed in
numerous hospitals and clinics and it has not been possible
to set up any uniform system of grading. The evaluations by
the instructors are complicated by the fact that the in
structors are graduate nurses and their concept of what
makes a good nurse will undoubtedly influence their evalua
tion of the students who may soon join their profession.
The voluntary dropout group in the present study dropped out
for the following reasons: romance and family complica
tions, financial difficulties, health problems, and change
of occupational interest. Such complex features involved in
the criterion and the lack of more than one measure of it
make it impossible to evaluate the reliability of the
criterion. In a correlational problem it is most important
to correct for reliability in the criterion, but it was not
possible to do so in this study. Thus, the validity corre
lations were not corrected for attenuation. The proportion
of dropouts in the present study were quite similar to the
one-third reported in the literature. For the control group
36 per cent did not pass and for the experimental group
34 per cent.
Statistical Procedures
The standard scoring keys were utilized for the GZTS
and the DFOS inventories. In addition, special scores on
the GZTS were obtained for the R.N. key of Beaver, the three
falsification scales of Jacobs and Schlaff, and for the two
popular scores derived in this study. Additional scores for
the DFOS consisted only of the two popular scores derived in
this study. Means and standard deviations were computed for
all groups under all conditions for all scores. To measure
validity of the 10 factor scores on each of the inventories
and for certain of the other scores, biserial correlation
coefficients using the dichotomous criterion of academic
success were computed. For significance of the difference
between experimental and control conditions, to evaluate the
effect of the experimental instructions, critical ratios
were computed, based on Fisher's transformation of the
validity r's. These were done for the fail criterion group
and the dropout criterion group. This is a test of the
major hypothesis. In order to satisfy the assumptions
necessary to justify computation of biserial r, frequency
distributions were made for the variable-criterion
relationships. Inspection of the distributions revealed the
regressions to be essentially unimodal and fairly symmetri
cal, so that homoscedasticity was assumed. It is assumed
that the fail criterion, although dichotomized, is basically
continuous and normally distributed in the student nurse
population.
To further evaluate the main hypothesis multiple
regression coefficients were determined by the Doolittle
method, for both inventories. Multiple R's and Beta weights
were derived for both criterion groups under both experi
mental conditions, comprising 4 multiple R's for each
inventory. Matrices of the correlation coefficients were
computed and are reported in Appendix A. Various t tests
were run between the experimental and control condition to
test the effect of the change in instructions. The total
experimental and control groups on all the factors were
evaluated this way as well as the Beaver R.N. key scores.
In order to relate the test factors to the criteria
for the instructors, significance of the difference tests
(t tests) were run between the instructors' scores and the
separate criterion groups on all of the factors. .To further
understand the relationship of the instructor's evaluation
of the factors, in terms of their selection of them in the
test, a profile analysis was made. The mean scores of the
instructors and the several criterion groups were trans
formed to standard scores, based on the mean T-score of 50
for the norm group on the GZTS. These profiles, based on
the T-scores were compared between the instructors and the
criterion groups by a method adapted from Guilford (1956)
for testing the significance of the difference between
paired means. For each factor, the standard scores for each
of the two groups was treated as a pair of scores and a
significance test for the mean difference was computed over
the 10 factors. This gives us a measure of distance between
the profiles.
43
The popular scores derived in this study were evalu
ated by means of validity coefficients to see how well they
predicted the criteria. For the falsification scales on the
GZTS, t tests of the difference between means were computed
between the experimental and control conditions. Also the
mean scores were compared with those derived under standard
and faking conditions in the study from which the falsifi
cation scales were developed (Jacobs and Schlaff, 1955).
An item analysis was also done on the GZTS since
there was much information from other sources regarding the
items, such as the work of Beaver on the R.N. key and the
falsification scales. Significance of the difference tests
between the proportions of each group scoring on each item
were computed for the experimental and control conditions,
because of interest in the effect of instruction on the
items. The frequency that each item was answered in the
direction keyed in the manual was tallied for each of the
criterion groups. The raw frequency count was converted to
proportions. The analysis of the items was done by
obtaining the significance of the difference between inde
pendent proportions as described by Guilford (1956, p. 221).
This method uses a pooled estimate, or weighted mean, of the
proportions in both groups in devising the standard error of
the difference between proportions. It was necessary in
many of the cases to take cognizance of and to appropriately
deal with the problem of the obtained pooled proportions
44
being too extreme. Since the sampling distribution of
critical ratios obtained from this method approaches
normality closely enough to allow use of the normal curve
distribution, provided that the smallest product of the
weighted mean of p or of q times or N2 is not less than
10, a correction for discontinuity, as recommended by
Guilford (1956), was used in those cases where this product
was between 10 and 5. This correction involves subtracting
one half of the sum of the reciprocals of the two N's from
the difference between the two proportions. Where the
product was less than 5, no test of significance was made
on the item.
In most comparisons where the N's were approximately
equal, a short method of obtaining the significance of the
difference between proportions was used based on the tables
by Jurgensen (1947). His method involves obtaining the phi
value from the proportions and then multiplying phi by the
square root of N to obtain the critical ratio. Several
randomly selected computations were made, using the
Jursensen method as well as the longer formula reported by
Guilford (1956) and the critical ratio obtained using phi
was always within .002 of the other, and when it erred, it
erred in the direction of being smaller, therefore being a
stringent test. An item analysis was also done on the items
used by Beaver in the R.N. key as a cross-validation of her
work. In addition to the items that were scored as keyed in
the Guilford manual, in Beaver's list some of the items were
keyed in the opposite direction and so they required addi
tional testing of the significance of the difference between
proportions using the method just described.
CHAPTER IV
RESULTS
The data available for analysis in this research
consist of scores obtained from two personality inventories,
for three groups of subjects, having various positions on
the criterion. The three subject groups consist of two
equated groups of student nurses, separated into experi
mental and control conditions, and one group of nursing
instructors. The test variables consist of the 20 factors
normally obtained from the GZTS and DFOS inventories, plus
additional measures on the GZTS: the falsification scales
and the Beaver R.N. score; and also two additional popular
scores on each of the two inventories which were derived in
the present study. Positions on the criteria consist of
pass, fail, or dropout. The level of probability adopted
for significance is .05.
For a first impression of the effect of the experi
mental instructions on the temperament trait scores, means
and standard deviations were computed for the total subjects
in each of the two test conditions, and t tests were run for
the 10 factors. Table 2 presents these data, wherein it can
be seen that the difference between the means for all the
46
47
TABLE 2
COMPARISON OF MEANS OF GZTS FACTORS UNDER EXPERIMENTAL
AND CONTROL CONDITIONS FOR TOTAL GROUPS
Factor
Control
(N=93)
Experimenta1
(N=93)
Mean S.D. Mean S.D. t
^G 17.34 5.07 19.24 3.93
**
2 .85
R 19.14 3.41 21.55 2 .73
**
5 .31
A 15 .00 5 .84 18.83 4.18
**
5 .14
S 21.02 5.20 25 .05 2 .67
**
6.65
E 20.26 5.12 24. 71 3.78
**
6.74
0 20.53 4.59 23 .63 3.53
**
5 .16
F 19.82 5.32 22 .93 4.13
**
4.46
T 18.73 4.26 21.10 3.64
**
4.08
P 21.69 5 .06 23 .84 3.57
**
3 .35
M 12 .80 4.03 14.66 2.59
**
3.74
icic
Significant at p. = .01
48
10 factors were found to be significant at the .01 proba
bility level, with the experimental conditions having higher
mean scores in all cases. This is clear evidence that the
experimental instructions have influenced test response.
The influence was in the expected direction for the 10
temperament traits, since they are scored in such a way that
higher scores indicate better adjustment. It was antici
pated that under the experimental instructions there would
be a tendency to bias responses in the direction of social
favorability. This effect is further analyzed later in the
study.
Inspection of the individual scores on the GZTS
reveals that under the experimental condition many of the
scores are considerably higher than those obtained under the
control condition. The means are higher and all of the
standard deviations are lower. This type of influence on
the test scores, particularly the influence on the varia
bility will tend to reduce the correlation between the
scores and the criterion under the experimental condition.
It might appear that a correction for restriction of range
is called for in this instance. However, according to
Guilford (1956), this condition cannot be treated as a
restriction of range. What has occurred is a condition
of the ceiling of the tests being too low. For a cor
rection of restriction of range to be applied the limita
tions, which are being corrected for, have to be in the
nature of limitations on the subjects. When the subjects
are pre-selected and as a result the variability of the
trait in the population is limited then restriction of range
may be applied. In this instance the experimental subjects
were tested without any restriction on their response range,
since they were chosen randomly and are equated with the
subjects under the control condition. Whatever restriction
in the range of scores has occurred has only occurred
because of the behavior of the subjects under the experi
mental condition. On the DFOS this problem is not raised at
all. The scores and the means under the experimental condi
tion do not deviate in any regular way from the control
condition, that is, some of the scores are higher and some
are lower under the experimental condition than the control
condition. Also, the standard deviations are not uniformly
smaller nor larger under the experimental condition.
In the early phases of this research it was con
sidered that the criterion of fail include both the subjects
who failed for academic reasons and those who dropped out.
However, inspection of the sub-group means revealed that
there were obvious differences between the fail group and
the dropout group. In Tables 3 and 4 are presented the
means and standard deviations for the separated pass, fail
and dropout groups for the experimental and control
conditions. For the control condition it was found that the
mean for the pass group fell between the means of the fail
50
TATBLE 3
GZTS MEANS AND STANDARD DEVIATIONS FOR CONTROL
CONDITION FOR PASS, FAIL AND DROPOUT GROUPS
Pass (n=5 9) Fail (N=18) Dropout (N=16)
Factor ---------------------------------------------------------
Mean S.D. Mean S.D. Mean S.D.
G 18.24 5 .10 14.5 0 4.95 17.26 3.97
R 19.58 3.39 18.11 3.03 18.69 3.76
A 15.31 5 .83 13 .56 6.57 15 .50 5 .07
S 21.53 4.68 19.11 6 .62 21.31 5 . 07
E 20.85 4.70 19.33 6.48 19.13 4.89
O 21.41 4.21 18.83 4.77 19.19 5.14
F 19.81 5 .12 20.72 4.95 18.81 6 .56
T 18.92 4.38 18.33 3.63 18.50 4.60
P 21.86 4.45 22 .11 5 .54 20.56 6 .68
M 12 .95 3.97 13.44 4.31 11.50 3.97
51
TABLE 4
GZTS MEANS AND STANDARD DEVIATIONS FOR EXPERIMENTAL
CONDITION FOR PASS, FAIL AND DROPOUT GROUPS
Pass (N=61) Fail (N=17) Dropout (N=15)
Factor ------------------------------------------------------
Mean S.D. Mean S.D. Mean S.D.
G 19.23 3 .83 17.88 4.41 20.80 3.49
R 21.39 2.59 21.00 3 .28 22 .80 2.57
A 19.07 3.37 15 .24 4.81 21.93 3.64
S 25 .28 2.39 23.12 2 .88 26.33 2.53
E 24.92 3.38 22 .47 5 .04 26.40 2 .56
0 24.10 2 .87 20.65 4.76 25.07 2.53
F 22.82 4.18 21.94 3.44 24.47 4.54
T 21.10 3.91 21.35 3.03 20.80 3.36
P 23.72 3.76 22.12 2.56 26.27 2.21
M 14.62 2.51 14.35 2 .79 15 .13 2.95
52
group and the drop-out group for 4 out of the 10 factors.
For the experimental condition this occurred for all 10 of
the factors. In 14 of the 20 instances observed then, the
biserial r's computed would be of opposite sign for the fail
than for the dropout subjects, indicating a distinct dif
ference between the fail and the dropout subjects. It is
evident that when fail and dropout would be combined these
differences would tend to cancel each other out and make the
pass group look very much like the combined non-pass group.
Therefore, for the remainder of this investigation the fail
and dropout subjects have been separated. It was fortunate
that these two groups had approximately equal N's. Under
the control condition there were 18 in the fail group and
16 in the dropout group, and for the experimental condition
there were 17 in the fail and 15 in the dropout group.
Tables 5 and 6 present similar data for the DFOS test. The
previously mentioned effect is not so evident in the DFOS,
but since much more is known about the GZTS, and information
is available about the social desirability valence of the
factors, cues are taken from the effects on the GZTS and the
DFOS was then treated in a similar manner.
Experimental effects directed toward the main
hypothesis, which concerns prediction under experimental
versus control conditions were dealt with by correlational
analysis in the rest of this study. Table 7 presents the
validity coefficients for the GZTS factors under the
53
TABLE 5
DFOS MEANS AND STANDARD DEVIATIONS FOR CONTROL
CONDITION FOR PASS, FAIL AND DROPOUT GROUPS
Pass (N=5 9) Fail (N=18) Dropout (N=16)
Factor -------------------------------------------------------
Mean S.D. Mean S.D. Mean S.D.
NA 12 .52 5.89 13.36 6.12 12 .62 5.87
LT 15 .60 5.88 13.56 5 .98 12 .49 5.61
AS 12 .58 5.40 11.74 5.46 12.32 5.01
SR 20.19 4.02 19.58 3.96 18.93 4.09
AA 17.09 6.59 16.97 6.74 16.63 6.60
CC 16.22 3.78 14.85 3.81 13.30 3.75
NF 9.39 4.21 9.35 4.27 9.74 4.13
RT 18.67 5 .22 17.41 5 .25 19.23 5 . 34
NP 11.73 6.87 10.43 6.92 8.69 6.54
ND 15.34 5.68 16.61 5 .74 15 .86 6.01
54
TABLE 6
DFOS MEANS AND STANDARD DEVIATIONS FOR EXPERIMENTAL
CONDITION FOR PASS, FAIL AND DROPOUT GROUPS
Pass (N=61) Fail (N=17) Dropout (N=15)
Factor -------------------------------------------------------
Mean S.D. Mean S.D. Mean S.D.
NA 11.10 4.92 11.35 4.97 11.01 5.02
LT 17.97 5 .86 16.05 5 .84 18.91 5.95
AS 11.09 5 .34 8.79 5 .32 9.78 5.37
SR 23.00 3.68 19.62 3.71 23.36 3.40
AA 19.89 5.97 15.17 6.06 19.84 5.72
CC 15 .96 4.16 16.10 4.21 17.42 3.80
NF 6.88 3.27 8.39 3.31 4.55 3.04
RT 21.87 4.25 19.44 4.27 21.91 4.17
NP 15.11 6.69 14.07 6.64 18.03 6.97
ND 17.26 4.57 16.23 4.57 15.98 4.58
55
TABLE 7
COMPARISON OF VALIDITY COEFFICIENTS FOR GZTS FACTORS
UNDER EXPERIMENTAL AND CONTROL CONDITIONS FOR FAIL
CRITERION, USING FISHER'S z TRANSFORMATION
Factor
Experiments1
(N=78)
Control
(N=77)
Critical
Ratio
(r bis) z (r bis) z
(E - C)
G .197 .200
**
.414 .441 -1.470
R .083 .083
k
.257 .263 -1.099
A
**
.550 .618 .170 .172
**
2 .723
S
**
.474 .516
*
.269 .276 1.464
E '
**
.362 .379 .172 .174 1.254
0
**
.549 .617
**
.338 .352 1.618
F .126 .127 -.105 -.105 1.416
T -.039 -.039 .082 .082 - .739
P
*
.257 .263 -.031 -.031 1. 795
M .061 .061 -.071 -.071 .806
*
Significant at
**
Significant at
P- =
P- =
.05
.01
56
experimental and control conditions for the fail criterion.
The biserial coefficients are presented, as well as the
Fisher's transformation which was used in the significance
of the difference tests between the experimental and control
conditions, also reported in Table 7 under critical ratio.
The critical ratios were computed in the direction of
experimental minus control; so for those factors for which
the control group has a higher coefficient, the critical
ratio appears as a minus value. Table 8 presents a similar
comparison, with data from the GZTS, for the dropout
criterion.
Based on the fail criterion, for 5 of the 10 GZTS
factors, under the experimental condition, the r's were
significantly different from zero, and under the control
condition 4 were significant; whereas, for the dropout
criterion, 7 of the 10 factors under the experimental condi
tion were found significant and under the control condition
only one. The major test of the difference, which is the
crux of the main hypothesis, that the experimental instruc
tions would increase the predictive ability of the tests, is
evaluated by the critical ratio test for the 10 factors.
For the dropout criterion 8 of the 10 critical ratios were
found to be significant, while for the fail criterion com
parison only one of the 10 critical ratios was significant.
It appears that Hypothesis I is partially substantiated at
this point; for the dropout criterion there is definite
57
TABLE 8
COMPARISON OF VALIDITY COEFFICIENTS FOR GZTS FACTORS UNDER
EXPERIMENTAL AND CONTROL CONDITIONS FOR DROPOUT CRITERION,
USING FISHER'S 2 TRANSFORMATION
Factor
Experimental
(N=76)
Control
(N=75)
Critical
Ratio
(r bis) z _ (r bis) z _
(E - C)
G
*
-.238 -.243 .117 .118 -2.181*
R
**
-.310 -.321 .148 .149
**
-2.840
A
**
-.459 -.496 -.020 .020
**
-2.876
S
*
-.250 -.255 .027 .027 -1.707
E
*
-.260 -.266 .210 .213
**
-2.894
0 -.196 -.199
*
.287 .295
**
-2.985
F
*
-.221 -.225 .105 .105 -1.994*
T .046 .046 .055 .055 - .054
P
**
-.399 -.42 3 .151 .152
**
-3.474
M -.113 -.114 .211 .214
•k
-1.982
*
Significant at p. = .05
Significant at p. = .01
58
evidence of increased prediction, but for the fail criterion
there is no such evidence.
An additional method was utilized to test the main
hypothesis. Multiple correlation coefficients were computed
for the 10 factors and the criterion for each of the cri
terion groups under each of the experimental conditions.
Four multiple R's were computed for the GZTS analysis; they
are, control fail, experimental fail, control dropout, and
experimental dropout. Multiple R is used in this instance
as a means of evaluating the combined effect of the corre
lation coefficients over all 10 factors in each of the
tests. Although there is no method presently available for
evaluating the significance of the difference between
multiple R's, and though this is an unusual use of that
statistic, it was applied because it was felt that it could
contribute some information. Table 9 shows the four mul
tiple R's and their levels of significance. The Beta
weights are reported in Appendix C. For the fail groups,
the multiple R's are quite similar in the comparison between
the control and experimental conditions, however, for the
dropout criterion groups there is a considerable difference j
between the two multiple R’s finding .286 for the control
condition and .483 for the experimental condition. This is
further support for the main hypothesis for the dropout
criterion.
TABLE 9
COMPARISON OF MULTIPLE CORRELATIONS, UNDER EXPERIMENTAL
AND CONTROL CONDITIONS, FOR THE 10 GZTS FACTORS
WITH THE FAIL AND DROPOUT CRITERION GROUPS
Criterion Group Condition N R
Fail Control 77
**
. 5041
Fail Experimenta1 78
2**
.508
Dropout Control 75 .286
Dropout Experimental 76
*
.483
Note: These values of R when calculated using
biserial r for the criterion are (1) .684** and (2) .717 *.
All other values of R were derived using point biserial r
for the criterion.
*
Significant at p. = .05
* *
Significant at p. = .01
60
For the DFOS inventory Tables 10, 11 and 12 present
a similar evaluation as that reported for the GZTS. For the
DFOS, Table 10 shows that 5 of the validity correlation
coefficients for the fail group under the experimental con
dition are significant and none is significant for the
control condition. For the dropout criterion (Table 11),
there are 2 significant r's under the experimental condition
and 3 under the control condition. Comparing the multiple
R's, in Table 12, the comparison between the experimental
and control condition favors the fail criterion group,
whereas for the dropout criterion the 2 multiple R's are
similar. The above appears to indicate that the fail
criterion is best predicted by the experimental conditions
of the DFOS test. On the GZTS it was found that the drop
out criterion was best predicted under the experimental
instructions. However, in observing the critical ratios of
the differences between experimental and control condition
in Tables 10 and 11, it is seen that for the dropout condi
tion there are 4 significant critical ratios and only 2 for
the fail criterion. This seems to run counter to the
previous indication. Inspection of Table 11 for dropouts,
shows that in 3 out of the 4 instances of significant
critical ratios, the control condition r's were higher than
the experimental, while for fails, Table 10 reveals that
experimental condition r's were higher for both significant
critical ratios. This argues against the dropout criterion
61
TABLE 10
COMPARISON OF VALIDITY COEFFICIENTS FOR DFOS FACTORS
UNDER EXPERIMENTAL AND CONTROL CONDITIONS FOR FAIL
CRITERION, USING FISHER'S z_ TRANSFORMATION
Factor
Experimenta1
(N=7 8)
Control
(N=77)
Critical
Ratio
(r bis) z_ (r bis) z_
(E - C)
NA -.029 .029 -.077 .077 .293
LT .192 .194 .207 .210 - .098
AS
*
.247 .252 .091 .091 .983
SR
**
.529 .589 .093 .093
**
3.028
AA
**
.447 .481 . 007 .007
**
2.894
CC -.024 .024 .209 .212 -1.441
NF
*
-.297 .306 -.003 .003 -1.850
RT
* *
.327 .339 .142 .143 1.197
NP .094 .094 .106 .106 - .073
ND .133 .134 -.126 .127 1.593
*
Significant at p. = .05
Significant at p. = .01
62
TABLE 11
COMPARISON OF VALIDITY COEFFICIENTS FOR DFOS FACTORS UNDER
EXPERIMENTAL AND CONTROL CONDITIONS FOR DROPOUT CRITERION,
USING FISHER'S z TRANSFORMATION
Factor
Experimenta1
(N=76)
Control
(N=75)
Critical
Ratio
(r bis) £ (r bis)
(E - C)
NA .011 .011 -.010 -.010 .127
LT -.086 -.086
kk
.319 .330
• k
-2 .514
AS .138 .139 .028 .028 .671
SR -.056 -.056 .177 .179 -1.420
AA .001 .001 .035 .035 - .205
CC -.219 -.222
k k
.450 .485
**
-4.272
NF .403 .428 -.062 -.062
**
2 .961
RT -.001 -.001 -.056 -.056 .332
NP -.242* -.247 .272* .279
**
-3.178
ND .162 .163 -.052 -.052 1.299
*
Significant at p. = .05
Significant at p. = .01
63
TABLE 12
COMPARISON OF MULTIPLE CORRELATIONS, UNDER EXPERIMENTAL
AND CONTROL CONDITIONS, FOR THE 10 DFOS FACTORS
WITH THE FAIL AND DROPOUT CRITERION GROUPS
Criterion Group Condition N R
Fail Control 77 .300
Fail Experimental 78
**
.523
Dropout Control 75 .401
Dropout Experimental 76 .402
**
Significant at p. =.01
64
being better predicted. Comparing this with the findings
for the dropout group, under the GZTS, Table 8, it is seen
that of the 8 factors with significant critical ratios only
2 of them have higher r's under control conditions. Thus,
it may be concluded that on the DFOS, the experimental in
structions appear to have increased the prediction for the
fail criterion.
The instructors of the student nurses, who are the
most important individuals in determining the position of
the students on the pass-fail criterion, took the test under
the experimental instructions. For the basis of comparison
between the instructors and the students, it is seen that
the instructors are closer to the students responding under
the experimental conditions than those under control condi
tions, in terms of the personality traits. This is because
the instructors are now at the occupational goal, they are
already successful graduate nurses. This is the goal-ideal
that is intended to influence the students under the experi
mental instructions. Therefore, it is felt that the nurses'
test protocols will not be changed much by the instructions
from their normal stance, and in addition it is hoped that
the pattern of responses derived this way from the instruc
tors will be close to an "ultimate profile" for the occupa
tional goal-ideal. Because the instructors primarily
influence the academic failure of the students and not the
behaviors of those who drop out for non-academic reasons,
65
this analysis is restricted only to the fail group.
Tables 13 and 14 present the means and standard deviations
for the instructors on the two inventories. It is desired
to relate the personality factors which the instructors view
as important to those traits that have demonstrated predic
tive validity for the subjects. If a correspondence is
found it will be evidence that the instructors use person
ality variables as a clue to arrive at the pass-fail
decision. It is hypothesized that this effect will be most
pronounced under the experimental conditions. Table 15
presents the results of comparisons between the means of the
instructors and the pass group, and the instructors and the
fail group separately for the control condition on the GZTS.
Ten t tests are reported for the factors under the two
criterion conditions. Table 16 presents similar data for
the experimental condition. To follow this analysis it may
be necessary to refer back to Tables 3 and 4 which present
the means for the pass and fail groups under the control and
experimental conditions. Also, see Table 17, for it pre
sents an estimation of the ranking of the traits for the
various groups. Table 17 lists T-scores which have been
standardized, based on the original standardization norms
for the GZTS, using 50 as a mean.
Taking the four traits that the instructors are
highest on, E, O, F and P, and comparing these with the
traits which demonstrated significant validity under the
TABLE 13
GZTS MEANS AND STANDARD DEVIATIONS FOR INSTRUCTORS (N=14)
G R A S E 0 F T P M
Mean 19.86 21.14 17 .29 23.93 26.36 24.57 24.64 18.86 25 .28 15 .00
S.D. 3.87 3.46 5.19 3.46 3.16 4.12 4.36 3.61 3.46 3.00
DFOS
TABLE 14 .
MEANS AND STANDARD DEVIATIONS FOR INSTRUCTORS (N=14)
NA LT AS SR AA CC NF RT NP ND
Mean 9.57 15.71 9.07 22.00 19.57 11.79 9.07 22.86 9.93 11.29
S.D. 5.29 6.71 4.36 4.00 6.78 4.58 3.16 4.79 7.21 6.16
cn
O ' !
67
TABLE 15
SIGNIFICANCE OF THE DIFFERENCE BETWEEN MEANS OF
INSTRUCTORS AND PASS GROUPS; AND INSTRUCTORS
AND FAIL GROUPS FOR THE CONTROL CONDITION
ON THE GZTS FACTORS
Instructor-Pass Instructor-Fail
Factor
Mean Mean
difference t difference t
G 1.62 1.321 5 .36
**
3.451
R 1.56 1.542 3 .03 2 .630*
A 2 .00 1.272 3.75 1.811
S 2 .44 2.140* 4.80 2.670*
E 5.50
**
5 .142 7.01
**
4.022
0 1.15 .921 5 . 75
i k *
3.660
F 4.80
**
3 .550 3 . 91
*
2.362
T - .06 - .051 .53 .422
P 3.40
**
3 .140 3.15 1.981
M 2.05
*
2.152 1.60 1.250
*
Significant at p. = .05
Significant at p. = .01
68
TABLE 16
SIGNIFICANCE OF THE DIFFERENCE BETWEEN MEANS OF
INSTRUCTORS AND PASS GROUPS; AND INSTRUCTORS
AND FAIL GROUPS FOR THE EXPERIMENTAL
CONDITION ON THE GZTS FACTORS
Instructor-Pass Instructor-Fail
Factor
Mean
difference t
Mean
difference t
G .63 .555 1.98 1.338
R - .25 - .258 .14 .116
A -1.78 -1.220 2 .05 1.128
S -1.35 -1.392 .81 .701
E 1.44 1.498 3 .89 2.604*
0 .47 .409 3 .92
it
2.470
F 1.82 1.412 2 .70 1.876
T -2 .24 -2 .090* -2.49 -2.078*
P 1.56 1.504 3.16
**
2.854
M .38 .448 .65 .628
*
Significant at p. = .05
**
Significant at p. = .01
69
TABLE 17
PROFILES OF MEAN T-SCORES, BASED ON GZTS NORMS (1949),
FOR INSTRUCTORS AND CRITERION GROUPS
Mean T-Scores
Group
G R A S E O F T P M
Norm, GZTS
Manual (N=389) 50 50 50 50 50 50 50 50 50 50
Instructors
(N=14) 56 61 57 57 69 65 69 52 66 60
Experimenta1
Pass (N=61) 54 62 60 59 66 64 65 56 62 59
Experimenta1
Fail (N=17) 52 61 53 56 62 57 63 57 59 59
Control Pass
(N=5 9) 52 58 53 53 59 59 59 52 59 55
Control Fail
(N=18) 45 55 50 49 57 54 60 50 59 56
70
experimental condition and then under the control condition
with the fail group, this relationship can be evaluated.
The validities just mentioned were presented in Table 7.
The findings of this comparison between the traits the in
structors value highly and/or ascribe to their own person
ality configuration, and the prediction level of these
traits, are that for the experimental condition 3 of the 4
traits deemed important for the instructors are also signifi
cant predictors, whereas, for the control condition this
occurs for only one of the traits. The 3 under the experi
mental condition are E, 0 and P, and for the control condi
tion, trait O. Although this difference between 3 and one
is not tested for significance it seems large enough to be
significant, or at least important.
A comparison between the mean scores of the in
structors and the pass group, and the instructors and the
fail group, under the control condition and then under the
experimental condition reveals the similarity between the
instructors and the criterion groups on the traits, under
the two test conditions. Tables 15 and 16 present these
data in terms of t tests of the significance of the dif
ference between means. For this analysis the differences
which were not found to be significant are considered.
Similarity is defined as differences which were not found
to be significant. Comparing the control condition and the
experimental condition it is observed that for the control
71
condition there is similarity between the instructors and
the pass group on 5 of the traits and between the instruc
tors and the fail group on 4 of the traits. For the experi
mental condition however, there is similarity between the
instructors and the pass on 9 of the 10 traits and between
the instructors and the fail on only 6. This appears to be
additional evidence that the experimental condition enhances
the correspondence between the personality variables self
ascribed to the instructors and the pass condition on the
criterion. Hypothesis II appears to be supported by the
above analysis.
Another method of analyzing the relationship between
the instructors and the experimental conditions of the tests
is through profile analysis. The standard scores reported
in Table 17 are listed in such a way as to be viewed as a
profile. The norms for the test in the first row of the
table are meant as a reference point, indicating that the
mean T-scores are 50 for each factor. If one attempts to
visualize lines connecting the different factors along each
row of the table it is possible to compare the profiles. In
this analysis parings of groups to be compared are: the
instructors with the experimental pass, the instructors with
the experimental fail, the instructors with the control
pass, and the instructors with the control fail. What is
desired is the comparison between the instructors and each
of the 4 condition-criterion groups. It is desired to
measure the similarity or difference between these profiles.
For the profile analysis a formula adopted from Guilford
(1956, p. 220) for the t tests of the difference between
pairs of means was used. The pairs being the values for
each factor for the two groups being compared. In Table 18
t tests for the significance of the mean difference are
presented for these comparisons. For all the comparisons
except that between the instructors and the experimental
pass group a significant difference was found, implying that
there was similarity between the instructors and the experi
mental pass group and difference between the instructors and
each of the other groups. It is noted that there is a
definite trend in the extent of the difference with the
largest difference being for the fail group under the con
trol condition, the next largest for the pass group under
the control condition and then a smaller difference for the
experimental condition in the fail group. It is not clear
why the fail group under the experimental condition should
be more similar to the instructors than the pass group under
the control condition, however, it must be realized that
what is noted here is a trend and that the difference
between successive values along this trend cannot be tested
for significance. What is known, however, is that only the
pass group under the experimental condition had a similar
elevation of profile to the instructors. This may mean that
under the set to project the concept of the occupational
73
TABLE 18
COMPARISON OF PROFILES OF T-SCORES FOR THE 10 GZTS
FACTORS BETWEEN INSTRUCTORS GROUP AND EACH
OF THE CRITERION GROUPS
Comparison
Groups
Mean
Difference t
Instructors -
Experimental Pass .05 0.55
Instructors -
Experimental Fail 3.3
*
2.57
Instructors -
Control Pass 5.3
**
5 .40
Instructors -
Control Fail 7.7
**
7.59
*
Significant at p. = .05 for 9 df
ic*k
Significant at p. = .01 for 9 df
74
image of the nurse those who were successful as students,
that is, passed on the criterion, were able to approximate
the projection of this concept whereas, those who failed on
the criterion were not able to do this. The method of this
profile analysis may only be approximately a correct statis
tical application. A more appropriate profile analysis
might have been that reported by Osgood and Suci (1952) and
also reported by Cronbach and Gleser (1953), the D, distance
or dissimilarity measure. Since Osgood and Suci (1952)
report that a significance of the difference between D's has
not been adequately developed as yet, it would have little
usefulness in indicating if there is a significant dif
ference between the various groupings. Cronbach and Gleser
(1953) report that a special form of the D, that of
Mahalanobis, of the Calcutta school, has a known distribu
tion function and thus a basis for testing the significance
of the difference between groups, but this distribution was
not available. This analysis presented additional support
for Hypothesis II.
The popular scores which were derived in the study,
which represent the modal item response pattern for the
several experimental conditions, both for the students and
for the instructors, were evaluated by correlational method
to determine the predictive validity. Tables 19 and 20
present the popular score means and standard deviations for
the GZTS and DFOS, for the various groups and conditions.
75
TABLE 19
GZTS POPULAR SCORE MEANS AND STANDARD DEVIATIONS
N Mean S .D.
Experimental Group
Student1s Key
Pass 61 239 23.94
Fail 17 220 31.98
Dropout 15 227 67.90
Instructor's Key
Pass 61 222 25 .86
Fail 17 203 31.55
Dropout 15 229 26.70
Control Group
Student1s Key
Pass 59 206 24.22
Fail 18 199 23.15
Dropout 16 2 02 25 .04
Instructor's Key
Pass 59 196 13.15
Fail 18 185 31.62
Dropout 16 191 30.87
76
TABLE 20
DFOS POPULAR SCORE MEANS AND STANDARD DEVIATIONS
N Mean S .D.
Experimental Group
Student's Key
Pass 61 216 26.19
Fail 17 200 27.85
Dropout 15 208 63.34
Instructor's Key
Pass 61 186 22.71
Fail 17 167 20.76
Dropout 15 181 20.42
Control Group
Student's Key
Pass 59 192 18.87
Fail 18 191 17.49
Dropout 16 190 15.43
Instructor's Key
Pass 59 140 24.96
Fail 18 140 18.76
Dropout 16 133 20.07
77
Each subject was tested on two keys, both the self-group key
and the instructor's key. The students under each condition
had the key which was derived under that condition, and also
the instructor-derived key.
The validities are presented in Tables 21 and 22 for
the two inventories respectively. On the GZTS, 3 of the 8
validities were found to be significant. They all dealt
with the fail criterion group. The dropout criterion group
was not predicted in any of the tests of this key. Under
the experimental condition both the student's key and the
instructor's key predicted fail groups. Under the control
condition only the instructor's key predicted the fail
group. The validities found with DFOS, Table 22, again
concur with the prediction of the fail criterion, but only
under the experimental conditions. The two significant
correlations found were for student's key and the instruc
tor's key, both predicting the fail criterion. On none of
these popular score correlational tests is the dropout
criterion predicted. Inspection of the mean scores,
Tables 19 and 20, shows that in most of the instances the
dropout means are quite close to the pass group means.
When the biserial correlation is computed the difference
between the means largely determines the size of the corre
lation and these mean popular scores for the dropouts were
always closer to the pass group than was the fail group.
Regarding the hypothesis to be tested in this experiment,
78
TABLE 21
GZTS POPULAR SCORE VALIDITIES (r bis)
Condition Key Criterion N r bis
Control Students Fail 77 .169
Control Students Dropout 75 .092
Control Instructors Fail 77
*
.248
Control Instructors Dropout 75 .202
Experimental Students Fail 78
**
.394
Experimental Students Dropout 76 .185
Experimental Instructors Fail 78
**
.404
Experimental Instructors Dropout 76 -.139
*
Significant at p. = .05
**
Significant at p. = .01
79
TABLE 22
DFOS1 POPULAR SCORE VALIDITIES (r bis)
Condition Key Criterion N r bis
Control Students Fail 77 . 003
Control Students Dropout 75 . 075
Control Instructors Fail 77 .000
Control Instructors Dropout 75 .169
Experimental Students Fail 78
**
. 353
Experimental Students Dropout 76 .127
Experimental Instructors Fail 78
**
.484
Experimenta1 Instructors Dropout 76 .140
ieib
Significant at p. = .01
80
Hypothesis III, under the experimental condition for both
inventories the fail groups were always predicted. For the
control condition only one out of the 4 tests for prediction
of the fail criterion was found significant. This supports
the hypothesis that the popular score would be more predic
tive of the criterion under the experimental conditions.
The scoring key which was made up of the items which
Beaver (1955) found significantly differentiated student
nurses from other college students was used in this study.
Table 23 presents means, standard deviations and validity
coefficients for the several condition and criterion groups,
and also t tests of the differences between means, where
run. When the total groups, comprising the pass plus the
non-pass students, were compared for the control and experi
mental conditions there was a significant elevation found
in the R.N. key scores. The experimental instructions
raised the scores significantly. For the control group when
the subjects who passed were compared with all of those who
did not pass, a significant difference was found between the
means on the R.N. key, however, under the experimental con
dition no significant difference was found between the pass
and the total non-pass group. The hypothesis, which
predicts that there will be higher validity under the
experimental condition, is partially borne out by the fact
that the validity coefficients are significant for the
experimental condition predicting the fail group; however,
TABLE 23
SUMMARY OF BEAVER R.N. KEY STATISTICS FOR CONDITION AND CRITERION
GROUPS AND t TESTS OF DIFFERENCES BETWEEN MEANS
N Mean S ,D. t r bis
Control, Total 93 22.34 4.52)
•J* ^
)
7.16
Experimental, Total 93 26.67 3.69)
Control, Pass 59 23.07 4.44)
* *
) 2.07 .270
Control, Non-pass 34 21.09 4.46)
Experimental, Pass 61 26.77 3.66)
)
0.37 .050
Experimental, Non-pass 32 26.47 3.80)
Control, Fail 18 21.72 4.28
Control, Pass and fail 77 .183
Experimental, Fail 17 25 .23 3.89
■ J U
Experimental, Pass and fail 78 .231
Contro1, Dropout 16 20.38 4.52
*■.
Control, Pass and dropout 75 .323
Experimental, Dropout 15 27.87 3.74
Experimental, Pass and dropout
- - - - - - - - - - - - S r . ? - - J ' j d l ' _ _ ' - - - - - - - - - - - - - =--7 \
76
c
-.104
*Significant at p. = .05
**Significant at p. - .01
82
the dropout group is better predicted under the control
condition. There is not sufficient evidence to support
Hypothesis IV, since no improvement in prediction occurs
under the experimental condition over that found under the
control condition.
When the instructions are changed as occurred in
this experiment, the tendency to respond in a socially
favorable way is enhanced. The falsification scales of
Jacobs and Schlaff (1955) were applied to the GZTS scores to
determine the extent and direction of this effect. It was
hypothesized that the falsification scale scores would be
higher under the experimental condition. It is also pos
sible to obtain information concerning whether faking good,
or in a socially desirable direction, is the only effect
that the instructions caused. Table 24 presents data on the
falsification scale comparisons between the control and
experimental conditions and also presents the standardiza
tion norm group means, as published in the scale manual
(Jacobs and Schlaff, 1955). Under the experimental condi
tions both the gross falsification and the subtle falsifica
tion scales were significantly elevated over the control
condition. These differences were found to be great and
highly significant. There is no doubt that the instructions
affected the test-taking attitude in the direction of
falsification. The carelessness-deviancy mean score was
lower under the experimental condition than the control
TABLE 24
FALSIFICATION SCALE COMPARISONS BETWEEN CONTROL AND EXPERIMENTAL CONDITIONS
AND STANDARDIZATION NORM GROUP FOR GROSS FALSIFICATION (GF),
SUBTLE FALSIFICATION (SF) AND CARELESSNESS-DEVIANCY (CD)
Control
(N=93)
Experimenta1
(N=93)
Mean
Difference
Norm Group^
(N=54)
Standard Faking
t Mean Mean Scale Mean S.D. Range Mean S.D. Range (E-C)
GF 17.40 6.68 4-32 24.26 5.45 11-33 6.86 7.673** 15.09 29.21
SF 10.08 4.26 2-17 13.11 2.73 4-17 3.03 5.771** 8.46 14.20
CD 2.37 2.35 0-13 1.67 1.59 0-9 -0.70 -1.47 3.40
■ ' ’ Based on data from Jacobs and Schlaff (1955) .
Significant at p. = .01
84
condition, but not significantly so. The same effect is
occurring here as for the other two scales; the effect of
movement in the direction of more favorability, that is,
less deviant impressions.
By comparing the mean scores obtained in this
research with those obtained in the standardization research,
it is possible to determine whether faking is the entire
effect. The scale manual reports the means for the standard
conditions and for the conditions when the subjects were
told to give the "best possible impression" of themselves
that they could possibly give. In the present study, for
GF the control group is higher than the standardization
group under the control conditions, while the experimental
group in the present study is lower than the faking group in
the standardization study. The difference between means
between the control and experimental conditions of this
study is quite a bit smaller than the difference between the
means in the faking versus standard conditions comparison of
the standardization study. This means that the control
group in the present study is more motivated for positive
impression than the standardization group, possibly because
in the present study the subjects signed their names to the
papers and were informed that the schools would have
knowledge of their test scores. This indicates that under
the control conditions a certain amount of bias in the
direction of social favorability was operating. It was to
85
study biases of this nature that this experiment was
conducted. The fact that the experimental instructions
elicited a mean score on the GF scale somewhat lower than
the mean for the fake good group in the standardization
sample indicates that the experimental subjects were not
merely faking good. The intent of the instructions was to
direct them to project the occupational goal-ideal, and
although social favorability is undoubtedly inherent in the
goal-ideal image, it is not only social favorability. A
similar analysis for the SF, the subtle measure of falsifi
cation, shows the same relationships occurring. The control
group is higher than the standardization group and the
experimental group is lower than the faking group, but in
this instance these differences are considerably smaller.
This again indicates the higher motivation of the control
group than the standardization group and the lower tendency
of the experimental group to merely fake a good impression.
It might also be that the experimental instructions in this
study were interpreted by the subjects as a suggestion
toward subtle faking rather than gross faking. Hypothesis V,
is well supported by these data.
In addition, it is possible to view this support as
a form of cross validation or substantiation of the falsifi
cation scales. The scales measured an actual tendency
toward falsification and both scales performed as expected,
based on the claims in the scale manual. The item analysis,
86
which is next reported, was surveyed regarding the 34 GF
items to see how they functioned. Thirty-one of the 34
were found to he significant. A significant difference
between the proportions selecting the item by the experi
mental versus the control conditions was the measure of
significance.
An analysis of the 300 items of the GZTS was em
ployed to determine the effect of the experimental instruc
tions on the direction of response of the items. It was
hypothesized that significant item differences would occur
with greater than chance expectancy. Statistical tests of
the significance of the difference between the proportions
of each group which answered the item in the keyed direction
was computed for each of the 300 items, once for each of
3 groups. The groups are: subjects who passed the
criterion, subjects who did not pass the criterion, which
includes both the fails and the dropouts, and the total
grouping of all the subjects under each condition.
Significant differences did occur on 150 of the 300 items.
Table 25 lists the number of items which were found signifi
cant for each of the factors on the GZTS, separately derived
for the groupings.
When large number of significance tests are com
puted, as in this instance, it can be assumed that some of
them will attain significance by chance probability.
Sakoda, Burton and Cohen (1954) have published a method
87
TABLE 25
ITEMS PER FACTOR, ANSWERED BY SIGNIFICANTLY DIFFERENT
PROPORTIONS OF THE CONTROL AND EXPERIMENTAL GROUPS,
BY CRITERION GROUPS; PASS, NON-PASS AND TOTAL
AND BY ANY OF THE GROUPS
Number of Items Out of 30
Factor
Pass Non-Pass-*- Total
Pass
Non-Pass
or Total
G 9 8 14 15
R 10 4 16 17
A 13 8 13 14
S 11 7 20 20
E 13 7 21 21
O 12 7 16 16
F 6 5 13 13
T 6 9 12 12
P 3 2 8 9
M 11 3 9 13
Sums 94 60 142 150
^"Non-pass group is the combined fail and dropout
groups.
88
which yields the chance probability of obtaining signifi
cance at the desired confidence level. At the .05 confi
dence level, adopted for the item statistics in the present
study, 25 out of 300 could occur by chance only one time in
100. This indicates that the number of items found signifi
cant, ranging from 60 to 142 for different groups, is very
significant. Thus, Hypothesis VI is supported.
This analysis concerns how the experimental in
structions have influenced the test items in a manner inde
pendent of the factor variance inherent in the item. It is
obvious, in consideration of the fact that 15 0 out of 300 of
the items were found to be thusly influenced; that some
drastic changes occurred in the test, when taken under ex
perimental instructions. Jacobs and Schlaff (1955), in
their development of falsification scales for this test,
found only 47 items for which there was a significant dif
ference between proportions responding in the keyed
direction between standard and faking instructions. However,
they used more stringent standards for the difference
between proportions, demanding that the difference between
the two groups be at least 40 per cent of the groups. In
the present study merely finding a difference between the
proportions which is significant at the „05 level was
involved.
Appendix D presents a tabulation of the items which
were found to have significant differences between the way
89
they were answered under the control condition and the
experimental condition. These are tabulated separately for
each of the 10 factors.
Table 25 reveals that certain of the factors seemed
more susceptible to the influence of the instructions than
were other factors. The number, out of the 30 possible in
each factor, varied from a low of 9 for trait P to 21 for
trait E. It may be that these differences in susceptibility
to change due to the instructions is related to how the
student nurses view these traits, in terms of being thought
important traits for a good nurse to have.
Table 26 presents the 22 items which were answered
in the opposite direction to the way they are scored in the
test manual by the majority of the experimental group.
Answered in this way they detract from rather than add to
the trait scores. These items present additional evidence
that the experimental instructions were not merely perceived
as instructions to "fake good" but have some more selective
influence.
Cross Validation of Beaver's R.N. Key Items
Of the 97 item responses for which Beaver (1955)
reports significant differences the 42 items which differen
tiated the nursing students from the education majors at the
.05 level of probability were utilized as the "R.N. key" for
this analysis. Of these 42 items, Beaver found that 16 of
90
TABLE 26
ITEMS PER FACTOR, ANSWERED BY SIGNIFICANTLY DIFFERENT
PROPORTIONS OF THE CONTROL AND EXPERIMENTAL GROUPS
AND ANSWERED BY OVER 50 PER CENT OF THE
EXPERIMENTAL GROUP IN THE
NON-KEYED DIRECTION
Factor
G R A S E O F T P M Sum
Number 5 4 1 2 2 1 1 3 0 3 22
91
them were significant indicators when keyed in the opposite
direction from the way they are keyed in the Guilford-
Zimmerman scoring manual. In the present study they were
keyed according to Beaver. Only the control group, under
standard instructions, was analyzed.
This cross validation of Beaver's items utilized the
statistical technique of analyzing the significance of the
difference between uncorrelated proportions (Guilford, 1956,
p. 221). The difference analyzed was that difference
between the proportion of passing student nurses who scored
on the item and the proportion of non-passing student nurses
who scored on the item. The two sub-groups thus used are
control pass and control non-pass.
Care was taken in this analysis to account and cor
rect for situations in which the proportions were very
extreme. Since the sampling distribution of critical ratios
obtained from this method approaches normality closely
enough to allow use of the normal curve distribution, pro
vided that the smallest product of the weighted mean of p or
q times or N2 is not less than 10, a correction for dis
continuity was necessary in those cases where this product
was between 5 and 10. Where the product was less than 5, no
test of significance was made on those items. A pooled
estimate, or weighted mean, of the proportions was used in
the formulas.
92
The analysis revealed that only 10 of the 42 items
were found to be significant at the probability level of .05
or better. When the corrections for continuity were made,
the probability level of one of these 10 items fell slightly
above .05 and so was not considered significant. Of the
remaining significant items, 2 of them were found to be
significant in the reverse direction than anticipated. For
these 2 items, the non-pass group had a significantly higher
proportion of subjects scoring than the pass group. For the
other 7 items, the pass group had significantly higher pro
portions than the non-pass group.
Table 27 presents a list of the 10 items, whether
keyed "yes" or "no, " the value of the critical ratio, and
the probability level. These results indicate that the
items substantially did not hold up in the cross validation.
It may be questioned whether this is an adequate method of
cross validating Beaver's items because this analysis was on
groups somewhat different from her groups. She found items
which significantly differentiated student nurses from
education majors. The present study deals only with student
nurses and compares those who were successful students with
those who failed. It logically may be assumed that most of
the items which would differentiate student nurses from non
student nurses would also differentiate good students from
failures in nursing training, because the failures ceased
being student nurses. Since this test administration was
93
TABLE 27
ITEMS OF THE BEAVER R.N.KEY FOUND TO DIFFERENTIATE
PASS FROM NON-PASS GROUPS UNDER
CONTROL CONDITION
Item
Number
Keyed
Response
Critical
Ratio
P
Level
3 No -1.964 corrected .05
107 No -3 .059 .01
*
151 No -1.916 corrected .055
185 Yes 2.065 .05
210 Yes 2.036 corrected .05
230 Yes 2.010 .05
231 Yes 2 .104 corrected •
o
U1
253 No 3 .408
« —1
o
282 No 2.041 corrected
in
o
•
286 Yes 2.756 corrected .01
*
Not significant
based on standard instructions it should stand as an ade
quate test of her items. There is at least one item whose
content would certainly not lead one to expect that poor
student nurses would answer them differently from good
student nurses. Reference is made, for example, to
item 260: "You would rather be a building contractor than
a nurse." Obviously, even the student nurses who failed
answered this "no," whereas Beaver found a significantly
greater proportion of the education majors answered "yes."
CHAPTER V
DISCUSSION
This study which concerns the effect or influence
upon the predictive power of factored personality and
interest inventories that may be due to control or changes
of the personal response set of the subject, is to be viewed
as concerning employment and selection testing situations.
Although in the present study student nurses who were
already selected were used as subjects, their testing
experience was comparable to the selection condition. The
tests were given in the first weeks of their educational
program. The students were quite aware of the rate of
attrition in nursing education and perceived themselves as
being on trial. The students signed their names to the test
papers and at the time of taking the test they were informed
that the schools would be given the results of the test.
There is good reason to believe, therefore, that the moti
vation of the students in this experiment was comparable to
that found in a selection situation.
This study is concerned with improving the predic
tion of already established personality inventories. The
principal hypothesis is that the experimental instructions,
designed to orient the subject in the direction of
95
96
projecting the response pattern of a highly valued occupa
tional ideal, will improve the predictive ability of the
test. This hypothesis was confirmed; prediction was shown
to be improved under the experimental instructions condition
over the prediction obtained under standard instructions.
Specifically, for the GZTS, prediction of the dropouts was
substantially improved, and for the DFOS, prediction of the
fails was substantially improved.
On the GZTS the dropout criterion was best predicted
under the experimental condition, but the fail criterion was
predicted about equally well between both test conditions.
The evidence for this increased prediction was: under the
experimental condition, 7 out of the 10 of the validity cor
relations were found to be significant, whereas, for the
control condition only one out of the 10 was significant in
the prediction of dropouts. In the critical ratio test of
the significance of the difference between the r's under the
experimental and control conditions for the dropouts 8 of
the 10 critical ratios were significant and 6 of those 8 had
the experimental correlations higher than the control
correlations. The multiple R that was computed for the
dropout criterion and the 10 factors was higher for the
experimental group than for the control group. However, for
the fail condition the evidence is all fairly non
significant. The validity correlations found significant
were of equal number for the control and experimental
97
conditions. The critical ratios between the experimental
and control condition for the validity r's yielded only one
which was significantly higher for the experimental
condition. The multiple R's were equal between the experi
mental and the control condition. Thus, there is clear
evidence that the dropout group was better predicted under
the experimental condition.
The analysis of the DFOS revealed that the fail
condition was best predicted under the experimental condi
tion, although the evidence for improved prediction is less
than that found for the GZTS. The dropouts were predicted
about equally well by the experimental and the control con
dition on the DFOS. For the fail group the evidence for its
superior prediction is as follows: the number of signifi
cant validity r's were 5 out of the 10 for the experimental
condition and 0 out of 10 for the control condition; the
multiple R was higher for the experimental condition than
the control condition, whereas, for the dropouts it was
equal for experimental and control; the critical ratio test
for the significance of the difference between the r's under
the experimental and control conditions found 2 out of the
10 significant and for both of these 2, the r under the ex
perimental condition was higher than the control. For the
dropout criterion prediction about equal numbers of signifi
cant r's were found, 2 for the experimental group and 3 for
the control, the multiple R's were approximately equal for
98
experimental and control, and the critical ratio test of the
significance between the r's yielded 4 out of 10 signifi
cant, but 3 out of these 4 were higher for the control con
dition than the experimental condition. This last statistic
is the only instance in the analysis of the first hypothesis
where the control condition had any superiority over the
experimental condition. Thus, it is observed that the fail
criterion was best predicted for the DFOS under the experi
mental conditions, although the evidence is not as weighty
as that reported for the GZTS. There is no single signifi
cance test which can be applied to these hypotheses, but
Chi square tests of independence were run between the
experimental and control conditions, for the number of
significant validity r's. For the GZTS dropouts, Chi square
is 7.5 (p = .01) and for DFOS fails, Chi square is 6.66
(p = .01), while for GZTS fails and DFOS dropouts the Chi
squares are near zero and not significant.
It is difficult to explain the differences in pre
diction which occurred between the two inventories. Why did
the personality inventory improve prediction for the drop
outs while the interest inventory improved prediction for
the fails? This may be due to peculiarities of the cri
terion when applied to student nurses. The criterion
certainly is not clear-cut, and as described above there
appeared distinct differences between the two groups who
did not complete the course. Changes have certainly
99
occurred in nursing education over the years, and now that
nurses are mostly trained in colleges rather than hospital
schools, the dropout may represent something quite different
than it has in the past. Students who drop out of the
nursing program could very well be dropping out because they
are capable and interested in more demanding courses of
training and in occupations which require abilities beyond
that of a nurse. It is known that some of the student sub
jects who dropped out did so to take four-year programs in
the college in other sciences. Although education was con
trolled for in this study, the intelligence above that
needed for college entrance was not controlled. Herzberg
(1954) found, in an actual employment selection and promo
tion situation, that "a comparison on the basis of education
shows higher mean scores for each scale on the GZTS with in
creasing education from grammar school through high school
to college graduation." His interpretation was that "faking
good adjustment" is related to education level. It might be
that higher levels of intelligence are related to the
ability to project the personality image when taking a test
under instructions such as those used in this study.
Regarding the DFOS results the fail group was best
predicted. It would be expected that an interest test would
better predict dropouts than fails, since dropouts have
voluntarily "changed their mind." The multiple R's under
the control condition reveal that when dropouts are compared
100
with the fail group, the multiple R's for the dropout
regression is .410 and for the fail it is .300. Therefore,
the DFOS does predict the dropouts better than the fails
under the control condition. This was not the type of com
parison that was the basis of this study though, the com
parisons were between experimental and control. It may be
that the prediction was improved for the fails under the
experimental instructions because those who failed do not
have inherent nursing interest patterns and therefore, were
unable to project that pattern under the experimental
conditions.
Hypothesis II is concerned with the relationship
that exists between the personality factors which the in
structors ascribe to themselves and the extent that those
factors are predictive of the criterion. It was hypothe
sized that positive relationships exist, and that they are
enhanced or pronounced under the experimental condition.
This hypothesis has been confirmed by the findings of three
analyses. Relating the four highest ranking traits for the
instructors with the predictive validity of the traits for
the students revealed that under the experimental condi
tion 3 of the 4 traits highest for the instructors were
found to be significant predictors of the fail criterion;
whereas, under the control condition only one of these 4
was found significant. This difference between 3 and one is
101
considered important if not significant, and confirmatory of
the hypothesis. The significance of the difference tests
between the means of the instructors and the criterion
groups under the experimental and control conditions, re
vealed that the experimental pass group was similar in 9 of
the 10 factors, whereas, for the control condition this was
true only for 5 of the factors. The experimental fail group
was similar to the instructors on 6 traits and the control,
similar on 4. This confirms the correspondence between
the personality variables self-ascribed to the instructors
and the pass condition on the criterion. A profile
analysis lends further support to the hypothesis; signifi
cant differences were found between the instructors and all
the condition-criterion groups except the experimental
pass group. These findings lend support to the impression
that instructors fail those students who seem to them to
not fit their concept of a potential nurse, possibly based
on values and characteristics they are unaware of. The
differences found between the experimental pass group and
the other groups in terms of being less like the instructors
is supportive of this inference. The underlying assumption
of this experiment, regarding the operation of this effect,
is that under the experimental instructions the students
are projecting a picture of their personality makeup which
apparently is perceived by the instructors throughout
the first year, during which the decisions are made
102
about the student's position on the pass-fail criterion.
Hypothesis III deals with the popular score which
was derived in this study based on the modal responses for
each group, on the total test. It was hypothesized that
these popular scores would predict the criteria better under
the experimental condition than under the control condition.
Each student obtained 2 popular scores, one based on his own
group response pattern and the other based on the instruc
tors' responses. This hypothesis was confirmed by the
superiority of the prediction under the experimental condi
tion for both inventories. Under the experimental condition
based on significant validity correlations, the fail groups
were predicted 4 times out of 4, while under the control
condition they were predicted once in 4 attempts.
Prediction of the dropout criteria was equivocal, there
being no significant r's under either of the conditions.
These findings indicate that the homogeniety of an
occupational or training group, based upon personality
variables, may be a good predictor of the validity of mem
bership in that group, when personality- test instructions
are changed as in this study. Without any reference to an
outside criterion, the modal group response can be an
effective predictor.
The response key for the GZTS which was derived by
Beaver (1955) was hypothesized to be a more effective pre
dictor of nursing success under the experimental
103
instructions than under standard instructions. The findings
of the present study do not support this hypothesis
(Hypothesis IV). Although the validity correlation for the
fail group was significant under the experimental condition
and not under the control; the control condition yielded
significant validity for the dropout criterion, while the
experimental condition did not. In addition, the control
condition was found to yield better prediction of the cri
terion when the total non-pass groups were evaluated. Thus,
the experimental instructions have not improved the
efficiency of the R.N. key. A cross validation of the items
was done, in this study, and the evidence does not support
the R.N. key as a valid measure.
The effect of the instructions upon response bias
was studied in this research by application of the falsifi
cation scales for the GZTS (Jacobs and Schlaff, 1955).
Hypothesis V predicts that the subject's falsification
scores would increase significantly under the experimental
conditions. This hypothesis was well supported by the data.
However, it was also found that the falsification scale data
supports the impression that other effects besides falsifi
cation occurred. The highest mean falsification values
found in this study were well below those found for the
faking condition in the standardization group. Also, the
control condition in the present study yielded falsification
mean values higher than the standard instructions of the
104
standardization group. These effects indicate that faking
good or social favorability, although present, are not the
main influences operating. The social favorability present
in the image of the occupational goal-ideal has had some
influence, but in addition the projection of the personality
attributes of successful nurses has been influential, as the
support for the major hypotheses of this study, the in
creased prediction under the experimental instructions,
indicated.
The performance of the falsification scales in this
study lends support to their use to accurately detect social
favorability falsification. This study tends to cross
validate the original research. The distinction between the
2 falsification scales was evident, and both scales operated
as claimed in the manual.
Hypothesis VI, regarding the influence on the GZTS
items, was confirmed by the significantly greater than
chance occurrence of response differences that were signifi
cant between the 2 test conditions. The instructions
changed the behavior of half the items, but all of the
factors were not equally susceptible to influence. It
appears likely that a new factor analysis, with the experi
mental type of instructions, would reveal changes in the
factor variance. It is probable that new factors would
emerge of the nature of "social perception" or "role-playing
ability." A larger sample than that used in the present
105
research would be necessary, particularly when the presence
of new factors is anticipated. When new factors are dis
covered, they should materially aid prediction when incorpo
rated into multiple regression equations. If the new
factors concern social perception or role-playing ability as
presumed, then they should be non-specific regarding any
particular occupational group. The factors which would
emerge from the groups used in this study, should then be
applicable to other selection problems.
The present research is a methodological study. It
concerns alterations of the formal aspects of testing for
selection purposes. The primary aim is to test the hypothe
sis that certain types of error variance could be reduced by
structuring the test situation. The crucial test of the
hypothesis is the extent and direction of changes in
prediction. Consequently, this study has not concerned
itself with interpretation of the test variables. The test
content has been considered primarily to the extent of
studying the literature to determine that the tests used
are appropriate instruments for prediction of the criterion.
Their acceptance and relevance for selection of nursing
students has been satisfactorily demonstrated elsewhere.
The influence of the experimental instructions of
the present study must be viewed in the framework of
response set and projection. It is believed that response
set in the formal sense as described by Cronbach (1946) was
not involved. Nor does the debate about test content and
response style as conducted by Jackson and Messick (1958)
correctly apply to the influences thought to operate here.
Rather, the explanation of Sherif and Cantril (1946) that
deals with less formal but more intrinsic aspects of
response predisposition seems appropriate. Their view of
response tendencies as frames of reference, conceives of
each person having his own reference point, yet being sub
ject to external influences, consciously or unconsciously,
which alter the reference frame. In the absence of struc
ture, as when the adventitious personal response sets that
the respondent brings to the test situation are left free to
act, the individual's personal reference frame influences
perception. When the situation is subtly structured, as in
this study, these frames of reference appear to be subject
to the influence of the instructions. Guilford (1954) men
tions the influence of added structure upon response
tendencies, and in the DFOS manual, Guilford, Christensen,
and Bond (1956) describe their use of projective question
naire items. Both of these influences are thought to bear
on the changes wrought by the instructions to respond as
the occupational goal-ideal.
It appears that industrial psychology might con
sider two findings of the present research as fruitful. The
confirmation of the major hypothesis leads to the conclusion
that personality and interest questionnaires can be made
107
more predictive by using special instructions designed to
aid projection of an occupational image and to reduce adven
titious response sets. The second value appears to be the
identification of the coincidence of the relationship of the
personality attributes of the supervisor, or in this case
instructor, to the predictive ability of those same person
ality variables among the ratees on the success-failure
continuum. The traits most highly self-ascribed to the
instructors are those which best separate the students who
pass on the criterion from those who do not pass.
CHAPTER VI
SUMMARY
The present study was aimed at the problem involved
in the use of self-report personality inventories, particu
larly in selection situations. It has been long recognized
that respondents are able to falsify their test records
when motivated by some reward, and under experiments when
asked to "fake good." However, the tendency to falsify
under real life motivation is personal and varies from
person to person. It is also known that these same inven
tories have been found quite valuable when used in coun
seling situations, in which the respondent is concerned with
having an accurate self-appraisal. The approach of the
present research was to change the test instructions in such
a way as to influence the respondents so that all would
simulate their concept of the image of the occupational
ideal person. The subjects, student nurses at the beginning
of training, were told to answer "the way you think a compe
tent graduate R.N. who likes her work would answer."
It was hypothesized that these instructions would
reduce the error variance due to the various personal
response tendencies that the subjects bring to the testing,
108
109
and consequently improve the prediction of academic success.
Response bias was thought to be focused in the direction of
the occupational image for all respondents similarly and the
tendency toward projection was thought to be enhanced,
reducing the attention toward self-reporting.
The main hypothesis, Hypothesis X, of the experiment
was confirmed: that the tests given under the occupa
tionally oriented instructions would yield scores which
would increase prediction of the academic criterion.
Specifically, for the GZTS, prediction of the dropouts was
substantially improved over that under standard conditions,
and for the DFOS, prediction of the fails was substantially
improved. Evidence was based upon: significant validity
coefficients; critical ratio tests of the difference between
experimental and control r's; and comparison of mulitple R's
between experimental and control conditions.
Hypothesis II, that the experimental instructions
would enhance the relationship that exists between the
temperament factors that the instructors ascribe to them
selves and the extent that those factors are predictive of
the criterion was confirmed. Evidence was based upon:
better prediction under the experimental condition found for
the traits which ranked highest for the instructors;
significance tests of the means between instructors and
students revealed greater similarity of pass groups to
instructors under the experimental condition; and profile
110
analysis, which reaffirmed the greater similarity of
experimental-pass to the instructors, than the other
criterion-condition groups.
Hypothesis III, that the popular scores, derived in
this study, would better predict the criterion under the
experimental conditions was confirmed. For both inventories
the fail criterion was significantly predicted under the
experimental condition. In only one instance of four, the
control condition had effective prediction.
Hypothesis IV, that Beaver's R.N. key for the GZTS
would be more predictive under the experimental condition,
was not supported. Significant validity r's occurred
equally under both conditions. Item analysis of this key,
for cross validation purposes, revealed that the key was not
found to be a valid measure.
Hypothesis V, that the Falsification Scales of
Jacobs and Schlaff, would increase significantly under the
experimental conditions, was confirmed by t tests of the
means. These scales also indicated that the effects of the
instructions were not merely to increase falsification, but
to create other influences. This study may also be seen as
further validating the Falsification Scales.
Hypothesis VI, was confirmed by the finding of the
item analysis, that a much greater number of items than
could have occurred by chance were found to be answered
differently between the two conditions. This great change
Ill
in test item response suggests that perhaps new factors
would emerge under the new test instructions, such as,
social perception or role-playing ability.
As a methodological study, the concern has been with
altering the elements of the testing situation in order to
reduce error variance and enhance prediction. It may be
I
concluded that certain personality and interest question
naires can be made more predictive by using special instruc
tions designed to reduce adventitious response sets and to
aid identification with a valued occupational image. This
study has also indicated the strong relationship of the
personality attributes of the supervisor or instructor to
the predictive ability of these same attributes among those
rated on the variable of pass or fail.
REFERENCES
REFERENCES
Beaver, A. P. Temperament and nursing. Psychol. Reports,
1955, 1, 339-344.
Benton, A. I. The experimental validation of the Rorschach
Test. Brit. J. of Med. Psychol., 1950, 23, 45-58.
Berg, I. A. Response bias and personality; the deviation
hypothesis. J. Psychol., 1955, 40, 61-72.
Berg, I. A. The unimportance of test item content, in Bass,
B. M. and Berg, I. A. (eds.). Objective approaches to
personality assessment. Princeton: Van Nostrand, 1959.
Cronbach, L. J. Response sets and test validity. Educ.
Psychol. Measmt., 1946, 6, 475-494.
Cronbach, L. J. Further evidence on response sets and test
design. Educ. Psychol. Measmt., 1959, 10, 3-31.
Cronbach, L. J. and Gleser, G. C. Assessing similarity
between profiles. Psychol. Bull,, 1953, 50, 456-473.
Edwards, A. L. Manual, Edwards Personal Preference
Schedule. New York: Psychological Corp., 1954.
Gordon, L. V. Validities of the forced-choice and question
naire methods of personality measurement. J. appl.
Psychol., 1951, 35, 407-412.
Green, R. F. Does a selection situation induce testees to
bias the answers on interest and temperament tests?
Educ. Psychol. Measmt., 1951, 11, 503-515.
Guilford, J. P. Psychometric methods. New York: McGraw-
Hill, 1954.
Guilford, J. P. Fundamental statistic in psychology and
education. New York: McGraw-Hill, 1956.
Guilford, J. P., Christensen, P. R., and Bond, N. A., Jr.
The D. F. opinion survey, manual of instructions and
interpretations. Beverly Hills: Sheridan Supply Co.,
1956.
113
114
Guilford, J. P., Christensen, P. R., Bond, N. A., Jr., and
Sutton, M. A. A factor analytic study of human
interests. Psychol. Monogr., 1954, 68, 4 (375).
Guilford, J. P. and Zimmerman, W. S. The Guilford-Zimmerman
temperament survey, manual of instructions and
interpretation. Beverly Hills: Sheridan Supply Co.,
1949.
Healy, I. and Borg, W. R. Personality characteristics of
nursing school students -and graduate nurses. J. appl.
Psychol., 1951, 35, 275-280.
Healy, I. and Borg, W. R. Personality and vocational
interests of successful and unsuccessful nursing school
freshmen. Educ. Psychol. Measmt., 1952, 12, 767-775.
Heineman, C. E. A forced choice form of the Taylor Anxiety
Scale. J. consult. Psychol., 1953, 17, 447-454.
Heron, A. The effects of real life motivation on question
naire response. J. appl. Psychol., 1956, 40, 65-68.
Herzberg, F. Temperament measures in industrial selection.
J. appl. Psychol., 1954, 38, 81-84.
Jackson, D. N. and Messick, S. Content and style in
personality assessment. Psychol. Bull., 1958, 55,
243-252.
Jacobs, A. and Schlaff, A. Falsification scales for the
Guilford-Zimmerman temperament survey. Beverly Hills:
Sheridan Supply Co., 1955.
Jarrett, H. A study of the validities of certain factor-
analytic personality measures administered to municipal
firemen under modified conditions. Unpublished Ph.D.
dissertation, University Southern Calif., 1957.
Jurgensen, C. E. Table for determining phi coefficients.
Psychometrika, 1947, 12, 17-2 9.
Kelly, E. L., Miles, C. C., and Terman, L. M. Ability to
influence one's score on a paper and pencil test of
personality. Character and Pers., 1936, 4, 206-215.
Meehl, P. E. and Hathaway, S. R. The K factor as a sup
pressor variable in the Minnesota Multiphasic
Personality Inventory. J. appl. Psychol., 1946, 30,
525-564.
115
Osgood, C. E. and Suci, G- A measure of relation determined
by both mean difference and profile information.
Psychol. Bull., 1952, 49, 251-262.
Ruch, F. L. A technique for detecting attempts to fake
performance on the self-inventory type of personality
test, in McNemar, Q. and Merrill, M. A. (eds) .
Studies in personality. New York: McGraw Hill, 1942.
Ruch, F. L. Ability of adults to fake desirable responses
on two personality self-inventories and an attempt to
develop a "lie detector" key. Psychol. Bull., 1945,
42, 539-540.
Rundquist, E. A. Form of statement in personality
measurement. J. educ. Psychol., 1940, 31, 135-147.
Sakoda, J. M. , Cohen, B. M., and Beall, G. Test of signifi
cance for a series of significance tests. Psychol.
Bull., 1954, 51, 172-175.
Sherif, M. and Cantril H. The psychology of attitudes,
Part II. Psychol. Rev., 1946, 53, 1-24.
Steinmetz, H. L. Measuring ability to fake occupational
interest. J. appl. Psychol., 1932, 16, 123-130.
APPENDICES
APPENDIX A
118
GZTS
CORRELATION COEFFICIENTS
CONTROL GROUPS1
G R A S E 0 F T P M Crit.
G 169 329 465 245 194 098 118 228 037 414
R 129 -055 -127 040 205 348 267 202 185 257
A 285 -116 571 278 281 -088 340 104 326 170
S 371 -166 536 538 457 049 208 132 015 269
E 136 249 12 9 345 700 317 070 275 330 172
0 094 312 206 410 699 548 022 498 375 338
F 102 447 -125 097 432 609 -047 699 359 -105
T 102 344 278 147 100 137 068 059 288 082
P 18-3 282 060 178 356 580 676 137 361 -031
M 076 321 417 131 371 466 376 290 406 -071
Crit. 117 148 -020 027 210 287 105 055 151 211
Above the diagonal are coefficients for pass and
fail. Below the diagonal are coefficients for pass and
dropout. All decimal points omitted.
119
GZTS
CORRELATION COEFFICIENTS
EXPERIMENTAL GROUPS1
G R A S E O F T P M Crit.
G 115 288 092 142 170 -015 133 023 124 197
R 099 079 -021 330 273 2 95 385 130 115 083
A 277 004 541 470 400 -035 057 190 193 550
S 038 -188 416 497 353 -019 105 128 -038 474
E 048 187 311 365 709 319 -001 451 243 362
0 071 093 15 9 193 646 377 -117 534 243 549
F -083 208 -115 -035 249 343 -117 373 158 126
T 162 389 074 089 -033 -197 -136 -155 -072 -039
P 104 200 158 121 496 619 394 -100 278 257
M 105 125 219 -068 195 341 202 -082 244 061
Crit. -238 -310 -459 -250 -260 -196 -221 046 -399 -113
Above the diagonal are coefficients for pass and
fail. Below the diagonal are coefficients for pass and
dropout. All decimal points omitted.
120
DFOS
CORRELATION COEFFICIENTS
CONTROL GROUPS1
NA LT AS SR AA CC NF RT NP ND Crit.
NA 352 236 -283 220 188 290 -568 162 399 -077
LT 283 428 211 492 134 -053 -111 649 387 207
AS 281 346 162 353 069 083 -206 434 461 091
SR -335 249 -025 152 -119 -397 527 273 -162 093
AA 181 452 266 097 -024 -195 045 371 346 007
CC 196 145 085 009 -089 125 -225 276 029 209
NF 236 -090 126 -395 -158 054 -510 -134 049 -003
RT -624 -141 -279 521 -058 -247 -403 -094 -463 142
NP 076 645 294 239 298 282 -136 -129 334 106
ND 514 409 527 -234 45 0 -006 088 -528 329 -126
Crit. -010 319 028 177 035 450 -062 -056 272 -052
Above the diagonal are coefficients for pass and
fail. Below the diagonal are coefficients for pass and
dropout. All decimal points omitted.
121
DFOS
CORRELATION COEFFICIENTS
EXPERIMENTAL GROUPS1
NA LT AS SR AA CC NF RT NP ND Crit.
NA 166 132 -313 05 0 150 2 78 -404 017 231 -029
LT 247 259 -014 436 375 -309 036 739 414 192
AS 111 169 118 403 -316 159 -101 241 444 247
SR -270 054 107 -022 -161 -318 475 -028 -134 529
AA 201 459 387 -051 -051 -140 088 321 458 447
CC 114 389 -237 074 125 -144 -128 271 018 -024
NF 221 -320 218 -073 -166 -171 -367 -384 -011 -297
RT -361 093 -070 436 034 -006 -306 -007 -390 327
NP -017 680 175 040 378 256 -423 174 382 094
ND 277 343 432 -020 360 -025 -001 -334 343 133
Crit. Oil -086 138 -056 001 -219 403 -001 -242 162
Above the diagonal are coefficients for pass and
fail. Below the diagonal are coefficients for pass and
dropout. All decimals points omitted.
APPENDIX B
123
RESPONSE CHOICES OF POPULAR SCORE KEYS FOR
CONTROL GROUP, INSTRUCTORS, EXPERIMENTAL GROUP
ON GZTS
of
Ltem
Cont. Inst. Exp.
NO. Of
item
Cont. Inst. Ex]
1 Y Y Y 30 Y Y Y
2 N
-
N 31 N N N
3 Y Y Y 32 N N N
4 Y Y Y 33 N N N
5 N N N 34 N N N
6 N N N
35 Y Y Y
7 N N N 36 Y Y N
8 N N N
37 - Y Y
9 Y Y Y
38 Y Y Y
10 Y N Y
39 N -
Y
11 Y Y Y
40 N N N
12 Y N Y
41 N N Y
13 N N N
42 -
N N
14 Y Y Y
43 N N N
15 N N N
44 Y Y Y
16 N N N
45 N N N
17 Y Y Y
46 Y Y Y
18 N N N 47 N N N
19 N N N
48
-
N N
20 N - N 49 - Y Y
21 N N N
50 N N N
22 Y Y Y
51 Y Y Y
23 Y Y Y 52 N N N
24 Y Y N
53 N Y Y
25 N N N
54 N N N
26
-
N N
55 N N N
27 Y Y Y
56 Y Y Y
28 Y
-
Y
57 N N N
29 Y Y Y
58 N N N
124
(continued)
>. of
tem
Cont. Inst. Exp.
No. of
item
Cont. Inst. Exj
59 N N N 89 N N N
60 N N N 90 N N N
61 N N N
91 N N N
62 N N N 92 Y Y Y
63 N N N
93 Y - Y
64 N N N 94 N N N
65 Y Y Y 95 N N N
66 N N N
96 Y Y Y
67 N N N
97 N N N
68 N N N
98 N Y N
69 N N N
99 N N N
70 N N N 100 N N N
71 Y Y Y
101 N N N
72 N N N 102 Y Y Y
73 Y Y Y
103 - - Y
74 N N N
104 Y - Y
75 N N N
105 Y Y Y
76 N N N
106 Y Y Y
77 N N N
107 Y - N
78 N - Y
108 N N N
79 N N Y
109 N N N
80 Y N N 110 Y Y Y
81 N N N 111 Y Y -
82 N N N 112 N N N
83 N N N 113 Y Y Y
84 N N N
114 N N N
85 N N N
115 N N N
86 N N N
116 N N N
87 N N N
117 N N N
88 Y Y Y
118 N N N
125
(continued)
No. of
item
Cont. Inst. Exp.
No. of
item
Cont. Inst. Exp.
119
-
Y Y 149 N N N
120 N N N 150 Y Y Y
121 Y Y Y 151 Y Y Y
122 Y Y Y 152 N N N
123 Y N Y
15 3 N N N
124 N N N
154 N N N
125 N N N
155 N N N
126 N N N
156 Y Y Y
127 N N N
15 7 N N N
128 N
-
Y
158 Y Y Y
129 N N N
159 Y Y Y
130 N N N
160 N N N
131 - Y . Y
161 N N N
132 N N N
162 N N N
133 N N N
163 Y Y Y
134 Y Y
164 N N N
135 Y Y Y
165 Y Y Y
136 N N N
166 N N N
137 N N N
167 N N N
138 N Y Y
168 Y Y Y
139 Y Y Y
169 Y Y Y
140 Y Y Y
170 Y Y Y
141 Y Y Y
171 N N N
142 N N N
172 N N N
143 Y Y Y
173 N N N
144 Y Y Y
174 Y Y Y
145 N N N
175 N N N
146 N N N
176 Y N N
147 N N N
177 N N N
148 Y Y Y
178 N N N
126
(continued)
No. of
item
Cont. Inst. Exp.
No. of
item
Cont. Inst. Exp
179
- Y N 209 N N N
180 N N N 210 Y Y Y
181 N N N 211 N N N
182 Y Y Y 212 N N N
183 Y Y Y
213 Y Y Y
184 N N N
214 Y Y Y
185 - N Y
215 N N N
186 N N N
216 N N Y
187 N N N
217 N N N
188
-
Y -
218 Y N Y
189 N N N
219 - Y Y
190 N N N
220 Y Y Y
191 N N N
221 N N N
192 - N N
222 N N N
193 N N N
223 Y Y Y
194 N N N
224 Y Y Y
195 N N N
225 N N N
196 Y Y Y
226 Y N Y
197 N N N
227 Y -
Y
198 Y Y Y
228 Y Y Y
199 Y Y Y
229 - N N
200 Y N N 230 N N N
201 N N N
231 N Y Y
202 N N N
232 Y - -
203 Y Y Y
233 N N N
2 04 N N N
234 Y Y Y
205 - - N
235 Y N Y
206 N N N
236 N N N
207 N N N
237 Y Y Y
208 N N N
238 Y Y Y
127
(continued)
No. of
item
Cont. Inst. Exp.
No. of
item
Cont. Inst. Exp.
239 Y Y Y
270
-
N N
240 Y Y Y
271 N N N
241 N N N
272 N N N
242 N N N
273 Y Y Y
243 Y
- N
274 N N N
244 N N N
275 Y Y Y
245 Y Y Y
276 N N N
246 N N N
277 N N N
247 N N N
278 Y - Y
248 Y N Y
279 Y Y Y
249 Y Y Y
280 N N N
250 N N N
281 N N N
251 N - Y
282 N N N
252 Y N Y
283 N - N
253 N N N
284 N N N
254 N N N
285 Y -
Y
255 N N N
286 N Y Y
256 N N N
287 Y Y Y
257 N N N
288 N -
N
258 Y Y Y
289 Y Y N
259 N N N
290 N N N
260 N N N
291 N N N
261 Y N N
292 N N N
262 Y N N
293 Y Y Y
263 N N N
294 N N N
264 Y Y Y
295 Y - Y
265 N - N
2 96 N N N
266 N N N
297 N N N
267 N N N
298 Y Y Y
268 N N N
299 Y Y Y
269 Y Y Y
300
—
N N
128
RESPONSE CHOICES OF POPULAR SCORE KEYS FOR
CONTROL GROUP, INSTRUCTORS, EXPERIMENTAL GROUP
ON DFOS
). of
.tem
Cont. Inst. Exp.
No. of
item
Cont. Inst. Exj
1 Y Y Y 30 N N N
2 Y N Y 31 N N N
3 N N N 32 Y Y Y
4 N N N
33 Y Y Y
5 N
-
N
34
-
Y N
6 N N N
35 Y Y Y
7 Y Y Y
36
-
N ■ -
8 Y Y Y
37 Y N -
9 N N N
38 N N N
10 N N N
.39 N N N
11 N N N
40 N N N
12 Y Y Y
41 N N N
13 N N N 42 Y Y Y
14 N N N
43 N N N
15 Y - Y
44 N N N
16 Y Y Y
45 Y Y Y
17 N N N
46 Y - Y
18 N N N
47 N N N
19 N N N
48 N N N
20 Y Y Y 49 N N N
21 N N N
50 N N N
22 N N Y
51 N N N
23 Y Y Y
52 N N N
24 Y N Y
53 N Y N
25 Y Y Y
54 -
N N
26 N N N 55 Y Y Y
27 Y Y Y 56 Y N Y
28 Y N -
57 N N N
29 N N N
58 N N N
129
(continued)
•. of
tern
Cont. Inst. Exp.
No. of
item
Cont. Inst. Ex]
59 Y Y Y 89 Y -
Y
60 Y Y Y
90 N - N
61 N N N 91 Y Y Y
62 Y Y Y
92 Y - Y
63 N N N
93 Y
-
N
64 N N N
94 Y Y Y
65 Y Y Y
95 Y Y Y
66 Y Y Y
96 N N N
67 N N N
97 N N N
68 Y
-
N
98 N N N
69
-
N Y
99 N N Y
70 Y Y -
100 N - N
71 Y Y Y
101 N N N
72 Y Y Y
102 - Y -
73 N N N 103 N N N
74 Y Y Y
104 N N N
75 Y Y Y
105 Y Y
76 Y Y Y
106 N N N
77 N N N
107 N N N
78 N N N
108 Y N
-
79 Y Y Y
109 Y Y Y
80 Y Y Y
110 Y - Y
81 N N N
111 Y - N
82 N N N
112 N N N
83 Y N N
113
- -
N
84 N N N 114 N N N
85 Y Y Y
115 Y Y Y
86 Y N Y
116 Y - Y
87 N N N
117 N N N
88 N N N
118 N N N
130
(continued)
i. of
,tem
Cont. Inst. Exp.
No. Of
item
Cont. Inst. Ex]
119 N N -
149 Y N Y
120 Y N Y
150 Y Y Y
121 N N N
151 N N -
122 Y N Y
152 N N N
12 3 N N N
153 N N N
124 - N N
154 N N N
125 - Y Y
155 Y Y
126 Y Y Y
156 Y - Y
127 N N N
157 N N N
128 N N N
158 N N N
129 Y - Y
15 9 N N N
130 N N Y
160 N , . N N
131 N N N
161 N N N
132 N
-
N
162 -
_
Y
133 Y Y Y
16 3 Y - N
134 N N N
164 N N N
135 N Y Y
165 Y Y Y
136 Y Y Y
166 Y Y Y
137 N N N
167 N N N
138 Y N N
168 Y N . N
139 Y Y Y
169 N N N
140 Y Y Y
170 Y
-
Y
141 N N N
171 N N N
142 N N N
172 N - Y
143 N N N
173 Y
-
N
144 N N -
174 N N N
145 N Y Y
175 N N N
146 N N N
176 - N N
147 N
-
N
177 N N N
148
-
N N
178 N N N
(continued)
131
ro. of
item
Cont. Inst. Exp.
No. of
item
Cont. Inst. Exp
179 N N N 209 N N N
180 Y Y N
210 Y Y Y
181 N N N
211 Y Y Y
182 Y Y Y
212 Y Y Y
183 N N N
213 Y Y Y
184 N N N
214 N N N
185 N N N
215 Y Y Y
186 N N N
216 Y N Y
187 N
-
N
217 N N N
188 N N N
218 - ■ Y Y Y
189 N ' N N
219 N N N
190 - -
N
220 Y
-
Y
191
-
N N
221 N N N
192 N N N
222 Y Y Y
193 N N N
223 N N N
194 - N N
224 Y Y Y
195 Y Y Y
225 N N N
196 Y Y Y
226
- -
N
197 N N N
227 - Y Y
198 Y - Y
228 N N N
199 N N N
229 N N -
200
-
N N
230 Y Y Y
201 N N N
231 Y Y Y
202 N N Y
232 N Y Y
203 Y Y N
233 N N N
204 N N N
234 Y Y Y
205 Y Y Y
235 Y Y Y
206 Y N N
236 Y Y Y
207 N - N
237 N N N
208 Y Y N
238 N N N
(continued)
132
No. of
item
Cont. Inst. Exp.
No. of
item
Cont. Inst. Exp.
239 N N
-
270 Y - Y
240 Y Y Y
271 N N N
241 N
-
N
272 Y Y Y
242 Y Y Y
273 Y Y Y
243 Y
-
N
274 N N N
244 N N N
275 Y N N
245 Y Y Y
276 Y N Y
246 N N N
277 N N N
247 N - N
278 N N N
248 - N N
279 N N N
249 N N N
280 N N N
250
- - N
281 Y - N
251 N N N
282 N N N
252 N N N
283 N N N
253 Y N N
284 N N N
254 N N N
285 N N N
255 Y Y Y
286 N N Y
256 Y Y Y
287 N N N
257 Y Y Y
288
_
N N
258 N N N
289 Y Y Y
259 Y Y Y
290 N N N
260 N Y N
291 N N N
261 - N N
292 N N N
262 N Y Y
293 N N N
263 - N N
294 N N N
264 N N N
295 Y Y Y
265 Y Y Y
296 N N N
266 Y Y Y
297 Y
-
Y
267 N N N
2 98 N N N
268 Y N Y
299 N N N
269 N N Y
300 Y - Y
APPENDIX C
134
GZTS BETA WEIGHTS DERIVED FROM
DOOLITTLE SOLUTION FOR MULTIPLE R
Variable
Test Condition and Criterion Group
Control
Fail*
Experimental
Fail*
Control
Dropout
Experimenta1
Dropout
G .390 .041 . 122 -.094
R .237 .034 - .002 -.240
A -.063 .354 -.117 -.248
S -.056 .258 -.047 -.128
E -.140 -.313 -.011 .054
O .695 .515 . 174 .071
F -.443 .030 -.119 -.097
T .024 -.044 .010 .165
P -.106 .025 -.004 -.172
M -.132 -.090 . 112 . 053
*
Derived by calculator, based on biserial corre
lation with the criterion. All others were derived by
computer, based on point biserial correlation with the
criterion.
135
DFOS BETA WEIGHTS DERIVED FROM
DOOLITTLE SOLUTION FOR MULTIPLE R
Test Condition and Criterion Group
Variable control Experimental Control Experimental
Fail Fail Dropout Dropout
NA -.065 .115 -.095 -.112
LT .183 -.040 .186 .121
AS .083 .059 -.022 -.046
SR -.040 .324 .041 -.112
AA -.051 .280 .001 .017
CC .157 .091 .272 -.081
NF .032 -.091 -.022 .266
RT .085 .057 -.071 .169
NP -.057 -.052 -.030 -.179
ND -.085 .012 -.066 .204
APPENDIX D
137
ITEMS ANSWERED BY SIGNIFICANTLY DIFFERENT PROPORTIONS
OF THE CONTROL AND EXPERIMENTAL GROUPS/ AS KEYED IN
THE GZTS MANUAL, BY FACTORS FOR CRITERION GROUPS;
PASS (P), NON-PASS (N) AND TOTAL (T)
Item Criterion
Number Groups
Item Criterion
Number Groups
16
21
36
46
51
56
71
86
91
101
121
126
131
141
146
P, T
N, T
N, T
N, T
P,N,T
P, T
T
T
P, T
N, T
P,N,T
P, T
P, N, T
P, N, T
P
7
17
27
32
42
47
52
62
67
72
77
87
102
107
122
137
142
T
T
T
T
N, T
N, T
P, T
P, T
P, T
P
P, T
P,T
N, T
P, T
P, T
P, T
P, N, T
138
(continued)
A
Item Criterion Item
Number Groups Number
3 P,N,T 9
28 P,T 24
38 P,N7T 34
43 P,T 39
48 P7N7T 44
53 P7N7T 49
63 P7T 54
73 P7N,T 64
78 P, T 69
88 ° P,N,T 74
93 N 79
118 P,T 84
128 P,T 89
138 P,N,T 99
109
114
119
124
144
149
Criterion
Groups
P, T
T
P7N,T
P,N,T
T
P7N,T
p7t
P, T
T
T
T
T
T
N, T
P, T
P,T
P,N,T
P, N, T
T
P, N, T
139
(continued)
E
Item Criterion Item
Number Groups Number
5 T 151
10 P,N,T 161
15 N, T 176
20 T 191
30 P,T 196
40 P,T 211
50 P,T 216
55 Px T 231
60 P,T 236
70 T 241
75 P, T 251
80 P,N,T 261
100 P,N,T 266
105 T 271
110 P,T 276
115 P,N,T 286
120 P,T
125 N,T
130 P,T
140 T
145 N,T
Criterion
Groups
P, T
T
P, N, T
P,N, T
P, N, T
P,T
P, T
P, N, T
T
T
P, T
P, N, T
P, T
P, N, T
P, T
N, T
140
(continued)
F
Item Criterion Item
Number Groups Number
152 P,T 153
172 T 163
187 P,N,T 168
192 N,T 178
202 N,T 183
222 T 193
227 P,T 208
232 T 213
262 P,T 233
267 T 243
282 N,T 248
292 P,N,T 253
297 P,T
Criterion
Groups
N/ T
P,N,T
P, N, T
T
P, N, T
T
P, T
P, N, T
N, T
P, N, T
N, T
N, T
Itei
umbi
154
15 9
194
219
234
239
274
279
289
141
(continued)
M
Criterion Item Criterion
Groups Number Groups
T 170 P,T
N,T 175 P,T
P 180 P,T
N, T 195 P,T
T 205 P,T
P,T 225 P,N,T
T 230 P
T 250 P,T
P, T 255 N, T
285 P
290 P
295 N
300 P,T

Asset Metadata

Creator Plittman, Jack C. (author)

Core Title The Influence Of The Control Of Personal Set Upon Prediction By Factored Tests Of Temperament And Interest

Contributor Digitized by ProQuest (provenance)

Degree Doctor of Philosophy

Degree Program Psychology

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag OAI-PMH Harvest,psychology, general

Format dissertations (aat)

Language English

Advisor Ruch, Floyd L. (committee chair), Jacobs, Alfred (committee member), Locke, Harvey J. (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c18-304013

Unique identifier UC11358945

Identifier 6402601.pdf (filename),usctheses-c18-304013 (legacy record id)

Legacy Identifier 6402601.pdf

Dmrecord 304013

Document Type Dissertation

Format dissertations (aat)

Rights Plittman, Jack C.

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA