Cognitive differentiation and occupational-profile differentiation on the Strong vocational interest blank


Material Information

Cognitive differentiation and occupational-profile differentiation on the Strong vocational interest blank
Added title page title:
Strong vocational interest blank
Physical Description:
iv, 260 leaves. : illus. ; 28 cm.
Krienke, John Walter, 1934-
Publication Date:


Subjects / Keywords:
Vocational interests   ( lcsh )
Cognition   ( lcsh )
Psychology thesis Ph. D   ( lcsh )
Dissertations, Academic -- Psychology -- UF   ( lcsh )
bibliography   ( marcgt )
non-fiction   ( marcgt )


Thesis -- University of Florida.
Bibliography: leaves 166-172.
General Note:
Manuscript copy.
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 022045510
oclc - 13432933
System ID:

Full Text









The author wishes to express his gratitude to Dr. Benjamin Barger,

the chairman of his supervisory committee, vho was most helpful in

clarifying and limiting the original ideas for this study and who pro-

vided support and Imptus throughout. Drs. Louis D. Cohen, J. Milan

Kolarik, and Charles W. Morris are also thanked for not only their

support and criticisms of various aspects of the dissertation but also

for the inspiration they provided throughout the graduate program.

Thanks also go to Drs. Wilse B. Webb and Henry S. Pennypacker, the

fonrihr for serving on the committee until he left for a period of study

in Europe and the Latter for being willing to replace Dr. Webb on the

corimttee at such a late date. Though he did not serve on the commit-

tee, Dr. Hugh C. Davis is also thanked for his continuing interest in

the research goals of the author and for his help in the design of this

study. The author would also like to express his deep gratitude to his

wife, Carolyn, vhosq invaluable aid in the collection and scoring of the

data and the typing of the dissertation-drafts shortersd imeasurably

this task. This gratittdo extends to his entire family, ircludLng his

children and his parents, all of wehom made those sacrifices which only

the families of graduate students can appreciate. Acknow:ledgment is

also made of the services of the University of Florida Conputing



ACKNO;'LEDGMSNTS . . . . . . .

LIST OF TABLES . . . . .... .




III RESULTS . . . a o .

IV DISCUSSION . . . . . .

V SUP AARY . . . . .

REFERENCES . . . . . . . .

APPENDICES . . . . . . . .

O 6 0 0 0

* 0 0 0 S 0

* 0 S S S S S S

* 0 S S S * S 0 0

* 0 0 6 5 6 0 S S

* S 5 0 5 S S S 0 5

* S 5 0 * S S 0

* 0 5 5 0 0 5 0 5 0

* S S S S S 0 5 S




















Table Page

1. Means and Standard Deviations of the
Paid and Non-Paid Groups on the
Four SVIB Criterion Variables . . . . 54

2. Percentages of Occurrence of Different Interest
Patterns in Three Different Samples * 55

3. Percentages of Occurrence of Primary and Secondary
Interest Patterns Within the Different
Occupational Families in Three Different Samples 56

4. Initial-Sample Canonical Correlations
with Associated Chi Squares a . . . 64

5. Canonical Coefficients of the First Canonical
Correlation of the Initial Sample . . . . 64

6. Cross-Validation-Sample Canonical Correlations
with Associated Chi Squares .. .. . . 65

7. Canonical Coefficients of the First Canonical
Correlation of the Cross-Validation Sample . 66

8. Intc,-correlations of the Seventeen Variables
in the Initial Sample .. . . . . . 68

9. Intercorrelations of the Seventeen Variables
in the Cross-Validation Sample . . . . a 69

10. Correlations Found to be Significant in the
Initial-Sample Correlation Matrix Compared
with the Corresponding Correlations in the
Cross-Validation-Sample Correlation Matrix . 70

11. Intercorrelational Patterns in the Initial
and Cross-Validation Samples . . . . 77

12. Moderated Predictor-Criterion Relationships Corre-
lated Significantly in the Initial Sample and
Receiving Support in the Cross-Validation Sample 83

13. Supported Moderated Relationships Between Criterion
Measures and Variables Used as Predictors
Arranged According to Possible Connon Variances 93

14. Supported Moderated Relationships Arranged
Accorflin to Modrating Conditions. . . . 108

15. Distribution of Scores of 252 Subjects
on the Group mbecIdod Figures Test . . . . 13?



For some forty-five years the Strong Vocational Interest Blank For

Men (SVIB) has been the preferred instrument of vocational counselors in

dealing with college students. In fact, as Cronbach has stated, the

Strong Blank is not only the "most highly developed and best understood

of the inventories," it also "ranks very near the top among psychologi-

cal tests of all types" (Cronbach, 1960, p. 434). Super and Crites have

agreed that "it is without question one of the most thoroughly studied

and understood psychological instruments in existence" (1962, p. 418).

Despite this amount of research, one problem in relation to the inventory

has existed for many years and still today remains one of the major un-

resolved aspects of the test. This problem is the one presented by those

men who end up with no occupations being selected by the SVIB from their

answers to the inventory questions; their profiles on the Strong are

"flat" or undifferentiated, and they are said to have low intensity or

weak SVIB vocational interests. Despite a number of studies, the char-

acteristics of men obtaining such profiles are still mostly a matter of


This problem was the beginning point of this study, but because

of so many previous failures in this area it was felt important to change

the usual strategy of attack on the problem. First, it was decided to

broaden the con,-crn of the study from just comparing groups of man with


and without flat profiles into a study of "flatness" or "height" of

the profile as a variable. Instead of studying differences between

groups, an attempt was made to determine how well certain predictor var-

iables were related to criterion variables obtained from different ways

of quantifying flatness or height.

Secondly, as was just intimated, this study also attempted to

evaluate some different ways of defining a criterion variable of height.

As Athelstan (1967) pointed out, one of the critical problems with many

of the studies in this area has been the problem of defining the crite-

rion group. By studying a variable, rather than a group, one does not

have to fear that by defining a group in a certain way certain important

individuals have bean cast aside; in this study all the individuals who

took the SVIB w'ere used. And by using the technique of canonical correl-

ation it was possible to make some evaluation of the usefulness of dif-

ferent ways of defining the criterion variable.

The third difference of this study from the previous ones was the-

attempt to relate SVIB occupational profile differentiation to a certain

class of cognitive style variables, those having to do with cognitive

differentA.ation, rather than to some kind of motivational (or "dynamic")

variable as has been the custom in the past. The main hypothesis to be

tested in this study was that in order for a person to display occupa-

tional differentiation on the SVIB he would need to be cognitively

differentiated to some degree; that is, a necessary condition (though

obviously not a sufficient condition) for occupational differentiation

is cognitive differentiation. Thus, it was expected that a low posi-

tive correlation would be found between various measures of cognitive

differentiation and the various criterion variables of occupational

differentiation on the S'JIB.

The fourth difference between this study and previous ones was the

attempt to study how the various cognitive variables interrelate to one

another and how their effect is "moderated" by the presence of certain

motivational variables found in recent studies to have important effects

on other personality variables. Through the use of this "moderator

analysis" it was hoped that non-linear interactive effects between these

variables would come to light.

The final difference in this study from the previous ones was the

decision to include a replication group within the study. That is, the

original sample was randomly divided into two samples called the initial

sample and the cross-validation sample. Only those results which were

significant in the initial sample and which were also present in the

cross-validation sample were considered to be valid.

Literature Review

Almost thirty years ago the inventor of the SVIB was complaining

of the lack of knowledge about those men who end up with no occupations

being chosen on the SVIB; he felt that these persons presented the
"toughest" kind of counseling problem (Strong, 1943). In 1966 in the

manual for the nervest revision of the test the authors were still feeling

that this problem needs "much more research" (Strong, 1966, p. 12), so

that despite the passage of some twenty-five years, the newest manual

can, for the most pert, only repeat the speculations Strong originally

made in 1943. At that tine Strong suggested that some flat profiles may

be produced by a chance distribution of answers to the test. Another

possibility, Stirong said, was that the individual who produced such a


low profile had interests for which there was no occupational scale,

e.g., mariner. On the other hand, Strong felt that there was some evi-

dence available to suggest that men with no primary occupational interest

patterns look like underachievers. Still, Strong also suggested that the

general elevation of the SVIB profile may be a measure of the "motivation

the individual has at his disposal for working hard," and therefore low

profiles often give the impression that their owners are "drifters"

(Strong, 1943, pp. L,4O-44l). But, on Lhe other hand, Strong also pointed

out that some low profiles are produced by successful students in the

various "business" areas, so a low profile may suggest a certain flexi-

bility required for advancing business executives: the interests of

business men "are widespread and on the average not so strongly held."

By 1955 in his next book Strong could report from his research

that older men obtain profiles with more intense interests than do younger

men. He also, by this time, had given up his motivational interpretation

of flat profiles, while retaining all the other 1943 speculations. In

place of the motivational interpretation he attempted to substitute a

theory involving attention (Strong, 1955). He saw intense interests as

being connected with freely given persistent attention, and low profiles

were thus partially due to the person's inability to focus his attention

on any one vocation.

In giving his counseling hints he repeated his concern about those

persons with low profiles, and he reported that in counseling he fol-

lowed the practice of classifying all the profiles into two main groups:

those with a V-shaped profile and those with a flat profile. "It is

the flat profiles that are troubleso-ime; men with such profiles need

thoughtful coirnseling" (Strong, 1955, P. 193).


The writer's guess is, and it is still a guess, that these men
will never have the keen enjoyment in their work which is char-
acteristic of nany professional men but, on the other hand, they
will never be so frustrated, whatever phase of business they are
engaged in, as professional men when these latter are employed
at work foreign to their interests. (p. 194)

Also actively involved for a long time in counseling with the

SVIB has been John G. Darley at the University of Minnesota. In 1941

he published a guide for the clinical use of the Strong and reported

a great deal of research, normative data, etc., on the SVIB (Darley,

1941). In 1955 he and Theda hagenah replaced his earlier work with a

much expanded one that included new research on the 1938 revision of

the SVIB (Darley & Hagenah, 1955). In this latter book the problem of

low profiles is recognized (p. 214), but there is very little discus-

sion of the reasons for such profiles. Their only suggestion is that

"there is a lower limit of interest differentiation" (p. 26) and that

this limit "is related, no doubt, to the level of ability, to the socio-

economic status, to the personality and motivation of the individual

student" (p. 27). Little evidence is offered to support this statement.

In the 1966 SVIB manual (revision by Campbell: Strong, 1966), many

of Strong's 1943 statements are repeated, including the statements aboat

the businessmen profiles: "The really tough counseling assignment occ-,i-s

when a person has no A ratings, few if any B+ ratings, and does not have

interests characteristic of business" (p. 11). In this virtually word-

for-word repetition of the reasons given by Strong in 1943, Campbell

includes the above mentioned businessmen interests, very specialized

interests, general lack of interests, or interests of workmen and not of

professionals. In place of the notion that those with flat profiles are

not willing to work hard, Campbell substitutes an idea from Layton (1958)

that "the lack of any pattern on the Strong Blank is related to


A person may be too fearful and unsure of himself to commit him-
self thoroughly to any vocation, or nay be characteristically
indecisive and ambivalent about most choices confronting him.
With young people, the lack of high scores may be associated with
irnaturity. -Mch more research is needed in this area. (Strong,
1966, p. 12)

Though the manual claims that some research has been done, a check with

Layton (1958) reveals only two studies by Berdie (1945, 1946) relating

constricted interests to maladjustment. A group of psychologically

abnormal people as well as some military trainees who did not adjust well

to training did not check as many recreational interests on a checklist

as did normals. The inference is that maladjusted people have a narrower

range of interests than normals and would probably have flat profiles on

the SVIB, but unfortunately the SVIB was never used to check this out.

Berdie (1943) did investigate what might be a very obvious hypoth-

esis about the flat profile: the person chose to respond with an unusual

number of "indifferent" choices. The SVIB, of course, is composed mostly

of items concerned with vocations, avocations, activities, school sub-

jects, personality types, etc., and to most of these items the subject

is asked to respond with Like, Dislike, cr Indifferent. Following the

example of Rulon (1931) Berdie marked three different answer sheets with

all Like, all Dislike, and all Indifferent responses.1 The Like-marked

answer sheet produced seven A occupations, the Dislike-marked sheet also

produced seven A occupations, and the Indifferent sheet produced five A

occupations. Birdie concluded that marking all the Likes produced a

1It is not clearly stated how Berdie marked those questions not
having the L-I-D categories, though he probably marked them the way he
reported Rulon had done. On his "Like" answer sheet Rulon marked such
questions in theh left column; on the "Dislike" answer sheet, in the
right column; on the "Indifferent" answer sheet, in the middle column.


profile characteristic of those obtained by people in social service

fields, business detail fields, the skilled trade fields and by produc-

tion managers. Narking all the Dislikes produced a profile similar to

those produced by people in the verbal-linguistic fields, such as

journalism and law, and by those in the scientific-creative fields.

Marking all the Indifferent choices produced a profile similar to those

of skilled tradesmen, musicians, and some social service people. It

thus appears that being indifferent to the choices on the Strong Blank

is not what produces a flat profile.

Despite these observations by Berdie, efforts have been made to

connect response styles to low profiles. Tyler (1960) suggested on the

basis of clinical observation that cases with no intense interest pat-

terns often had very few Dislike responses on their blanks. It was her

hypothesis that vocational interests mature within a person as he forms

more interest-rejections, so that, up to a point, the clearer the person

perceives his role the more Dislike responses he will give. A recent

study (Athelstan, 1967) of 151 low profiles, however, found that com-

pared with a randomly selected control group the low profile answer

sheets contained fewer Like and Indifferent responses and more Dislike

responses, the exact opposite of Tyler's predictions. Nevertheless,

other studies (Stewart, 1960; Armatas & Collister, 1962) have indicated

that the L-I-D responses reflect not only the responder's vocational

interests but also his personality traits.

In the studies which have been actually concerned with the problem

of low profiles, the comncn concept which has been used to describe the

occurrence of these profiles has been "primary patterns." This concept,

first published by Da-ley in 1941, is basod u-on the fact that the SVIB


profile sheets group the occupational scales, according to content and

factorial relationships, into various occupational "families." Darley

proposed a set of rules by which one could say how strong a person's

interests in a particular occupational family were:

The primary pattern is the interest type within which he
shows a preponderance (plurality or majority) of A and B+ scores
on the specific occupational keys; the secondary pattern is the
interest type within which he shows a preponderance of B+ and
B scores. (Darley, 1941, p. 17)

The problem with this "plurality or majority" rule, as later writers

pointed out, is that it was devised for counseling procedures, and it

was thus of no great concern to Darley that one counselor could judge

a particular profile as having a certain primary pattern while a second

counselor might feel it was only a secondary pattern. Despite this

ambiguousness, Darley used this rule in publishing normative data on

three years of freshmen (ending in 1941), including in the data the

number of primary and secondary patterns encountered in this group

(Darley & Hagenah, 1955).

It was in this later book that Darley also introduced the related

concept of the "reject pattern," as a more useful and meaningful pattern

than the & rlier defined tertiary pattern. A reject interest pattern

was recorded in this study "whenever the majority or plurality of the

scores lay to the left of the 'chance,' or shaded, area on the profile"

(Darley & Hagenah, 1955, p. 79). The 1eject pattern was interpreted to

mean that not only was the subject not interested in that particular

occupational faraily but he was also positively rejecting the family.

Though it might appear that the number of reject patterns is

another way of characterizing a flat SVIB profile, the normative data

collected by Darlrey and hagpnah indicate otherwise, Tn the 1000 cases


they studied, 94 percent had one or more reject patterns, while 20 per-

cent had no primary pattern and 2.2 percent had neither a primary nor a

secondary pattern. The cases with reject patterns were distributed as

follows: single reject with no primary--9 percent; single reject with

single primary--9 percent; single reject with multiple primary--5 per-

cent; multiple reject with no primary--7 percent; multiple reject with

single primary--29 percent; multiple reject with multiple primary--33

percent (Darley & Hagenah, 1955, P. 90). It seems then that reject

patterns tended to be associated with the presence of primary patterns

rather than with their absence.

It was not until 1961 that alternative rules for classifying

primary patterns began to appear, and the first was Stephenson's (1961).

Noting the difficulties with Darley's rules, he arbitrarily resolved the

ambiguous situations and published a table of general instances of these

resolutions. It is possible to use his table to classify, with perfect

repeatability, any occupational family that includes from one to nine

occupational scales, into classifications of primary, secondary and

reject. The one great change from Darley's rules is with the reject

decisions, Stephenson has changed this rule so that in general any

family containing a majority of C range scores is ruled to be rejected.

A set of slightly more lenient primary and secondary rules was

published a year later by Korn and Parker (1962) along with a much

stricter rule for reject patterns. They also published normative data

on the occurrence of each classification in the 1959 freshman class at

Stanford University. These rules are as follows:

1. Primary pattern--a majority of B+ or higher scores on the
specific occupational keys.
2. Secondary pattern--a majority of scores are B or higher,


provided that it does not qualify as a primary pattern.
3. Reject pattern--a majority of scores at standard score 15 and
below. (Korn & Parker, 1962, pp. 223-224)

An inspection of Stephenson's (1961) table reveals that more patterns

would be classified primary and more as secondary under Korn and Park-

er's rules than under Stephenson's. However, Korn and Parker would find

fewer reject patterns with their rules than would Stephenson using his

rules. Hov many fewer can be estimated by the fact that Korn and Park-

er found that 78 percent of their sample had no reject patterns, while,

as stated above, Darley and Hagenah found only 6 percent of their sample

had no reject patterns.

The various studies of low SVIB profiles appear to be divided

between those which used Darley's rules for classifying primary patterns

and those which used Korn and Parker's revision of these rules. Most of

the former can 'be found reviewed in Athelstan's dissertation (1967),

while Athelstan's work and one other dissertation have used the latter


Studies of Low Profiles

Possibly the earliest study of this problem, and one not cited by

Athelstan (1967), is that done by Kinsala and Phillips (1949). Unfortu-

nately this work was published in a mimeographed "journal," and a copy of

it could not be obtained for this study. From the source of its citation

in The IMiPI Handbook it evidently was concerned with the relationship

between low profiles and MNIPI variables.

Five other studies were done before Athelstan's, and they have

received such an excellent review by him that they will be only briefly

summarized here. Eichsteadt's (1951) dissertLation, completed in 1949,

dealt with the relationship between the number of primary patterns,


YM4I scores, and responses on a general information form used at the

University of Wisconsin Counseling Center, These counselees seen at

the Center who had zero, one or three primary patterns did differ on

some of the information items and on one 12PI clinical scale (the no-

primary group had higher Psychasthenia scores than did the three-

primary group). As Athelstan points out, the differentiating items

are rather heterogeneous and Eichsteadt offers no unifying theory to

tie them together in a meaningful way.

The next study reviewed by Athelstan is Bernstein's (1953) study

of 142 ninth grade boys with no primary patterns. His hypotheses were

that the absence of primary patterns would be associated with scholastic

underachievement and adjustment difficulties and with having fathers in

lower occultional levels. None was supported.

Cooper (1954) took advantage of the 1938 revision of the SVIB

having "group scales," which represented a way of estimating whether a

person's interests would be high on the scales within a particular occu-

pational family without having to laboriously score all the scales within

the family (this was useful before the advent of electronic scoring

devices and is not often used on the 1966 revision). She found that

students -wth no A on a group scale were no different from students with

at least one A group scale, in academic ability, achievement, stability

of vocational choice, or drop-out rate.

The most carefully done study reviewed by Athelstan was that of

Cross (1955), who defined a low profile in a more stringent manner. His

low-intensity interest group had not only no primary patterns but also

no A scores at all. In comparing his low-intensity group with his


high-intensity group (three or more primaries, exclusive of group IV),

he used measures such as the PI, the Minnesota Personality Scale, the

American Council on Euducation Psychological Examination (ACE), work

experience, hobbies, and socio-economic variables. Among all these

variables the only differences he found were that a greater proportion

of the high-intensity group had worked full-time during the previous two

summers and that a greater percentage of the low-intensity subjects were

undecided about their vocational choice. However, among those in the

low-intensity group who had made a vocational choice, "certainty" of

choice was just as great as those making a choice in the high-intensity

group. Athelstan emphasized the work Cross did in searching for rela-

tionships between the SVIB and the IMPI, including comparisons of means

between the two groups on the clinical and ten non-clinical scales,

between the two groups and a representative sample control group on the

same variables, an analyses of profile types including high- and low-

point codes, and so forth.

The final study found by Athelstan and reviewed by him is the one

of Zytowski (1965) done on 35 low-intensity subjects seen at the Univer-

sity of Washington Counseling Service over a five year period. Using a

comparison group of subjects having at least one primary pattern and

matched with the low-profile group on initial of last name and year in

which they were seen at the Service, Zytowski examined comparative

scores on the Kuder, the SVIB non-occupational scales and the distribu-

tion of the L-I-D responses. He hypothesized that the low intensity

profiles would show lower scores on the Immaturity Scale and on the

Occupational Level Scale; would have fewer L and more I and D choices;

and would have less "jagged" Kuder profiles. Only the hypothesis about

less maturity was supported.2

Athelstan's general criticism of these studies is rather scathing,

except for Cross' study. He raises a variety of methodological criti-

cisms, but the failure he considers to be most important is that of

being too lenient in defining a low profile. He heartily approves of

Cross' method of including the restriction of having no A score (besides

having no primary pattern) before one can say a profile is truly low.

Not to do so, as the other studies did, would probably yield "a sample

very heterogeneous with respect to interests" (Athelstan, 1967, p. 10).

Although it is technically possible (though certainly unlikely) from

the way primary patterns are defined to have as many as twenty-two A

scores on the 1966 SVIB and still not have a primary pattern, Athelstan

does not indicate why heterogenity of interests should be more of a

problem for defining the low profile group than it would be for defining

the high profile group. It would seem that the problem of differentiat-

ing between high and low profiles on the SVIB by definition ignores the

variety of interests of the subjects and attempts to account for the

difference in profiles by means of other areas of personality variables.

In any case, Athelstan seems correct in summarizing the data frcm

the studies preceeding his as indicating that "the meaning of low pro-

files on the SVIB is still unclear," and "the prevalence of negative

results among the few relevant studies suggests that a new approach to

studying low profiles is needed" (1967, p. 18). This new approach by

Athelstan consisted basically of a re-test of a low-profile and a

2It should be noted that the 1966 revision of the SVIB no longer
includes the im scale, since it "consistently failed to hold up in
validity studies" (Strong, 1966, p. 52).


control group four years after initial testing and an "essentially

exploratory, descriptive study of low profiles" (p. 21). The data he

used in the study consisted of a selection of SVIB's given to all incom-

ing freshmen in 1961, the corresponding MI scores, Minnesota Scholastic

Aptitude Test (MSAT) scores, Cooperative English Test scores, year of

graduation from high school, high school percentile rank, history of

counseling, as well as the follow-up SVIB scores and the responses to a

vocational history and interest questionnaire. Defining his low-profile

subjects as those having no A scores and no primary patterns as freshmen

and his control group as every nth remaining person in the freshman

class, he tested, among others, these hypotheses:

1) There is no evidence that low profiles result from random
responding to the Blank.
2) The item response patterns of men having low profiles will
differ from those of an appropriate comparison group of men
not having low profiles.
3) Low profiles will be as reliable upon retest as the profiles
of a comparison group.
4) The item responses associated with low profiles will be as
stable upon retest as the responses of a comparison group.
5) Men having low profiles will not have interests characteristic
of those in occupations for which there are no existing scales.

8) The interest'scores of'low-profile men will change'up'on retEt
in ways similar to those of a comparison group, and in ways
consistent with the maturationall" changes reported in other
studies. (Athelstan, 1967, p. 21)

Athelstan admitted that not all of these hypotheses could be tested

formally, but could be evaluated only "indirectly by considering various

kinds of evidence relating to each of them" (p. 21).

in his data he found support for hypothesis 1) by examination of

the answer sheets for "obvious" random answering or "responding in any

other systematically unusual fashion" (p. 82). Hypothesis 2) was sup-

ported in that the low-profile sheets had a greater proportion of Dis-

like responses and fewer Like and Indifferent responses than did the


control group sheets. Hypothesis 3) was supported by the finding that

the median test-retest individual profile reliability coefficient for

the low-profile group was only slightly lower than that for the control

group. Hypothesis 4) was not supported; the responses of the low-

profile subjects to many items changed considerably on retest. Hypothe-

sis 5) was given indirect support in that a rescoring of the blanks of

the low-profile subjects on ten new scales which had just become avail-

able yielded only a "few A scores, which could as well have resulted

from scoring changes as from the use of new scales" (p. 83). Hypothesis

8) was interpreted as being supported, although the evidential basis for

this decision was not clear. In fact, at an earlier point Athelstan had


The present distribution of primary patterns among the low-profile
subjects also differs from the original distribution exhibited
four years earlier by the control group, suggesting that the low-
profile subjects are not simply catching up with late-maturing
interests. (P. 73)

This statement seems to contradict the position of hypothesis 8) that

changes would be of a maturational nature.

Athc'Istan also interpreted his data as supporting Strong's (1943)

idea that riany low-profile men should go into business areas. Again,

it is not clear how this conclusion was drawn, for while his data

reported that 32 percent of his retested low-profile group had primaries

in area LX (Business Contact), only 10 percent had primaries in area VIII

(Business Detail) while 27 percent and 26 percent had primaries in areas

X (Verbal-Linguistic) and V (Social Service) respectively. Taking into

account that the largest change in scores in the low-profile group was in

area V, it seems more important to interpret the number of primaries in

area IX as having to do mere with interest in people than with interest

in business.

Finally, Athelstan added to those researchers3 who have found no

relationship between the MPI and the SVIB; his two groups did not dif-

fer on mean scores on any of the common MPI scales.

Since Athelstan's study, two additional studies have been done

which have dealt with the problem of low profiles. The first, by

Deutscher (1966), was a study which managed to use a personality inven-

tory other than the MPI: the Edwards Personal Preference Schedule

(EPPS). Deutscher used the EPPS as well as measures of academic ability

(the American College Tests) and a personal information inventory to

examine high-intensity profiles (those having one or more primary inter-

est patterns) and low-intensity profiles (those having no primary pat-

terns), as defined by Korn and Parker's rules (1962). He found that the

two kinds of profiles did not differ on the 'CT, but they did reveal

different manifest need structures. The lower intensity subjects ap-

peared to have lower needs for deference, order, abasement, and nurtur-

ance and a higher need for autonomy. Also, the personal information

inventory revealed that the low-intensity students had more mothers with

college degrees than did the high-intensity students, and the low-inten-

sity students reported more often than the high-intensity students that

they were attending college for social enjoyment.

Dolliver's (1966) study dealt only incidently with the problem of

low profiles. In a study which was primarily concerned with the agree-

ment of inventoried and expressed interests, he hypothesized that the

3Anker, Tonscnd, and O'Connor (1963) factor analyzed scores on the
MMPI, Work Values Inventory, SVIB, and various decision-making ability
scores and also found no factors which contained NMPII and SUB scales


more spread the interests on the SVIB the more personal confusion the

person would manifest. This hypothesis failed to be confirmed, but

Dolliver did find that among those persons whose expressed and inven-

toried interests agreed there were more Dislike responses and a greater

number of occupations scoring low.

It should be noted that all of these studies have treated the

problem as a class distinction: those subjects with no primary patterns

have been compared with subjects with one or more primary patterns.

This treatment of the problem as a distinction among groups has some

difficulties with it. For one thing, it is very wasteful of subjects,

in that the percentage of men with no primary patterns ranges from

around 6 to 40 percent (cf. Table 2 in this study). For example, Athel-

stan's study reported that only 151 out of 1535 students met the criterion

of having no primary patterns and no A occupations, while only another

200 of those having primary patterns or A's were used as a comparison

group. Some 1100 records were not used in this study.

Another difficulty encountered in using criterion groups, is,

as Athelstan pointed out, the problem of defining the groups. Some have

defined their criterion groups as including those with no primary pat-

terns; some have used no primary nor secondary; some would have it simply

no A occupations, ignoring the possibility of having primary patterns on
"no-A" profiles; and Athelstan decided on no primary and no A occupa-

tions. Further, these contradictory ways of defining the criterion

groups do not produce simply subsets of one another; they produce groups

which include different subjects, making them qualitatively different

from one another.

It seemed eppropricte to attempt to bypass these problems by using


criterion variables in place of criterion Rroups. The problem becomes,

then, not simply what is it that causes a person to have no primary

pattern (or no secondary pattern or no A occupations, etc.) on the SVIB

but what characteristics of the person are related to profile height on

the SUB? What is there about a person which allows him to have many

SUB occupational interests, and what about the person tends to limit

the number of occupational interests on the SVIB? Stating the problem

in terms of variables instead of groups allows full use of the subject

population and does not allow one to wonder if the wrong subjects have

been included or excluded from the criterion groups.

Cognitive Style Literature

Most of the studies cited so far as attempts to uncover relation-

ships between patterns on the SVIB and personality variables would prob-

ably be labeled as studies of personality dynamics, as opposed to person-

ality structure, according to the distinctions made in the Annual Review

of Psychology in the last seven years (cf. Messick, 1961). Studies of

personality dynamics are supposed to have been concerned with "motives,

needs, drives, and the functional interplay of forces involved in

transient relations with the environment' (Klein, Barr, & Wolitzky,

1967, p. h67), while personality structures have been interpreted as

being "stable, relatively enduring components of personality organiza-

tion that are invoked to account for recurring similarities and consis-

tencies in behavior over time and over situations" (Messick, 1961, p. 93).

Wiggins (1968) points out that a great deal of the content of the

structuralist position is furnished by the traditions of Eysenck,

Cattell, and Guilford, with a concomitant methodological emphasis upon

multivariate proondu-es of analysis, especially upon factor analysis.


However, one notices in both Wiggins' (1968) and Klein, Barr, and

Wolitzky's (1967) reviews that another large source of content for

structural studies is that furnished by the various "cognitive style"

theorists. Though some attempts have been made to relate this large

area of personality-structure investigations to the SVIB, as yet no

attempt has been made to investigate how the various cognitive style

variables may account for some of the variance in interest intensity on

the SVIB.

The concept of cognitive style (also called cognitive attitudes,

cognitive abilities, cognitive strategies and cognitive control varia-

bles) seems to have been used very early by Klein (1951), who remains

today one of its most eloquent spokesmen (Klein, Barr, & Wolitzky, 1967).

In this latter article Klein seems to be moving away from the psycho-

analytic ego-psychology in which he started toward a more general infor-

mation-processing point of view. He still sees the cognitive structures

with which he is concerned as having "the steering functions of regulat-

ing, promoting, or dampening adaptive feedback" (1967, p. 483), but he

sees his approach today as essentially a "functional" one. Regulative

constancies such as "personal construct" or "field-articulating tendency"

should be viewed in "terms of the achievement of an optimum adaptive

balance of some kind in relation to the demands of both the situation

and motivating tendencies" (p. 484).

A regulative mode thus functions as a cognitive transform, giving
to the "stimulus" situation its meaning and, therefore, its back-
ground and foreground characteristics, as well as both conscious
and unconscious aspects of response. In its application to the
problem of personality constancy, cognitive control implies ways
of dealing with situations--attitudes or orientations--that become
organized into economical strategies of encounter. Some of these
are of high generality, others are of low generality, and research
on p-rsonality is the study of this heirarchy. (pp. 484-485)


It is Klein's further belief that these accommodative patterns are to be

found in the behavior of any person who has the intention (not necessar-

ily conscious) of making a realistic appraisal and assesment of a

situation (Klein, 1958).

There are several other groups actively engaged in cognitive

personality research, many of whom, like Schroder, Driver, and Streufert

(1967), understand the cognitive processes to include orientations that

act like "filters" in selecting certain kinds of information from the

environment and to include something like a program or set of rules

which regulates how the information is transformed and combined in spe-

cific ways. A crucial aspect of this processing of information is the

way it is grouped into categories through various decision processes.

Here Bruner (1957) has provided seminal theorizing for understanding the

importance of this process. His rallying cry of "No perception without

categorization:" stresses the unity of the cognitive and perceptual

processes, and even makes such areas as signal-detection theory relevant

to cognitive processes.

It may be possible to increase our understanding of the SVIB by

analyzing its stimulus-demand characteristics in terms of this general

cognitive theory. Basically, the task confronting a person taking the

SVIB is that of making decisions (or recalling decisions already made)

about 399 items concerning vocational and avocational interests, school

subjects, personality types, etc. Most of the items require free cate-

gorization into the categories of Like, Indifferent, or Dislike, but

others require the person to rank various items according to how inter-

esting they are, to compare the intensity of interest on tTo kinds of

activities, etc. In any case, the basic process seems to be a decisional


one, requiring the person to attend to a particular kind of activity or

person and to sort through his memory for stored information about his

emotional reactivity to experiences with the particular activity or

person. In some cases (i.e., with some items) the information is imnedi-

ately available and may even produce a secondary emotional reaction upon

its introduction into the attentional center. In other cases the person

may have had no experience with the item and may have to generate some

kind of emotional reaction by searching various memory items which some-

how seem associated with the particular one being attended to (this is

not to imply that this is a conscious search nor that in the first case

associations are not furnishing information to the decisional process; in

these hypothetical distinctions the point is, which seems to furnish the

major part of the information for the decision--the primary stimulus or

the associations to it?). In any case, the basic decision to be made

(for most items) is whether the emotional reaction is one that can be

classified as a "liking," one that can be classified as a "disliking," -

or one that seems so v:eak that it can be safely assumed that it fits

into an "inlifferent" category.

Moro is required, however, of the subject than just being able to

respond emotionally to the different items of the SVIB if he is to end

up with a profile which displays positive vocational interests: he must

match the interests of a particular occupational group. However, this

match of interests is not accomplished in the SVIB in an absolute way,

as the manual points out: "... the measures of interests by the SVIB

depends on differences in interest patterns between occupational groups,

not on any absolute indication of interest" (Strong, 1966, p. 26). This

is true because ". .. the items that are weighted on an occupational

scale are not necessarily the ones that men in an occupation select most

often--rather they are items that men in that group answer differently

from other tren" (ibid.). The "other men" in this process were, as the

manual points out, all of the other men used in the standardization of

the other occupational scales, and this group was given the label "men-

in-general," or "MIG."4 For an individual to score high on a particular

occupational scale, he must answer the specified items in a way different

from men-in-general and in a way which matched the typical response of

the men in that particular occupational standardization sample. When a

large number of occupational scales receive high scores from an individ-

ual, then it means that he rather consistently differentiated himself

from the men-in-general group and that he was capable of including sev-

eral different patterns in his responses.5 If he does not differentiate

4Yet another group of men called men-in-general was involved in
establishing the validity statistics of the SVIB, according to the manual.
The use of the first, larger group for the item analysis weighted all
the occupations involved equally. The occupations in the second group
of men-in-general, many of whose representatives were selected from the
first group, were weighted approximately according to their frequencies
in the general college-educated population. After the items for each
scale were selected through the use of tha first group, the scales were
used to score the responses of the men 3v the second group, and a statis-
tic was calculated for each particular scale describing how well the
scale separated the average score on a particular scale of the men in
that scale's standardization sample from the average score of the
weighted men-in-general group on the same scale. This statistic was used
to help determine if a scale functioned well enough to be included in
the SVIB. In addition, the range of the middle third of scores of the
second men-in-general group on each scale is now indicated on the 1966
SVIB profile sheet by a shaded area; formerly the shaded area was the
middle two-thirds of the distribution of chance scores on each scale
determined by using dice throws. The present discussion is concerned
only with the process involving the first men-in-general group.
5It is possible to score high on mrany different SVIB scales, be-
cause each scale vses only a portion of the 399 items (from 35 to 98 in
the 1966 revision). Thus those interests of the different occupations
which arc contradictory do not necessarily involve cormnon items which
would have to be scored in contradictory directions.


his responses from those of men-in-general or if his interest patterns

are not any of those of the inen in the standardization samples, then he

will end up with what has been called a low profile on the SVIB.

From a cognitive point of view, the requirement that the individ-

ual, in order to produce a high scale score, be able to give his respon-

ses in fairly well-articulated patterns suggests that his information

receiving, storage, and retrival systems must be differentiated to some

minimal degree. In order to have differentiated responses to the indi-

vidual interest items, he must be able cognitively to make the distinc-

tions among the different interest elements to which he is asked to

respond. That is, the more an individual tends to make only global or

gross distinctions among these items, the less selectively he would be

able to respond to the items. If the items which weight a particular

SIB occupational scale require finer distinctions among interests than

the individual prefers to make, or is capable of making, then he will

not be able to obtain a high score on that particular scale, even if he

claims interest in that particular occupation. However, the condition

of being cognitively differentiated does not necessarily mean that the

individual will make a large number of high occupational-scale scores.

It is hypothesized that cognitive differentiation is a necessary but not

a sufficient condition for matching the different occupational interest

patterns. Another necessary condition is that the individual have rea-

sons, or motivations, for patterning his differentiated interests ele-

ments in particular fashions. These motivational conditions are not of

primary interest at this point--they have been investigated by many

other studies.

A general hypothesis of this study was, then, that the "cognitive


map" a person has of his vocational, avocational, school subject, per-

sonality, etc., worlds must be first of all differentiated and then

organized in a particular fashion if he is to match a scale on the SVIB.

More generally, the person must have a cognitive structure which permits

this differentiated cognitive map, or have cognitive styles which permit

the erection of a differentiated cognitive structure. The less differ-

entiated his cognitive structures, the less capable he should be of

producing the differentiated responses necessary to match an SVIB pat-

tern, no matter which particular occupational-scale pattern this may be.

Possibly further complicating the responses to the SVIB are the
"risk" characteristics of the task. That is, are there desirable con-

sequences to be gained and undesirable consequences to be avoided by

means of answering the questions? Perhaps these risks can be seen as

being of two general kinds: the risks attendant upon gaining information

about oneself, and the risks attendant upon giving important-others

information about oneself. Most late adolescents (and perhaps most

people) seem to experience some anxiousness about psychological tests,

with perh; ps the main concern being whether or not the test will reveal

the person to be abnormal in some way.

This risk involved in gaining information about oneself can be

related to taking the SVIB in at least three ways. First, there is the

information one gains about oneself in answering the Strong item by

item: whether it is easy or hard to decide the items, whether one has

heard of these occupations or not, etc. Second, a person can get infor-

mation from his overall impression as to how he is answering the items

as he progresses through the test (when this writer took the Strong he

began to have ths feeling as he was going through it that he was


"indifferent" to too many items, that he maybe was "wishy-washy"). The

final risk is the unknown way the test results will categorize the per-

son, as for example, failing to support an interest in being a physician

or indicating the occupation "one should be in" is really quite undesir-


The other major risk involved in taking the SVIB is that of letting

some important other person know some possibly unflattering information

about oneself. In answering the items, then, the person must choose

between trying to be honest and possibly revealing some undesirable

information about himself or of trying to present himself in a socially

desirable light but at the same time risk not receiving any helpful

information about himself.

Whether or not these risk and cognitive task characteristics of

the SVIB influence in a systematic way the patterns of occupational

interests is something that has to be determined empirically. Morris

(1966), for example, found that "propensity for risk taking" influenced

strongly the way high school seniors made hypothetical choices of voca-

tions, within which they had earlier rated their probabilities of

achieving success. Mahone (1960) found that those students who were

fearful of failure (and thus not willing to risk much) were more likely

to be judged as having unrealistic vocational aspirations than students

less fearful of failure. However, this "debilitating anxiety" did not

discriminate between those who did and did not predict accurately their

primary and secondary patterns on the Strong (as, incidently, scored by

Darley's 1941 unclear rules).

Not many studies have been undertaken relating cognitive variables

to performance on the SVIB. Witkin, famous for his Field Independent-


Dependent variable (now being called Articulated versus Global Field

approaches), thought that a person with a non-articulated cognitive

style might come out with Service Area type occupations on the SVIB

(Witkin, Dyk, Faterscn, Goodenough, & Karp, 1962), but evidently he had

little data to back this up. Pierson (1965) reported in his dissertation

that he had found some relationships between cognitive styles and vari-

ous of the broad vocational area scales of the pre-1966 SVIB. Field

Independence (Witkin's variable) was related positively to Physical

Science and Technical Sales but negatively to Sales Interests and Verbal-

Linguistic interests. Pierson concluded that Field Independence was

related to the impersonal types of occupations, while Field-Dependent

subjects were more interested in interpersonal concerns.

Chung (1967) in his dissertation looked at the relations between

vocational preferences and identification and the cognitive styles

called Field Independence-Dependence, Leveling-Sharpening, Constricted-

Flexible Control, and Equivalence Range (all variables studied by

Gardner, Holzman, Klein, Linton, & Spence, 1959, and Gardner, Jackson,

& Messick, 1960). Using discriminant analysis he found that scores on

the various measures of the cognitive variables could discriminate among

the eight areas of college majors, but he could find no relationship

between Kuder scores, vocational commitment or vocational identification

and the cognitive style measures.

Levine (1967) in his dissertation chose to look at only one vari-

able of cognitive style, that of Constricted-Flexible Control (CF), and

he attempted to assay how it related to the cognitive handling of impul-

ses in three different vocational groups. He concluded that the instru-

ment traditionally used to measure CF (Gardner et al., 1959) really


measures some kind of ability to tolerate stress and not a particular

style of coping.

This not very extensive literature indicates that not a great deal

of thought and study has becn attempted in relating cognitive variables

to vocational interests, let alone characteristics of the SVIB profile.

The attempts at providing a theory relating the two seem to be limited

to psychoanalytic ego-psychology (Levine, 1967) and to some kind of
"value dimensions which were reflected in the cognitive styles and voca-

tional interests" (Pierson, 1965). In fact, as has already been inti-

mated, the SVIB has not been well integrated into any kind of established

personality theory, but has more or less rested its case upon the empir-

ical statement that "it works." Thus it may be premature to attempt, as

did Levine, to unite cognitive theory and vocational choices through a

carefully reasoned, deductive model which can economically assay only

one hypothesized relationship. Since these relationships are just begin-

ning to be explored, it may be better to attempt, as did Chung (1967),

a more haphazard but more general study which would include a variety of


It vas felt that if a more preliminary kind of study which would

involve a variety of cognitive measures was to be undertaken, then again,

in the interest of economy of research, it would be important that these

measures do not involve a great deal of individual attention to each

subject. Similarly, if it were conceivable that cognitive measures may

be able to account for some of the flatness of the profiles on the SVIB,

then it would be important that these measures be easily administered by

counselors or testing assistants. For a preliminary investigation it

thus seemcd appropriato to choose instruments which could b3 group-

administered and which could be scored easily without any required

expert judgment.

With these thoughts in mind a survey was made of tho various cog-

nitive-variable research traditions for pertinent instruments and vari-

ables, and the following discussion includes those which were felt to

be promising in attacking the problem of varying SVIB profile flatness.

One of the longest traditions of working on "cognitive structures"

is that of Rokeach (1951), who considered himself to be working in the

traditions of Tolman and of Lewin. Culminating his early work on
"narrow-mindedness" was his book, The Open and Closed Mind (Rokeach,

1960), and the publication of his Dogmatism Scale. Another long tradi-

tion, that of Witkin and his associates, stems from laboratory research

in perception (Witkin, 1949; Witkin, Lewis, Hertzman, Machover, Bretnall

Meissner, & Wapner, 1954). Using a variety of perceptual instruments to

measure the dimension called Field Dependence-Independence, Witkin and

his associates have since broadened their concern to the more general

"Global-Differentiated cognitive style" (Faterson, 1962; Witkin, 1965).

The study of another dimension of cognitive functioning, Complex-

ity-Simplicity, was reported by Barron (953; Barron & Welsh, 1952).

No theoretical starting point for this dimension existed, but rather it

sprang from the factor analysis of responses to an art-preference test

(Welsh, 1949). The term "Complexity-Simplicity" was rendered less clear

when Bieri took it over (with no qualifying adjectives) to label his

work on interpersonal constructs, in the Kelly tradition (Bieri, 1955).

Yet another group moved into the area with different instruments and

from a different theoretical background (Lewinian, as with Rokeach) to

add to the confusion (Harvey, Hunt, & Schroder, 1961). Finally,


Vannoy (1965) collected a large number of these instruments purporting

to be measuring this Complexity-Simplicity dimension and through factor

analysis found, as many suspected, that there was no single large fac-

tor common to these tests, but rather there seemed to be at least

eight different factors among the measures.

Another long tradition of cognitive work and yet another theo-

retical orientation can be found in the research of Klein and Gardner

and their associates at the Menninger Foundation. Stemming from psycho-

analytic ego-psychology, their work has dealt with variables they have

called cognitive control variables, i.e., variables which regulate the

style with which a person adapts to his environment in his attempt to

satisfy his drives. They have included Witkin's Field Independence-

Dependence variable as one of their six cognitive control variables

(Gardner et al., 1959).

Another more loosely related group, which has involved men work-

ing at the Educational Testing Service, at Pennsylvania State, at

Duke, and other places, has included Messick, Kogan, Jackson, and

Wallach. Seemingly followirng no particular theoretical orientation,

they have researched broadly throughout the cognitive area. Two of

them, Jackson and Messick, have worked with Gardner (Gardner, Jack-

.son, & Messick, 1960), and have continued work on Gardner's (1953)

own unique contribution to cognitive variables, Equivalence Range

(Clayton & Jackson, 1961; Sloane, Gorlow, & Jackson, 1963; Messick &

Kogan, 1963). Closely related to this group's work has been the devel-

opment of a quantative measure of Category Width by Pettigrew (1958) and

the comparison of several measures of Equivalence Range by Fillenbaum

(1959) t is also of interest to note how often these investigators


have cited Bruner as being of particular encouragement (cf. Pettigrew,

1958), and to note that Bruner was also influential on the early work

of Klein (cf. Smith & Klein, 1953).

Finally, a fairly recent entrance to the list of cognitive vari-

ables has been Rotter's variable of the expectation of Internal-Exter-

nal Control of reinforcements (Rotter, 1966; Lefcourt, 1966). This

variable, too, can be thought of as a style variable, a way of cog-

nizing the world.

This survey of the relevant cognitive research literature turned

up five general variables which it was felt could be influencing profile-

flatness on the SVIB. As will be noted, it was felt that the probable

effect of most of these variables would be accomplished through the

general concept of cognitive differentiation. That is, as was already

pointed out, in order for a person to have an A occupation on the SVIB

he must answer the questions in a way which differentiates his responses

from those of men-in-general on a particular occupational scale. And it

was proposed that in order for a person to make such differentiated

responses he must have an underlying differentiated cognitive map of the

relevant vocational, personality, avocational, etc., areas which are

tapped by this test. Any cognitive style which promotes this kind of

differentiation should also promote occupational differentiation or high

occupational profiles on the SVIB. Any style which hinders such cogni-

tive differentiation should also make it more unlikely that an individual

will produce a differentiated pattern of answers which will closely

match any of the given patterns of the occupational scales. As mentioned,

at least five general cognitive style variables were found which would

seem to be related to this cognitive differentiation.

Hypothesized Relationships

Predictor Variables

Field-Dependence-Independence. As mentioned, this variable is now

claimed to be a perceptual manifestation of a broader dimension of per-

sonality, the articalateness-of-experience dimension (Faterson, 1962).

The variable is measuring the individual's ability to keep an item sep-

arate from a field or embedding context. A variety of instruments have

been used to measure this dimension, including the rod-and-frame test, the

tilting chair, and an Embedded Figures Test (EFT) (Witkin, Dyk, Faterson,

Goodenough, & Karp, 1962). Only the last instrument could be cheaply

produced, and it became much more useful when a group version was devel-

oped (Messick & Fritzky, 1963) and found to correlate very highly with

the individually administered version (Jackson, Messick, & Myers, 1964).

The contrasting styles which a test such as the EFT is getting at include,

as Wallach (1962) has described it, on the one hand a style of cognitive

functioning that is active, analytical, articulated, specific, and crit-

ical, and, on the other, a style that is passive, global, vague, diffuse,

and uncritical. The articulated" person has, according to Witkin (1965),

a better sense of separate identity from his environment and from other

people; thus, his structural self is more articulated also. One could

hypothesize that the more field-independent person would also experience

his vocational environment in a more differentiated fashion than would a

field-dependent person. The field-dependent person would see less dif-

ferences between vocations and would thus have more difficulty experi-

encing differentiated emotional reactions to the various vocations.

Other things being equal, the field-dependent person should have fewer


pri-rary interest patterns on the SVIB, because he does not have the

differentiated interests necessary to match a particular occupational


CoMPexity-Simlicity. As was mentioned, the first instrument used to

measure this dimension was the Barron-Welsh Art scale (Barron & Welsh,

1952), but an agree-disagree scale was also developed by Barron from

his Independence of Judgment scale which correlated highly with the

Art scale (Barron, 1953). In Vannoy's (1965) factor-analysis study of

the various instruments which had been used to measure this dimension,

Barron's Complexity-Simplicity (CS) scale was located on Factor I

along with a similar kind of scale by Budner (1962) called the Intol-

erance of Ambiguity scale. Since the former loaded .72 and the latter,

-.68 on Factor I, it was proposed for this study to reverse the scoring

on Budner's scale and combine the two for a measure of what Vannoy termed

Complexity of Verbal Apparatus. The possession of a more complex verbal

apparatus, he said, seems to be conducive to, or correlated with, a more

varied and possibly more "equivocal" interpretation of experience. In

Berkowitz' (1957) investigation of the relationship between Klein's vari-

able of leveling tendencies and Barron's attitude scale on CS, evidence

was found for the hypothesis that "individuals preferring simple, order-

ly phenomenal experiences tended to achieve this simplicity by leveling,

that is, by forgetting some of the details of prior experiences" (Berk-

owitz, 1957). It would seem, then, that CS would also affect the per-

son's way of categorizing his knowledge about vocations, this time more

explicitly in terms of verbal categories. A more verbally complex per-

son would have more differentiated patterns of response available to a

verbally constructed test such as the SVIB. In so far as a person is not


capable of responding differentially, then his interest profile on the

SVIB should be flat, because he is not making the distinctions among

occupations necessary for him to match the distinctive ansi:ers of the

standardization sample of a particular occupation.

Eguivalence Range. This dimension has to do with the person's style in

the size of categories he forms, i.e., how many different objects he is

comfortable in saying belong together. Gardner et al. (1959) used an

Object Sorting Test (OST) to measure this dimension, scoring simply the

number of different categories a person used in grouping together a

variety of objects. As mentioned, Pettigrew (1958) devised a scale (CW)

to measure a similar dimension, and attempts were also made to relate the

CW scale to Complexity-Simplicity. It seemed reasonable that the person

who used narrow categories needed many of them and thus needed a more

complex cognitive structure. However, Vannoy (1965) found that CW with

the Quantative score on the SCAT formed a separate factor in his study,

so CW was poorly related to the other measures of complexity. Fillenbaum

(1959) correlated several of these equivalence range tests as well as

some new cr:es of his orn devising (such as the Synonymity Task) and

found then to be more independent than he had anticipated. Clayton and

Jackson (1961) developed a group version (paper and pencil) of Gardner's

OST, with which Messick and Kogan (1963) discovered the necessity of

obtaining at least two different scores from sorting tests, and which

Sloane, Gorlow, and Jackson (1963) demonstrated could be substituted

for the original, individually administered test.

It was proposed for the present study to use several measures of

equivalence range in an attempt to account for low intensity of inter-

ests on the SVTB. As Sloane et al. (1963) pointed out, equivalence

range scales can be considered to be measuring a "preferred mode of act-

ing so as to maximize similarities or differences among objects." If

the style of maximizing similarities between objects is generalizable to

maximizing similarities between vocations, avocations, etc., then the

person with such a style will again not have the requisite differentiated

responses in matching the responses of the standardization samples of

the various occupational scales. Those who tend to maximize differ-

ences, or who have smaller equivalence ranges, should have greater

chances of producing the differentiation of responses necessary to match

particular occupational scales.

Internal-External Control. Though not often labeled as a cognitive style

variable, Rotter's (1966) personality dimension (I-E) seems to function

in a manner similar to these variables. In general, the dimension taps

the person's belief as to how much his own behavior produces the posi-

tive and negative events of his life, or, how much control he believes

he has over his own reinforcements. This basic cognitive set seems to

influence the way the person "knows" other aspects of his environment.

For example, Lefcourt (1966) produced a great deal of evidence to show

that Southern whites feel much more in control of their reinforcements

than do Southern blacks. Another study (Davis & Phares, 1967) sought

the relationship between information-seeking and the I-E variable and

found some support for the hypothesis that Internals sought more rele-

vant information than did Externals to whom the information appeared as

relatively useless. Liverant and Scodel (1960) found that Externals,

in a betting situation, played hunches on previous outcomes more than

did Internals, who seemed cautious and played a planned selection of

probabilities, presumably as a way of maintaining control. Finally,


one study has sought correlations between I-E and the SVIB scales and

found that the Artist, Musician, and Author-Journalist scales correlated

most highly with external control (the occupiers of these vocations have

a greater belief in the role of chance and luck in their lives) while

a scattering of various other technical and administrator scales corre-

lated most highly with the expectancy of internal control (Zytowski,

1967). The correlations in this study were, for the most part low, and

the correlations with the internal-control end of the I-E scale did not

make a very meaningful pattern. Nevertheless, it appeared possible that

I-E could have some general relationship with vocational choice through

differing motivations to collect vocational information. It would

appear that Internals would have more reason to seek and attend to in-

formation about vocations than would Externals, who would probably per-

ceive vocational choice and advancement more as a matter of luck than as

a matter of choosing and planning for. Again, other things being equal,

the person who has more finely articulated information about vocations,

etc., will probably be more capable of making the distinctions on the

SVIB necessary to match a particular occupational scale.

Dogmatism. Rokeach's (1960) Dogmatism "cale was intended to be a measure

of the relative openness and closedness of conceptual systems. Kaplan

and Singer (1963) have suggested further that a highly dogmatic person

also lacks self-awareness and even suffers from an alienation from his

sensory modalities. Finding such a general variable cutting across

these different information-processing realms warrants calling this

dimension a dimension of cognitive style also (cf. Witkin, 1965). Its

relationship to the SUIB profile was difficult to predict, but it seemed

highly probable that its effect would be oderited by the presence of


another variable, level of occupational aspiration. That is, a rela-

tively closed mind would produce a very sharp peak on the profile if

the person were professionally minded; if he had interests more in the

"blue-collar" jobs, then it is possible that his rejection of the SVIB

occupations would be more complete, yielding a lower profile, than would

the rejection of a more open-minded "blue-collar" aspirer.

Specialization Level. In addition to the occupational scales on the

SVID, there are four non-occupational scales, Specialization Level,

Occupational Level, Masculinity-Femininity, and Academic Achievement.

Of these four, it seemed possible that the first one, Specialization

Level, was tapping a kind of cognitive differentiation. It was devised

by comparing the answer patterns of medical doctors who were interested

in very specialized fields of medicine with those doctors who were inter-

ested in general practice. A high score could represent, then, an inter-

est in stressing differences among things, while a low score could sug-

gest an interest in similarities among, and generalities about, objects

in the cognitive world. Though Super and Crites (1962, p. 455) have

argued that, "Research in the 'undifferentiated' group both in college

and elsewhere should presumably be pressed, using other points of ref;r-

ence than that of the standard scales," it was felt to be important that

some understanding be gained of the relationship of this scale to the

other cognitive scales as well as to profile height.


It may be helpful at this point to summarize the predicted rela-

tionships between the various cognitive-style variables and SVIB profile

height. The general concept through which most of these variables were

hypothesized to interact with the SVIB was the person's ability to make


patterned distinctions among the various vocational, avocational, etc.,

items of the Strong. Any style which hinders a person from making such

distinctions should increase the probability of the person obtaining a

flat profile on the SVIB; conversely, other things being equal, any

style which helps a person to make distinctions should permit the person

to obtain more primary interest patterns on the SVIB.

These hypotheses were admittedly general and were to be thought of

as being preliminary in nature. They were intended as guidelines in

searching for more specific hypotheses, not only relating each of the

individual cognitive-style variables to SVIB profile height, but also

interrelating the variables in complex interactions with each other as

the complex cognitive structures affect interest intensity. The plan for

this study was to follow a design similar to Noonan's (1967), which used

initial hypotheses to guide a search for more specific hypotheses in an

initial sample of subjects. Upon analyzing the data from the initial

sample, Noonan noted the more subtle relations between the predictors

and the criteria and used these as refined hypotheses to cross-validate

on a final sample of subjects.

The initial hypotheses were:

(1) SVIB interest profiles are partially a function of the extent

of the field dependence or independence of the subject. The more

field dependent, or less well-articulated the subject is, the less

articulated or differentiated he experiences the vocational world,

and he is, therefore, less capable of producing the differentiated

responses necessary to match occupational scales on the SVIB.

Other things being equal, the more field independent the person

is, the more he will be able to match the occupational scales on

the SVIB.

(2) SVIB interest profiles are partially a function of the extent

of the person's preference of cognitive simplicity or complexity.

The more a person simplifies his cognitive structure the less

capable he is of responding differentially in a manner which will

match occupational scales on the SVIB. The more a person prefers

cognitive complexity the more capable the person will be of per-

forming such matching.

(3) SVIB interest profiles are partially a function of the broad-

ness of the categories a person prefers to use. The broader the

categories he prefers to use, or the more he maximizes similari-

ties among objects, the less capable he will be of responding

differentially to the SVIB and, thus, of matching any particular

occupational scale. The narrower the categories he prefers to

use, or the more he maximizes differences, the more capable he will

be of differentially responding.

(4) SVIB interest profiles are partially a function of the extent

to which a person believes luck or chance plays a role in his life.

The more a person sees his reinforcements as being externally

controlled the less reason he has to attend to vocational differ-

ences and the less capable he is, therefore, of responding in a

differentiated fashion to the SVIB in order to match a particular

interest scale. The more a person sees his reinforcensnts as

being internally controlled the more reason he has to differen-

tiate his Iknoldge of vocations (in order to make a wise choice)

and tho more capable he will ba of differentially responding to

the SVIB.


(3) SVIB interest profiles are partially a function of the extent

to which a person's conceptual system is open or closed and of the

extent to which the person's aspirations are on more professional

or on more non-professional levels. Among those who aspire to

lower occupational levels, those with relatively closed systems

should reject more completely the various professional alternatives

on the SVIB and thus end up with a lower average profile height

than those with relatively open systems. It is not certain how

the open and closed aspirers to high occupational levels will dif-

fer on profile height.

As mentioned, these hypotheses were intended to be used as guides

in investigating the relationships between the cognitive differentiation

variables and SVIB occupational profile height. Through investigating

these relationships in a preliminary sample, it was hoped that the var-

ious possible interactions between the predictor variables and the cri-

teria would become clearer, making possible the formulation of more

specific hypotheses as to how the various dimensions of cognitive dif-

ferentiation would interact either to produce flat interest profiles or

to allow high interest profiles on the SVIB.



This study began when responses were obtained from a large number

of male university students to the Strong Vocational Interest Blank For

Men (SVIB) and to a battery of cognitive-differentiation tests and

various control tests. After the data were collected from these students,

a mixture of beginning psychology course "volunteers" and paid volunteers,

the tests were scored and the subjects' data randomly divided into an

initial sample and a cross-validation sample. Intercorrelations among

all variables were obtained from the two samples, as well as canonical

correlation analyses. A moderator analysis was performed between the

variables in the initial sample, and those moderated relationships found

to be statistically significant in this initial sample were examined in

the cross-validation sample for significance. These procedures are

explained in further detail below.

Selection and Description of Instruments

Predictor Instruents

Field Dependence-Independence. In this study the Field Dependence-Inde-

pendence instrument was the group version of the Embedded Figures Test

(Messick & Fritzky, 1963) as altered by Jennings (1968) and further al-

tered for this study. It consisted of sixteen complex figures, each of

which contained one of five smaller, simple figures within its design.


The subject's task was to decide which one of the five simple figures was

within each of the sixteen complex figures and to mark his decision on a

multiple-choice answer sheet. It was a timed test, with a limit of six-

teen minutes to complete all sixteen figures. The score was the number

of embedded figures correctly identified and thus was a measure of


Complexity-Simplicity. As was pointed out in Chapter I, a large variety

of instruments exists purporting to measure this dimension. In this

study two instruments were combined to measure Vannoy's (1965) first

factor of Complexity-Simplicity, Complexity of Verbal Apparatus. The

instruments were Barron's (1953) twenty-two item Complexity-Simplicity

attitude scale and Budner's (1962) twenty-four item Intolerance of Ambi-

guity scale. Both of these scales are agree-disagree, or true-false,

scales, but they were loaded in opposite direction on Vannoy's Factor I.

It was necessary, then, to reverse the scoring on the Intolerance of

Ambiguity scale in order that its score cmld be added directly to the

score of the Barron attitude scale. Thus, tolerance of ambiguity was

added to preference for complexity to yield a total score measuring Con-

plexity of Verbal Apparatus. Since the two scales were combined in this

fashion, only the total score was computed and used in this study.

Equivalence Range. This dimension also had been measured by a variety of

instruments, most of which were fairly simple to administer and score.

The instruments which were selected for this study were:

(1) The Category Width Scale. This is a twenty item scale (Petti-

grew, 1958) composed of statements about various statistical aver-

ages for a variety of events, e.g., "The average annual rainfall

for New York has been 410 inches." The subject is then asked to


choose an estimate of what the greatest and the least amount of

the event has been in a given time (with two scores per item,

forty scores are thus obtained from this scale). The choices

are weighted so that those furthest from the average receive the

most points while choices closest to the average receive the

least points. A greater number of points thus indicates a use of

broader Category Widths, or a greater equivalence range.

(2) The Synonymity Task. This instrument consists of thirteen

statements with an adjective in the statement singled out for

replacement by other, almost synonymous adjectives. The subject's

task is to indicate how many, from the list of adjectives for

each statement, he thinks could be substituted for the original

adjective without substantially changing the meaning of the sen-

tence. Fillenbaum (1959) reported a reliability of .92 for his

instrument with a range of scores from 12 to 88 (variance--389),

indicating it had the properties desirable for measuring this kind

of Verbal Equivalence Range. Since the scores are obtained by

simply counting the number of adjectives circled, the higher the

scor-e, the broader the Verbal Equivalence Range of the subject.

Another way of stating it is to say the subject prefers to maxi-

mize similarities among words.

(3) The Group Object Sorting Test. As mentioned, the group ver-

sion of the OST was devised by Clayton and Jackson (1961) to sub-

stitute for the individually administered version in which the

subject actually manipulated various objects into sorting-groups.

The paper-and-pencil version used in this study consists of a list

of fifty nanes of common objects and an answer sheet divided into


several large blank spaces. The subject is asked to write into

the various spaces the names of those objects he thinks belong

together (for any reason whatsoever) and to label the groups A, B,

C, etc. The instructions indicate the person is to sort all of

the object-names, but in case he cannot find any group in which

to include a few of the names he may place each of them into a

space by itself. The original scoring for the OST was simply to

count the number of spaces, or categories, used, with a higher

score indicating a finer, or less coarse, category system. How-

ever, Messick and Kogan (1963) found that two fairly independent

scores could be derived from an OST: the number of categories

which include two or more objects and the number of categories

containing only a single object. They found both these scores to

be reliable -nd to correlate differently with other cognitive and

personality measures. They labeled the first score Conceptual

Differentiation and the second Compartmentalization, or Restricted

Productive Ideation. It seemed appropriate in this study to call

them Conceptual Differentiation (CD) and Conceptual Compartmentali-

zation (CC). The correlation between these two scores in their

study was only .10, while Compartmentalization correlated .93 with

the previously used total score. Thus only the CD and CC scores

appear to be needed to capture the useful dimensions of this instru-

ment. High scores on the Conceptual Differentiation dimension mean

the subjects have used several two-item-or-more categories and are

thus stressing differences among objects. Therefore their equiva-

lence ran:os are narrower than those of subjects scoring low on

this dimension, A high score on Conceptual Compartmentalization


represents some kind of extreme emphasis on differences and thus

suggests a low equivalence range.

Internal-External Control. The I-E scale consists of twenty-nine pairs

of matched statements between which the subject is asked to choose. Only

twenty-two of the items are scored, the rest being fillers. For this

study the scale was scored so that a higher score represented more belief

in the person's Internal Control of his reinforcements; the lower scores

thus indicated belief in External Control.

Dogmatism. Rokeach's forty item Dogmatism scale, Form E, was used in this

study to measure the relative openness and closedness of the person's cog-

nitive system. The subject has six ways to respond to each item, ranging

from +3, "I agree very much," to -3, "I disagree very much." Because the

total scores in this study included negative numbers down to -80, they

were converted by adding +80 to each of them. High scores on this varia-

ble represented a more closed system, or more Dogmatism, while low scores

represented more openness of the cognitive system.

Specialization Level. This standard scale of the SVIB was included as

one of the predictor variables of profile height in an effort to gain

more understanding of its relationship to profile height and to the other

cognitive variables. The score reported is a standard score (with the

mean at 50), with the higher scores representing interest in specializing

and the lower scores, interest in being a generalist.

Moderator Instruments

At the conclusion of their rather thorough study of risk taking,

Kogan and Wallach stated that "consideration of potential moderator var-

iables is nothing less than essential in psychological research involving

the study of correlations" (1964, p. 188). Their reason for this strong


statement was that in most of their work "the assumption of linearity

that has typically been made in previous studies of the overall relation-

ship between two psychological variables simply does not hold" (ibid.).

The use of moderator variables was first suggested by Saunders (1956) and

was further promoted by Ghiselli (1960a, 1960b, 1963). Wallach (1962)

warned against the dangers of working with only one variable at a time or

even of trying to generalize one construct as broadly as Witkin and Fater-

son were attempting with their global-analytic dimension. He noted that

some were trying to say that leveling-sharpening was a variant of analyti-

cal-global, but contradictory studies had also been found to this proposi-

tion. In fact, the personality area has included many contradictory

studies, an effect that Wallach felt was partially due to uncontrolled

moderating effects. That is, the contradictions may have been due to the

differing positions of the samples upon various moderator variables.

As Wallach (1962) pointed out, a moderator variable represents a

dimension along which a sample may be divided into smaller, more homoge--

neous groups within which relationships between other variables may be

revealed more strongly. In fact, the moderator-variable groupings may

reveal strong correlations within subgroups where, when the sample as a

whole is considered, only weak correlations turn up by the usual methods.

Thus Kogan and Wallach (1967) reported that in their 1964 study they

would have found little consistency in a person's preferred level of risk

from one task to the next if it were not for their segregating their sub-

jects into four groups along the dimensions of test anxiety (fearfulness)

and social desirability (defensiveness). Those who were high in both

fearfulness and defensiveness were very consistent in their preferred

level of risk, whereas those low in both were more free to vary their


level of risk according to the different demands of each situation. The

latter subjects were found to be, in fact, most adaptive of the four

groups. Without the use of moderator variables this study would possibly

have been another correlational failure.

Conservatism-Riskiness. As was suggested in the first chapter, some amount

of risk seems to be involved in taking the Strong, and a convenient instru-

ment for assessing risk-taking is Kogan and Wallach's (1964) Choice Dilem-

mas Procedure. They have recently reported that, based on a wide variety

of studies, they have "much confidence" in the validity of this procedure

for assaying this dimension (Wallach & Kogan, 1965, p. 8). It was not

apparent that this dimension would have any direct effect upon the over-

all intensity of interests on the SVIB, but it seemed conceivable that

this variable could moderate the effects of other variables upon the SVIB.

That is, it seemed possible that a highly risky person with a low pro-

file could have his low profile as a result of different cognitive fea-

tures from those of a highly conservative person with a low profile.

The Choice Dilemmas Procedure consists of twelve different proble-

matic situations in which a person is faced with two alternative courses

of action, one of which is more attractive than the other but which is

also less likely to be attained than the other. The subject is asked to

indicate the minimum odds of success he would demand before recommending

that the more attractive or desirable alternative be chosen. The items

are scored such that a high score indicates the person would demand high

odds of success before recommending the attractive alternative and would

thus be Conservative, or non-risk-taking. Low scores characterize a

person willing to take chances, or to be Risky.

Test AnxietX. Kogan and Wallach (1964) found that a useful moderator


variable was that supplied by Alpert and Haber's (1960) Test Anxiety

Scale. The former have done a great deal of work with this variable, and

they are satisfied that the test-anxious person generally views his envi-

ronment in terms of its possible implications for his failure or success

(Wallach & Kogan, 1965). This person is "rendered especially sensitive

to whatever payoff characteristics could be construed as relevant to any

decision situation" (p. 209). Further, Burnstein (1963) found, in a

study of the relationships between test anxiety and occupational aspira-

tions, that fear of failure contributed to a willingness to settle for

less prestigious occupations. Thus, if fear of failure leads to a pat-

tern of interests characteristic of the less prestigious, non-profes-

sional occupations which are not on the SVIB, then it is also possible

that fear of failure may contribute directly to a low profile.

The scale consists of nineteen items, divided so that approximately

half of the items ask how much anxiety hinders test-taking and half of

the items ask how much anxiety benefits the test-taker. The scale was

scored so that high scores represented Test Anxiety, or fear of failure,

while low scores represented unconcern or confidence.

Social Desirability. The second moderator variable used by Kogan and

Wallach (1964) was that supplied by Crowne and Marlowe's (1960) Social

Desirability scale. This scale was designed to replace Edward's SD

scale, which Marlowe and Crowne felt was too dependent upon items con-

cerned with pathology. Christie and Lindauer (1963) in their annual

review agreed that the Crowne and Marlowe SD scale had successfully

replaced Edward's. At first the scale was interpreted as measuring the

person's need for approval, but Crowne and Marlowe (1964) began to

change their interpretation of their scale, and studies since then have


presented data that a defensiveness interpretation is more appropriate

than the need-for-approval concept. Thus, the scale seems to be getting

at a denying or repressing mode of behavior (Jacobson & Ford, 1966) or a

repressive ego-defensiveness (Breger, 1966). It seemed probable that

performance on the SVIB would vary considerably according to the per-

son's defensiveness, but how this would actually affect the other vari-

ables' relationships to the SVIB was difficult to predict.

The scale consists of thirty-three items to which the subject is

asked to respond True or False, that is, whether the items describe him

accurately or not. A high score on this scale means the person is trying

to make his actions appear Socially Desirable; he is defensive. A low

score represents a kind of openness.

Occuational Level. Strong (1943) speculated that two contradictory ways

might exist in which a person could get a low level profile: one is

through not being interested in any kind of professional occupation and

the other is through having such a wide variety of interests in upper

level careers that no area is strong enough to produce a peak on the SVIB.

The non-occupational scale OL (Occupational Level) of the SVIB measures

the amount of general interest the individual has in the careers of tho

upper socio-economic classes. It seemed likely that dividing subjects

along this variable would allow the other variables to show a relation-

ship with profile height which would otherwise perhaps be cancelled out.

Criterion Measures

As was mentioned, since a variety of studies had failed using the

primary pattern concept to define criterion groups, it seemed appropriate

to change to using criterion variables and to investigate several differ-

ent ways of quantifying the criterion of profile height. The method of


evaluating different measures of the criterion will be discussed later,

but it is based upon Hotelling's (1936) canonical correlation technique.

The criterion variables were as follows:

(1) Primary-Secondary Pattern Score. The traditional way of des-

cribing the interest intensity of different SVIB profiles is by

means of the classification system of primary and secondary occu-

pational interest families, as discussed above. It was proposed

to generate a variable by counting the number of primary and sec-

ondary occupational families in each profile (using the classifica-

tion rules of Korn and Parker, 1962), multiplying the number of

primary patterns by two, and summing the thusly weighted number of

primary patterns and the number of secondary patterns. It was

felt that this weighted total would reflect the relative impor-

tance of the two patterns in counseling. In order to accomplish

this, Korn and Parker's procedures had to be altered slightly

because of the addition of three new occupational scales and the -

shifting of others within the occupational families in the new

revision of the SVIB. In the pre-1966 Strong, four of the occu-

pational families contained only one occupation each, and the

traditional way of classifying primary and secondary families was

done by collapsing these four single-occupation families into

those other, larger families with which they correlated most high-

ly. The 1966 revision contains only two such single-occupation

families, VII, CPA Owner, and XI, President, Manufacturing Con-

cern. For this study, XI was combined with Group IX, Sales, and

VII was combined with Group X, Verbal-Linguistic, giving a total

number of nine occupational families to be classified. The


weighted sums of these classifications was called the Primary-

Secondary Pattern Score (PSPS).

(2) Letter Range Score. A way of including more of the variance

of the profiles in a criterion measure is to attend to the indi-

vidual occupational scales instead of the occupational families.

The scores on forty-eight different occupations are reported in

two different ways. A standard score is given for each scale

which represents how much the subject's way of responding to the

test matched the answers given by selected groups of occupation-

ally successful men. These standard-score scales are also divided

into ranges of scores headed by letters of the alphabet. Thus, an

A score on an occupational scale represents a standard score of 45

or higher, indicating a high level of concordance between the sub-

ject's preferences and those of men in that occupation. The B+

range includes standard scores of 40-44, B, 35-39, etc., down to

the C range which includes any standard score of 24 or less. It

was proposed to generate another criterion variable by arbitrarily

assigning five points for each A score on a profile, four for each

B1+, three for B's, two for each B-. and one for each C+. These

forty-eight different scores on each profile were summed, and the

total was called Letter Range Score (LRS).

(3) Total Standard Score. A third possible criterion variable,

and a way of including all the variance in the SVIB profile, was

to be obtained by simply adding together all of the individual

standard scores on each scale. It seemed possible that this total

variance would be necessary to understand the contribution of cog-

nitive variables to the strength of the profiles. This variable

was called Total Standard Score (TSS).


(4) Rejection Pattern Score. The final criterion variable, which

was included for investigatory reasons, was the total number of

reject patterns per profile. Apparently little has been done with

this as a variable, and no hypotheses were generated to predict

relations between it and the predictor variables. As noted

earlier, three different rules have been proposed for judging

reject patterns, and Korn and Parker's (1962) rules were the

strictest of the three. Because their rules were so strict as to

turn up reject patterns in only 22 percent of their profiles,

this rule was rejected in favor of Stephenson's (1961) rule. A

reject pattern was thus ascribed to an occupational family when

a majority of the occupations in a particular family were in the

C range (at a standard score of 24 or less). The total number of

reject patterns per profile was called the Rejection Pattern

Score (RPS).

Hypothesized Correlations

It may be helpful at this point, with the variables having been

defined quantatively, to restate the hypotheses in terms of the correla-

tions which were expected in the study. These restated hypotheses werps

(1) Field Independence correlates positively with Primary-Second-

ary Pattern Score (PSPS), Letter Range Score (LRS), and Total

Standard Score (TSS).

(2) Complexity of Verbal Apparatus correlates positively with


(3a) Category Width correlates negatively with PSPS, LRS, and TSS.

(3b) Verbal Equivalence Range correlates negatively with PSPS,

LRS, and TSS.


(3c) Conceptual Differentiation correlates positively with PSPS,

LRS, and TSS.

(3d) Conceptual Compartmentalization correlates positively with


(4) Internal Control correlates positively with PSPS, LRS, and


(5) Dogmatism is confounded by Occupational Level and is not

significantly correlated with PSPS, LRS, and TSS, except within

moderator subgroups defined by the Occupational Level variable.

Subjects and Procedure

The subjects used in this study were male undergraduate university

students, approximately half of whom were enrolled in one of two basic

psychology courses which required them to take part in experiments for

course credit. Because an insufficient number of these students volun-

teered to take the tests, additional students were recruited through an

offer of five dollars to the first 120 male undergraduates who would

appear to take the three hours of tests. The total number of subjects

came to 252, with 130 being class "volunteers" and 122 being paid volun-


The tests were administered to the subjects in the following order:

(1) Embedded Figures Test

(2) Clayton and Jackson's Object Sorting Test

(3) Rokeach's Dogmatism Scale

(4) Choice Dilemmas Procedure.

(5) Marlowe-Crowne Social Desirability Scale

(6) Synonymity Task

(7) Category Width Scale

(8) Rotter's I-E Scale

(9) Complexity-Simplicity Scale (Vannoy's Factor I)

(10) Alpert-Haber Test Anxiety Scale

(11) Strong Vocational Interest Blank For Men.

The Embedded Figures Test was administered first, because it was the only

timed test used. After the time on it had elapsed, the subjects were

instructed to turn in their copies and the answer sheets. They then

began work on the Object Sorting Test and the ones following it. The

tests were given in an order which, it was hoped, would provide variety

and thus maintain interest. Before the testing began all subjects were

informed that a session would be scheduled later at which they would be

informed of the test results and also receive their own personal voca-

tional interest profile.

The necessity of using paid volunteers to fill out the subject-

total unfortunately led to some variation in the test-taking procedure.

The psychology class volunteers took their tests on two different nights,

the first night being given to the first ten tests and the second night,

to the SVIB. The paid volunteers took a!l of the tests the same evening.

Because this difference in procedure could have caused the paid students

to take the SVIB with more "test fatigue" than the class volunteers had,

it was considered important to test any possible differences between

these two groups on the four criterion variables gained from the SVIB.

The data in Table 1 show that the paid subjects as a group had

significantly fewer primary-secondary patterns and fewer rejection pat-

terns than the non-paid subjects. Though with 250 degrees of freedom

these differences are statistically different, they do not appear to be

meaningfully large. Nevertheless, bGcause of these differences it was




Criterion Standard
Variables Groups Mean Deviation t

Primary-Secondary Paid 3.11 1.84
Pattern Score Non-Paid 3.64 1.87 2.23*

Letter Range Paid 65.89 12.54
Score Non-Paid 66.57 11.32 0.45

Total Standard Paid 320.05 63.62
Score Non-Paid 314.30 54.31 0.77

Rejection Paid 2.87 1.09
Pattern Score Non-Paid 3.23 .99 2.75**

*p< .05
**p< .01

considered appropriate to make some further checks, which will be dis-

cussed in the Analysis section which follows.

The question of the motivation of the paid students to take care

in answerbig the SVIB may possibly be answered by examining the relative

percentages of those students who showed up at the later-scheduled

interpretation sessions. Fifty-two percent of the 252 students received

their copies of the SVIB profiles, with 70 percent of the paid students

and only 36 percent of the non-paid students showing this interest.

These deviations from the expected percentages were significant in a

chi-square test beyond the .001 level, suggesting that the paid students

were much more interested in receiving the SVIB results than were the

non-paid students. Thus it is probable that the motivation of the paid

students while taking the SVIB was at least as high or higher than the


motivation of the non-paid students.

The type of subjects used in this study differed from the type

used in several similar studies. Most other studies have used freshmen,

who received the SVIB as part of an orientation battery, or counselees

in a counseling center. It seemed appropriate, then, to make some com-

parisons of the performance of the total subject pool in this study with

some normative data from other studies.

These comparisons are presented in Tables 2 and 3. The Minnesota

figures in the tables are from a study of 1000 freshmen conducted in the

early 1940's by Darley using the 1938 revision of the Strong (Darley &

Hagenah, 1955), while the Stanford figures were obtained, also using the

1938 revision, from 853 freshmen entering Stanford in 1959 (Korn & Park-

er, 1962). Table 2 shows that the occurrence of different interest pat-

terns within the Florida sample was similar to the percentages in the



Pattern Type Minnesota* Stanford** Florida

No Primary 19.3 41 26.6
Single Primary 41.0 37 40.9
Double Primary 30.3 18 22.6
Triple Primary 8.8 3 9.1
Quadruple Primary .6 -- .8
No Secondary 26.0 28 30.2
Single Secondary 42.9 44 42.5
Double Secondary 23.4 21 19.8
Triple Secondary 4.2 6 6.7
Quadruple Secondary .3 1 .8
No Primary or Secondary 2.2 7 7.1

*From Darley & Hagenah, 1955, pp. 86-87, 90.
*From Korn & Parker, 1962.


Minnesota and Stanford groups. Thus it seemed probable that the varia-

tions of profile height in the present sample are representative of the

general student population. And it would seem that the 1966 revision of

the SVIB produces patterns similar to the ones produced by the 1938


Table 3 indicates some differences in the occurrence of primary

and secondary patterns within the different occupational families among

these three samples. However, it is difficult to make rigorous compar-

isons here, since, as was already pointed out, the 1966 revision of the

Strong allows for the inclusion of two more occupational families than



Occupational Primary Patterns Secondary Patterns
Family Minn.* Stan.** Fla. Minn. Stan. Fla.

Biological Science 9.5 17 20.6 11.8 18 13.5
Physical Science 15.5 10 15.1 12.7 8 9.1
Technical Supervision .. .. 13.5 .. .. 13.1
Technical-Skilled 32.6 10 2.4 18.8 16 6.?
Social Service 11.8 12 11.1 13.3 14 13-9
Aesthetic-Cultural .. .. 16.7 .. .. 16.3
Business-Accounting 24.9 9 6.0 17.8 19 9.1
Sales 22.1 16 12.7 16.6 14 9.1
Verbal-Linguistic 12.8 10 18.7 12.8 15 14.7

*Minnesota data from Darley & Hagenah, 1955, p. 98.
*Stanford data from Korn & Parker, 1962.

did the 1938 revision, and further, some of the occupations have been

moved to different families in the new revision. With all this, the

present Florida sample seemed very similar to the Stanford and Minnesota

samples in the areas of Biological Science, Physical Science, and Social

Service occupations. Interestsin Technical and Business areas were

lower in the Florida sample while there seemed to be a slight increase

in interests in the Verbal-Linguistic occupations. All in all, the

Florida sample does not appear to be unrepresentative of students at


Research Design and Analysis

The scoring of the ten objective scales followed the usual proce-

dure outlined by the various authors of the tests, except where noted

differently above. The SVIB was scored by the National C'omputer Systems,

Inc., of Minneapolis, following their usual procedure. The scores sup-

plied by this company included those of the forty-five occupational

scales and four non-occupational scales (as well as ten experimental

occupational scales not used in this study). On each of these supplied

profile sheets the four SVIB criterion variables were then scored by

the investigator.

Once all the data were collected, the subject-scores were randomly

divided into two samples, the initial sample and the cross-validation

sample. In imitation of the designs of Ghiselli (1960b) and Noonan

(1967), the initial sample included 150 subjects, while the remainder,

102 subjects, composed the cross-validation sample. This unequal divi-

sion of subjects was due to the different purposes of the two samples.

The initial sample was used not only to test the initial hypotheses but

also, in a sense, to generate more exact hypotheses, which were then

cross-validated on the second sample. The larger number of subjects in

the initial sample gave the greater statistical power to the sample which

was used to discover hypotheses, leaving what was considered to be a


minimum but adequate number in the cross-validation sample. The experi-

mental design called for the division of each of these samples approxi-

mately into thirds, and the resulting subgroups in the initial sample of

about 50 subjects each and in the cross-validation sample of about 34

subjects each seemed to provide a reasonable compromise between the

needs of stability and the needs of power (Cohen, 1962).

Because of the finding that the paid and non-paid subjects differed

on two of the criterion variables, the distribution of these two kinds of

subjects between the two samples was checked. Of the 150 subjects in the

initial sample, 74, or 49 percent, were paid subjects; in the cross-vali-

dation sample, 48 out of 102 subjects were paid, or 47 percent. Neither

of these percentages is significantly different from the 48 percent of

paid students in the total sample. Thus it seemed reasonable that any

differences between the paid and non-paid subjects were equally dis-

tributed between the initial and the cross-validation samples. Differ-

ences between the two samples could not be attributed to the two ways -

the subjects were recruited and treated.

Initial Smple Analsis

Canonical Correlation. The first step in the analysis of the initial

sample was to carry out Hotelling's (1936) canonical correlation between

the first seven predictor variables (those for which there were direc-

tional hypotheses) and the four criterion variables. The actual pro-

cedure used was that of Cooley and Lohnes (1962), who described it as a

way of maximizing the correlation between linear functions of two sets

of variables. Using two sets of simultaneous equations involving the

predictor and criterion variables, this procedure determines a set of

weights for the two sets of variables which will maximize the correlation

between the derived canonical variates. A chi-square test is available

to determine if the linear combinations significantly relate the two

sets of variables. Since there are as many linear combinations possible

as there are variables in the smallest sot (whether criterion or predic-

tor set), it is possible to have more than one significant relationship

defined by the equations. Included in the computerized procedure of

Cooley and Lohnes (1962) is the chi-square testing of the significance

of each of these independent linear combinations. Further, the weights

used to maximize the correlations between the canonical variates can

also be used to help in judging which of the criterion meamres is the

more important measure in relation to the predictor variables.

Intercorrelation Matrix. The second step in the analysis of the initial

sample was to calculate product moment correlations between all seven-

teen of the vara-bles. From the resulting intercorrelation matrix the

initial hypotheses were tested. This matrix was also examined for sig-

nificant relationships not hypothesized for this study.

Moderator Variable Analysis. The third step in the analysis of the

initial sample was to test the moderating effects (Saunders, 1956) of

each of the predictor and moderator variables upon the relationships

between the remaining variables and the criterion variables. The test-

ing of the moderating effects of a particular variable was accomplished

by first ranking all the subjects upon the variable and then deciding

upon cutting scores which would divide the sample approximately into

thirds. These decisions for the cutting scores in the variables in the

initial sample wore made with the distributions of the scores in the

cross-validation sample in mind, so that the uso of the same cutting

scores in both samples would produce the best possible equal subgroups


in both samples. BPcause the distributions of scores within the two

samples varied considerably on some variables, some of the subgroups

deviated quite a bit from the hoped-for equal third divisions. In the

initial sample, where the goal was 50 subjects in each subgroup, the

number of subjects actually in each subgroup ranged from 33 to 63.

The corresponding range in the cross-validation analyses was 21 to 42,

with the ideal size being 34. (The cutting scores on each of the twelve

variables used in the moderator analyses of the samples, as well as the

consequent subgroup sizes, are given in Table 16 in Appendix A.) These

differences arose because it was judged to be more meaningful to use the

same cutting scores in the samples rather than to use equal subgroup

sizes. Since cutting scores hopefully categorize subjects meaningfully,

the above procedure should provide more transference to future research.

These approximately equal third divisions were made within the

scores of twelve of the criterion and moderator variables (Specializa-

tion Level was not investigated as a moderator variable), resulting in -

36 different (though not independent) subgroups (the subjects were sorted

into three groups twelve different ways). Within each of these 36

groups, pr-iduct moment correlations were run between twelve variables

and the four criterion variables, which produced, then, 1728 different

correlations within the initial sample. For example, 41 subjects made

up the High subgroup on the variable Field Independence. The scores of

these 41 subjects on the remaining twelve variables (Specialization

Level was included in these correlations) were each correlated with each

of the four SVIB criterion variables, resulting in 48 different corre-


As will be noted in the n~xt chapter, a way to group the results


in a fashion more meaningful than had to be used to calculate these corre-

lations is to sort the correlations by variable, rather than by sub-

group. For example, the correlations of Field Independence with each of

the four criterion variables were extracted from each of the computer

subgroup tables and placed in a table by themselves (with the appropri-

ate identification of their subgroup source).

After these arrangements were made, the 1728 different correla-

tions in this initial sample moderator analysis were examined for statis-

tical significance. Because any significant relationships found were

used as hypotheses to be re-examined in the cross-validation sample, and

because of Cohen's (1962) warning that typically this kind of research

does not use enough power in its statistical tests, the decision was

made to use the .10 level of significance in the t-test of the correla-

tions in this initial-sample moderator analysis. Changing the signifi-

cance level from .05 to .10 raises the power of the t-test to detect

correlations which are in reality different from zero but which may, in

any given sample, be low because of sample error. The significance

level was returned to .05 for the cross-validation moderator analysis.

Cross-Validation Sample Analysis

Using the hypotheses gained from the initial-sample analysis,

moderator analyses were run on the 102 subjects making up the cross-

validation sample. Though it was necessary for efficiency reasons to

imitate in entirety the analysis run on the initial sample, only the

correlations actually involved in the hypotheses are being reported,

since the purpose of the cross-validation sample was not to look for

new significant relationships but only to test those previously dis-

covered. However, an intercorrelation matrix involving all the


variables was produced, just as in the initial sample, and the canonical

correlation analysis was again run on this sample, in order to give an

indication of the stability of the findings in these areas.

Because the canonical correlation analysis involves matrix-inver-

sion calculations and because 3728 correlations were necessary in the

other analyses, the facilities of the University of Florida Computing

Center were used.



In this chapter the results of the canonical correlation analyses

of the initial and cross-validation samples will be presented first,

followed by the correlation matrices of these two samples. These corre-

lation matrices provided the test for the initial hypotheses. Following

this will be the summarized results of the moderator analyses, which

includes only those relationships found to be significant in the initial

sample and also supported by the cross-validation sample.

Canonical Correlation Analyses

As was mentioned in the last chapter, canonical correlation anal-

ysis using the Cooley and Lohnes (1962) procedure was performed between

seven predictor and four criterion variables of the initial and cross-

validation samples. With seven predictor and four criterion variables

there are four independent ways of relating these variables, and thus

the analysis produces four different canonical correlations. These

correlations of the initial sample are presented in Table 4.

As can be noted from the table, the maximum canonical correlation

of the initial sample was .33, but the chi-square analysis revealed this

was not a statistically significant relationship. That is, the seven

variables of Field Independence, Complexity of Verbal Apparatus, Cate-

gory Width, Verbal Equivalence Range, Conceptual Differentiation, Con-

ceptual Copartmentalization, and Internal Control, when taken as a




Number of Largest Latent Corresponding Chi Degrees of
Roots Removed Root Remaining Canonical R Square Freedom

0 .ill .33 29.04 28
1 .o48 .22 12.03 18
2 .023 .15 4.99 10
3 .oU .10 1.58 4

group, were not significantly related in the initial sample to the four

criterion variables, taken as a group.

Despite this lack of significance, it was still considered appro-

priate to examine the canonical coefficients which produced the largest

canonical correlation (cf. Stewart & Love, 1968). These are displayed

in Table 5. The coefficients suggest that the predictor variable most

related to the criterion variables was Complexity of Verbal Apparatus,





.567 Complexity of Verbal Apparatus .688 Letter Range Score
-.249 Field Independence .567 Rejection Pattern Score
.169 Internal Control -.410 Total Standard Score
-.144 Conceptual Compartmentalization -. 192 Primary-Secondary
.081 Verbal Equivalence Range Pattern Score
.072 Conceptual Differentiation
.022 Category Width

while the criterion variable which was most importantly related to the


predictor variables was Letter Range Score, That variable which reflects

the most popular criteria, primary and secondary patterns, was least

related to the predictor variables in this largest root of the initial

sample canonical analysis.

The cross-validation sample canonical analysis produced similar

results. Table 6 presents the four canonical correlations, and again

the largest correlation was not significant.



Number of Largest Latent Corresponding Chi Degrees of
Roots Removed Root Remaining Canonical R Square Freedom

0 .118 .34 29.05 28
1 .093 .31 16.96 18
2 .047 .22 7.54 10
3 .030 .17 2.95 4

Again it was decided to note the canonical coefficients of the

variables which had enabled their variates to be related with a correla-

tion of .34, the largest canonical correlation. Table 7 shows that in

the cross-validation sample there were two predictor variables, Field

Independence and Internal Control, of similar importance in relating

to the criterion variables as a group. Letter Range Score continued to

be the most important criterion variable in this first correlation of

the cross-validation canonical analysis, with Primary-Secondary Pattern

Score close behind it. (The coefficients for the remaining three corre-

lations in each sample are located in Tables 47-52 in Appendix E.)




Predictors Criteria

-.298 Field Independence .624 Letter Range Score
.272 Internal Control -.594 Primary-Secondary
.178 Complexity of Verbal Apparatus Pattern Score
-.043 Conceptual Compartmentalization .487 Rejection Pattern Score
-.041 Conceptual Differentiation .142 Total Standard Score
.015 Category Width
.012 Verbal Equivalence Range

Thus the canonical correlation analyses of the two samples agreed

that no predictor and criteria linear equations could be found in this

data which would significantly relate the two sets of variables. That

is, taken as a whole, the various measures of cognitive differentiation

were related to the various criterion measures significantly in neither

of the two samples. However, the coefficients of the variables as they

were used in the largest correlations of the two samples did reveal

some agrev ent between the two samples. That is, in both samples Com-

plexity of Verbal Apparatus, Field Independence, and Internal Control

were the three most important predictor variables in relating to the

criterion variables, while Category Width was the least important be-

tween the two samples. In both samples, the most important criterion

variable was Letter Range Score, with Total Standard Score being least

related to the predictors between the two samples. Primary-Secondary

Pattern Score was least important in the initial sample, but it moved

to being very close to the first in importance in the cross-validation

sample. Thus the functioning of the criterion variables in the two

samples was somewhat varied.


Intercorrelation Analyses

The seventeen-variable correlation matrix of the initial sample

(Table 8) revealed that when the variables were individually related to

one another (instead of being related through groups, as in the canoni-

cal analyses) only Complexity of Verbal Apparatus, among the eight pre-

dictor variables, was significantly correlated with any of the criterion

variables. This variable had a low positive correlation (p<.05) with

Primary-Secondary Pattern Score. Thus in the initial sample Hypothesis 2

found some weak support, while Hypotheses 1, 3a, 3b, 3c, and 4 failed in

this sample. Hypothesis 5 was not contradicted, in that Dogmatism was

not correlated with any of the criterion variables, but its evaluation

must be reserved for the moderator analysis which is to follow. In the

initial sample, then, only that hypothesis received support which sug-

gested that the more verbally complex a person is, the more capable he

is of differentiating his responses to the SVIB and of therefore differ-

entiating himself enough from men-in-general that he will have more pri-

mary and secondary patterns.

The cross-validation sample correlation wtatrix (Table 9) revealed

that only one predictor variable was significantly related to a criterion

variable in the cross-validation data. Field Independence had a low

negative correlation (p< .05) with Letter Range Score, an effect that

was in the direction opposite to that hypothesized. That is, in the

cross-validation sample the more field independent, or articulated, or

differentiated men tended to have lower SVIB profiles, rather than

higher profiles as predicted. There were no other significant relation-

ships between predictor and criterion variables in this sample.

Besides these two significant relationships between predictor and



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

2 .16*
3 .27** .16*
4 .04 .14 -.07
5 -.07 -.12 .10 -.12
6 -.03 -.07 -.01 -.13 .19*
7 -.07 .07 .05 .11 .10 -.01
8 -.05 -.38** -.07 .02 .15 .12 -.06
9 -.10 -.40** -.17* -.01 -.03 .06 -.27** .19*
10 .08 -.04 -.01 .01 -.04 .10 -.09 -.06 .04
11 .00 -.07 .00 .08 .02 .03 .02 .15 -.15 -.09
12 .13 .14 .26** -.11 .00 -. 04 .29** -.07 -.25* -.07 -.18
13 .08 .42** .05 -.01 -.03 .07 .11 -.13 -.15 -.07 -.13 .26**
14 .05 .16* .03 .03 -.05 -.08 .02 .00 -.05 .06 -.05 .18* .07
15 .01 .09 -.05 -.06 -.05 -.05 .00 .04 .00 .17* .00 .20* -.11 .71**
16 .07 -.05 -.06 -.11 -.11 -.02 .00 .05 .02 .13 .12 .09 -.22** .36** .75**
17 -.04 .23** .07 .10 -.03 -.09 .14 -.13 -.25** -.05 -.11 .09 .28** .23** -.18* -.40**

*P<,.05 **p< O1

Field Independence
Complexity of Verbal Apparatus
Category Width
Verbal Equivalence Range
Conceptual Differentiation
Conceptual Compartmentalization

7 Internal Control
8 Dogmatism
9 Conservatism-Riskiness
10 Test Anxiety
11 Social Desirability
12 Occupational Level

Specialization Level
Primary-Secondary Pattern Score
Letter Range Score
Total Standard Score
Rejection Pattern Score



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

2 .06
3 .10 .30**
4 -.16 .17 -.06
5 .01 .03 -.05 -.07
6 .10 .07 -.02 -.06 .23*
7 -.14 -.06 .17 .03 -.10 -.07
8 .05 -.38** -.19 -.03 .11 -.01 -.13
9 .16 -.25* .11 -.03 .00 -.03 -.01 .04
10 .11 -.20* .02 -.13 .08 .03 -.18 24* 10
11 .12 -.09 .05 .09 .05 .00 -.02 .24* .11 .06
12 -.11 .20* .18 -.14 -.03 .08 .23* -.18 -.27** -.24* -.04
13 -.15 .39** .19 .19 -.08 -.01 .24* -.25* .04 -.05 .05 .34**
14 -.11 .10 -.13 -.01 .05 .04 -.08 -.01 -.21* .05 -.08 .12 -.09
15 -.21* .07 -.12 .04 -.03 .07 -.01 -.04 -.26** .03 -.13 .17 -.14 .78**
16 -.09 .0. -.lo .12 -.12 -.05 .06 .03 -.07 -.03 -.03 .18 -.03 .43** .68**
17 -.07 .14 .09 -.02 .08 -.09 .12 -.19 -.06 .03 -.04 .03 .16 .23* -.12 -.32**

*p< .05 **p< .01

Field Independence
Complexity of Verbal Apparatus
Category Width
Verbal Equivalence Range
Conceptual Differentiation
Conceptual Compartmentalizat ion

Internal Control
Test Anxiety
Social Desirability
Occupational Level

13 Specialization Level
14 Primary-Secondary Pattern Score
15 Letter Range Score
16 Total Standard Score
17 Rejection Pattern Score



Initial Validation
Relationships Sample Sample

Field Independence and .16* .06
Complexity of Verbal Apparatus
Field Independence and .27** .10
Category Width
Complexity of Verbal Apparatus and .16* ,30**
Category Width
Complexity of Verbal Apparatus and -.38** -.38**
Complexity of Verbal Apparatus and -.40** -.25*
Complexity of Verbal Apparatus and .42** 39**
Specialization Level
Complexity of Verbal Apparatus and .16* .10
Primary-Secondary Pattern Score
Complexity of Verbal Apparatus and .23* .14
Rejection Pattern Score
Category Width and -.17* .11
Category Width and .26** .18
Occupational Level
Conceptual Differentiation and .19* .23*
Conceptual Compartmentalization
Internal Control and -.27** -.01
Internal Control and .29** .23*
Occupational Level
Dogmatism and .19* .04
Conservatism-Riskiness and -.25** -.27*
Occupaticnal Level
Conservatism-Riskiness and -.25** -.06
Rejection Pattern Score
Test Anxiety and .17* .03
Letter Range Score
Social Desirability and -.18* -,04
Occupational Level
Occupational Level and .26** .34*
Specialization Level
Occupational Level and .18* .12
Primary-Secoidary Pattern Score


TABLE 10 (continued)

Initial Validation
Relationships Sample Sample

Occupational Level and ,20* .17
Letter Range Score
Specialization Level and -.22** -.03
Total Standard Score
Specialization Level and .28** .16
Rejection Pattern Score
Primary-Secondary Pattern Score and .71** .78**
Letter Range Score
Primary-Secondary Pattern Score and .36"* 43*
Total Standard Score
Primary-Secondary Pattern Score and .23** .23*
Rejection Pattern Score
Letter Range Score and .75** .68**
Total Standard Score
Letter Range Score and -,18* -.12
Rejection Pattern Score
Total Standard Score and -.40"* -
Rejection Pattern Score

*p< .05
**p< .01

criterion variables, there were a number of other significant relation-

ships among the predictor, moderator, and criterion variables. In order

to display the stability of these relationships, Table 10 was prepared

from the two intercorrelation matrices. This table lists all those sig-

nificant relationships in the initial sample and shows the corresponding

correlations in the cross-validation sample. Of 29 significant rela-

tionships in the initial sample, 13 also found significance in the cross-

validation sample. These cross-validated findings can be divided into

four general groups of relationships: those among the cognitive and mod-

erator variables, those among the criterion variables, those among the

SVIB non-occupational scales and criterion variables, and those among


the cognitive and moderator variables and non-occupational scales.

(None of the relationships among the cognitive variables and criterion

variables, the hypothesized relationships, was significant in both sam-

ples, though the relationship between Complexity of Verbal Apparatus and

Rejection Pattern Score did receive some mild support in the cross-vali-

dation sample with a correlation of .14.)

First, among the cognitive variables and those variables added to

the study for moderator analysis, it appears that Complexity of Verbal

Apparatus was the variable most related to the others in this category.

In both samples it was significantly related to Category Width (positive

correlation), to Dogmatism (negative correlation), and to Conservatism-

Riskiness (negative correlation).

The relationship of Complexity of Verbal Apparatus to Category

Width appears to be opposite to what one would predict using a theory of

cognitive differentiation. What was found was that the more complex a

person's verbal apparatus, the wider quantative categories he tended to

use. However it would seem to be more reasonable that a more complex

verbal apparatus would call for narrower categories; broad categories

would seem to make complexity more difficult. The findings of this

study, however, are similar to what Vannoy (1965) found in his factor-

analytic study of a wide variety of measures of complexity. Though

Vannoy's analysis placed Complexity of Verbal Apparatus and Category

Width in two different factors, the factors were not oblique. His.

intercorrelation matrix also reported a correlation between Category

Width and the scale formed from Barron's Independence of Judgment scale

(Complexity) of .20. Similarly a correlation of -.20 was found between

Intolerance of Ambiguity and Category Width. The Intolerance of


Ambiguity scale was, of course, combined with the Complexity scale in

this present study to form a measure of Vannoy's first factor, Complex-

ity of Verbal Apparatus, by simply reversing the scoring of the Intoler-

ance of Ambiguity items and adding them to the Complexity items. The

approximate correlation of .20 in Vannoy's study between Complexity of

Verbal Apparatus and Category Width is similar to the correlations of

.16 and .30 found in this present study.

Coplexity of Verbal Apparatus was related also to Dogmatism, such

that the more complex persons were less dogmatic than verbally simpler

persons. Another way to state it is that persons with a simpler verbal

apparatus tended to be more dogmatic, or more closed to new beliefs and

information, which seems to be a reasonable relationship. Finally, it

was found in the two samples that those people who preferred complexity

also tended to recommend more risky behavior than did those who preferred

more simplicity; those with a simpler verbal apparatus tended to be


It should also be noted in this part that the two scores drawn

from the Object Sorting Test, Conceptual Differentiation and Conceptual

Compartmentalization, had a significant correlation of approximately .20

in the two samples. This compares with a correlation of .10 which

Messick and Kogan (1963) found in their study originally defining these

two variables.

The second general group of findings in the correlation matrices

of the two samples was that almost all of the criterion variables were

significantly related to each other. Positive correlations existed

between Primary-Secondary Pattern Score and the other three, but only

between Letter Range Score and two others and between Rejection Pattern


Score and two others. That is, of all possible comparisons of the cri-

terion variables, only Letter Range Score and Rejection Pattern Score

failed to be significantly related in both of the two samples. However,

even this relationship, with its correlation of -.12, can be said to

have received some mild support in the cross-validation sample.

The two criterion-variable relationships with the highest correla-

tions were those between Primary-Secondary Pattern Score and Letter

Range Score and those between Letter Range Score and Total Standard

Score. Thus if the variables are ranked as to the amount of variation

of the profiles they include, such that Primary-Secondary Pattern Score

includes least, Letter Range Score is next, and Total Standard Score

includes most, then the highest correlations are between adjacent vari-

ables in this ranking. The fact that, when the two extremes of the

ranking are correlated, the common variance between criterion variables

drops from around 56 percent to around 16 percent indicates that the

first three criterion variables are not simply different ways of quan-

tifying the same effect.

The Rejection Pattern Score variable was related to only two of

the other criterion variables, Total Standard Score and Primary-Secone'ary

Pattern Score. The relationship between Total Standard Score and Rejec-

tion Pattern Score was negative, as would seem appropriate from the make-

up of these two variables. Rejection Pattern Score and Primary-Secondary

Pattern Score were related with a low positive correlation, which seems

similar to the relationship between primary and rejection patterns in

Darley and Hagenah's (1955) data, as discussed in Chapter I. Primary

patterns and rejection patterns tend to go together, so that the more

differentiated a Drofile is in a positive direction the more differen-


tiated it tends to be in a negative direction. The fact that Rejection

Pattern Score correlates positively with Primary-Secondary Pattern

Score, negatively with Total Standard Score, and not at all with Letter

Range Score again seems to indicate that the latter variables are not

just different ways of quantifying the same thing.

The third general area of findings in Table 10 were the relation-

ships between the SVIB non-occupational scales and the criterion vari-

ables. Several of these relationships were significant in the initial

sample, with almost all of them also being close to significance in the

cross-validation sample. These included Occupational Level and Primary-

Secondary Pattern Score; Occupational Level and Letter Range Score; and

Specialization Level and Rejection Pattern Score. These indicate some

tendency for interest in higher socio-economic occupations to be asso-

ciated with a more differentiated SVIB profile, and a tendency for inter-

est in specialized occupations to be associated with the rejection of

more occupations. Furthermore, the two SVIB non-occupational scales

used in this study were significantly related in both samples. These

correlations are slightly higher than the correlation of .23 between

Occupational Level and Specialization Level reported in the current SVIB

manual (Strong, 1966). Thus an interest in higher level occupations tends

to be associated with interest in specialized occupations.

The final significant relationships in Table 10 were between the

non-occupational scales and various of the cognitive and moderator vari-

ables. Occupational Level was related to three of these variables, with

Specialization Level correlating significantly with just one. The high-

est correlations in Table 10 are between Specialization Level and Com-

plexity of Verbal Apparatus, indicating a strong tendency for preference


for complexity to be associated with interest in an occupation which

requires a great deal of specialized knowledge.

Occupational Level was associated with Internal Control, with

Conservatism-Riskiness, and to some extent with Category Width. The

person who sees himself in control of his own reinforcements also tends

to have interests compatible with the interests of men in higher socio-

economic occupations, as does the person who is more willing to take

risks. The view that there is a lot of luck involved in one's life and

also a hesitancy in taking risks seem to be related to being interested

in lower socio-economic occupations. Also existent was some tendency

for those who use broader quantative categories to be interested in the

higher level occupations.

In all of these correlations, the three variables which stood out

were Complexity of Verbal Apparatus, Occupational Level, and Specializa-

tion Level. It was noted that these variables were also rather strongly

interrelated, and there seemed to be some pattern to their relations to-

the other variables. In order to detect any possible patterns more

easily, T.ble 11 was prepared to display the trends in these relation-

ships as i,,ell as those relationships significantly cross-validated.

Table 11 readily shows the tendencies of Complexity of Verbal

Apparatus to relate to both Primary-Secondary Pattern Score and Rejec-

tion Pattern Score, of Occupational Level to relate to Primary-Secondary

Pattern Score and Letter Range Score, and of Specialization Level to

relate to Letter Range Score and Rejection Pattern Score. In addition,

both Complexity of Verbal Apparatus and Occupational Level were related

to Category Width and to Conservatism-Riskiness, both Complexity of

Verbal Apparatus and Spocialization Lovel were related to Dogmatism,




Occupational Specialization Of Verbal
Level Level Apparatus


PSPS .18* .12 too .16* .10
LRS .20* .17 -.11 -.14
RPS too so#28** .16.3* .1

OL ... ... .26** .34** .14 .20*
SL .26** .34** 00 4.*. .2**
CVA .14 .20* .42** ,39** .o

FI ... .. .16* .06
Cw .26* 1.16* 3o**
VER -.11 14 .14 .17
IC .29** .23* 0. .2,
Dog to -.13 -.25* -.38** -.38**
CR -.25** -.27** see g* -.40** -.25*
SD ... ... .. -.07 -.09

*p< .05
**p< .01

and both Occupational Level and Specialization Level tended to be related

to Internal Control. In these three patterns, the direction of the

relationships did not contradict the positive correlations among Occupa-

tional Level, Specialization Level, and Complexity of Verbal Apparatus,

although there was a tendency for Occupational Level and Complexity of

Verbal Apparatus to relate to Verbal Equivalence Range in such a con-

tradictory fashion. The consistency of the directions of these inter-

correlations appears to argue against attributing the significance of

the correlations to chance factors.


Another interesting fact about Table 11 is that it contains almost

all of the trends between variables in Tables 8 and 9 (except for those

among the criterion variables). Only four sets of consistent trend cor-

relations are excluded. These are the correlations of .19 (p< .05) and

.23 (p< .05) between Conceptual Differentiation and Conceptual Compart-

mentalization (already pointed out above), the correlations of .27

(p< .01) and .10 between Category Width and Field Independence, the

correlations of .15 and .24 (p< .05) between Dogmatism and Social Desir-

ability, and the correlations of -.13 and -.19 between Dogmatism and

Rejection Pattern Score (initial and cross-validation sample correla-

tions, respectively).

The correlations here between two cognitive variables--the Cate-

gory Width-Field Independence correlations--were in the direction con-

sistent with the correlations between these two and Complexity of Verbal

Apparatus. Thus Complexity of Verbal Apparatus, Field Independence, and

Category Width tended to be positively intercorrelated with each other.

Occupational Level almost joins this circle, for it shared in tendencies

toward positive correlations with Complexity of Verbal Apparatus and

Category Width. The correlations between Dogmatism and Rejection Pat-

tern Score are also consistent with the intercorrelations among Complex-

ity of Verbal Apparatus, Specialization Level, Dogmatism, and Rejection

Pattern Score. Thus Complexity of Verbal Apparatus, Specialization

Level, and Dogmatism were all consistent in predicting Rejection Pattern

Score. If Specialization Level can be thought of as a measure of some

kind of cognitive differentiation, then the correlations of Complexity

of Verbal Apparatus, Dogmatism, and Specialization Level with Rejection

Pattern Score all suggest that the more cognitively differentiated the


subject was, the more rejections of occupational families he made.

Other patterns in Table 11 include the intercorrelations of Com-

plexity of Verbal Apparatus, Specialization Level, and Occupational

Level, as already mentioned. The more the subjects preferred verbal

complexity, the more interested they were in specialized occupations

and in higher status occupations. Also, Internal Control tended to be

related to the two SVIB non-occupational scales, such that the more the

individual saw the reinforcements in his life as being under his control,

the more he tended to be interested in specialized and high status occu-

pations. The final group of intecorrelated variables in Table 11

involved Complexity of Verbal Apparatus, Occupational Level, and Con-

servatism-Riskiness. Here, the more non-risky conservative the individ-

ual, the lower were his status interests and the simpler were his prefer-

ences in verbal organization.

In general, then, the intercorrelation matrices revealed that three

out of the thirteen cognitive and moderator variables accounted for most

of the important correlations among the variables. The cognitive varia-

bles were poorly related to the criterion variables and to each other,

but there were some important correlations between the cognitive variables

and the two SVIB non-occupational variables included in this study. Al-

most all of these intercorrelations were consistent in the directions of

their intercorrelations with each other. However, from the viewpoint of

cognitive theory the direction of the correlations of Category Width

with both Complexity of Verbal Apparatus and Field Independence was op-

posite to what one might predict, though Field Independence and Complex-

ity of Verbal Apparatus tended to correlate in the expected positive

direction. Finally, the variable Social Desirability tended to correlate


only with Dogmatism, suggesting that a set to respond in a socially

desirable way had very little influence in producing these intercorre-


Moderator AnalYses

The first step in the moderator analyses of the two samples was to

rank the subjects on each of the twelve predictor and moderator varia-

bles (Specialization Level was not used as a moderator variable) and

then to divide the subjects, on each of these variables, into three

approximately equally numbered groups labeled High, Medium, and Low.

Within each of these resulting 36 groupings, correlations were run be-

tween the remaining twelve predictor and moderator variables (Speciali-

zation Level was included in these correlations) and the four criterion

variables. Thus even though some variables were primarily of interest

for their ability to predict, their ability to moderate other predictor

variables was also checked. And similarly those variables included as

moderator variables were also checked for their ability to predict. As

mentioned, this resulted in 1728 separate correlations in the initial

sample to he evaluated.

In jider to begin the evaluations of the moderator-analysis results

of the initial sample, the correlations were rearranged so that all of

the correlations between a particular variable and the four criterion

variables were placed in one table. Within a particular table, these

correlations were labeled according to the particular moderator subgroup

from which they were obtained. Thus, for example, all of the 132 corre-

lations between Field Independence and the four criterion variables

were placed in a table, and it was possible to examine the table and

note which moderator subgrouping resulted in significant correlations


between Field Independence and one of the criterion variables and which

did not. (Because of the length of these thirteen tables, they are

presented in Appendix B, Tables 17-29.)

After the correlations in the initial-sample moderator analysis

were examined for significance (using a significance level of .10, as

explained in Chapter II), all of the relationships attaining statistical

significance were rearranged in additional tables according to criterion

variables (these tables are in Appendix D, Tables 43-46). It was noted

from these tables that out of 432 correlations between the variables

used as predictors and Primary-Secondary Pattern Score, 44 were signifi-

cant at at least the .10 level. Of the same number, Letter Range Score

was correlated significantly with 56; Total Standard Score, with 58; and

Rejection Pattern Score, with 81.

In order to make the comparisons called for by the experimental

design, a moderator analysis was obtained on the cross-validation sample,

and the results were placed in tables similar to those of the initial-

sample analysis (the cross-validation moderator results are in Appendix

C, Tables 30-42). Only those correlations corresponding to the signif-

icant relationships in the initial sample were noted and placed in tho

four tables of significant relations arranged according to criterion

variables (Tables 43-46, Appendix D). These latter four tables made it

possible to see which of the significant initial-sample relationships

were cross validated in the cross-validation sample.

These four tables were quite lengthy, and a yet shorter table was

distilled from them to display the important results of the moderator

analyses. In order to evaluate also the trends in the data (those rela-

tionships not quits reaching cross-validation significance), the data


selected for display in this new table, Table 12, were chosen by three

rules. First, this table includes all those relationships which were

significant in the initial sample at the .05 level and which were sup-

ported by a correlation of at least .20 in the cross-validation sample.

Second, those additional relationships were selected which were signifi-

cant at the .10 level in the initial sample and which were supported by

a correlation of at least .25 in the cross-validation sample. This

second requirement was instituted in order to make use of the initial-

sample significance level of .10 in a way which would not be excessively

"liberal" but which would allow one to notice trends. Third, all of the

moderating conditions thus utilized by the first two rules were examined

for yet weaker trends between the same predictors and one or more of the

other criterion variables. These latter, weak trends were placed in

parentheses in Table 12. This last rule allows one to see if a particu-

lar predictor-criterion relationship is unique or if there was at least

some tendency for the same predictor to be related to one of the other

three criterion variables within the same moderating condition. Table

12 thus became the effective starting place for reporting the important

moderator analyses results,

The first characteristic of Table 12 which requires some comment

is the relative paucity of statistically significant cross-validated

correlations: there are only thirteen such correlations. Thus, of the

1728 correlations in the initial-sample moderator analysis which were

examined for significance at at least the .10 level, and of the 239

correlations in the initial sample which did achieve this significance

level, only thirteen were supported with a significance level of .05

in the cross-validation sample. (It should also be pointed out that



Variables Cross-
Used As Moderating Initial Validation Criterion
Predictors Conditions Sample**** Sample***** Variables


Complexity of
Verbal Apparatus

High Complexity of
Verbal Apparatus

Low Complexity of
Verbal Apparatus

Low Field

Medium Verbal
Equivalence Range

High Conceptual

High Conceptual

High Internal

Medium Dogmatism .41**P


-. 37**





- ,3*













Low Social

High Test


Equivalence Range


C ompartmentalization





High Test


High Verbal
Equivalence Range

High Conceptual

High Test Anxiety

Low Test Anxiety

High Internal Control

High Test Anxiety

High Social

Low Social Desirability

(-. 37**






-. 30**









- 31*








-.21 PSPS

(. 37*






TABLE 12 (continued)

Variables Cross-
Usod As Moderating Initial Validation Criterion
Predictors Conditions Sample Sample Variables

Test Anxiety





High Field

Medium Categoy

Low Category Width

Medium Conceptual

High Conceptual

Medium Dognatism

Medium Conservatism-

Medium Test Anxiety .25*

e 31"*


















.26 LRS

Low Social 5l*** .37 LRS
Desirability 37** .27 TSS

Specialization Low Field Independence .31** .43** RPS

Medium Complexity of .29** .41** RPS
Verbal Apparatus

Low Complexity of 33* -. 36** LRS
Verbal Apparatus

High Verbal .41*** .3l** RPS
Equivalence Range

High Conceptual -.27* -.29 TSS
Differentiation (.43*** .18 RPS)

Medium Conceptual -.27* -.29 LRS
Compartmentalization (-.42*** -.11 TSS)
33"* .20 RPS

High Internal .41*** .23 RPS

Low Internal -.28* -,28 TSS

High Test Anxiety -.36** -.31 LRS

Medium Social .38*** .20 RPS

High Occupational Level .27** .24 PPS

TABLE 12 (continued)

Variables Cross-
Used As Modeatirg Initial Validation Criterion
Predictors Conditions Sample Sample Variables

Low Occupational Level -.39** -.35** LRS
(-,38"* -.16 TSS)

*p< .10
**p< .05
***p< .01
****The N in each moderating Condition in the initial sample was approximately 50. The exact N for
each condition may be found in Table 16, Appendix A.
*****The N in each moderating condition in the cross-validation sample was approximately 34. The
exact N for each condition may be found in Table 16, Appendix A.


Primary-Secondary Pattern Score
Letter Range Score
Total Standard Score
Rejection Pattern Score


in Tables 43-45, Appendix D, there were three significant reversals in

moving from the initial sample to the cross-validation sample: Category

Width and Primary-Secondary Pattern Score, moderated by Medium Verbal

Equivalence Range; Conservatism-Riskiness and Letter Range Score, moder-

ated by Medium Dogmatism; and Test Anxiety and Total Standard Score,

moderated by High Conservatism-Riskiness.) These thirteen significant

cross-validations involved these variables used as predictors: Field

Independence, Complexity of Verbal Apparatus, Conservatism-Riskiness,

Occupational Level, and Specialization Level. Two of these were cog-

nitive-style variables (Field Independence and Complexity of Verbal

Apparatus); one was of the motivational variable type included for mod-

erator analysis but also checked for its ability to predict directly

(Conservatism-Riskiness); and the last two were non-occupational SVIB

scales, one of which was included for moderator analysis, but both of

which were also checked for direct predictability (Occupational Level

and Specialization Level). The last two together produced significant

relationships in more moderator subgroups than the variables in any

other classification. In addition to ths predictors related signifi-

cantly to the criterion variables, four other cognitive-style variables,

Category Width, Conceptual Differentiation, Internal Control, and Dog-

matism, tended to be related to the criteria also.

Looking at Table 12 from the viewpoint of the criterion variables,

one can note that the criteria correlated significantly in the largest

number of moderating conditions were Rejection Pattern Score and Letter

Range Score, with seven and four significant correlations respectively.

PrLmary-Secondary Pattern Score and Total Standard Score were each cor-

related significantly in one condition. In addition, the "trend"


correlations (but not the correlations in parentheses) break down into

fourteen conditions for Letter Range Score, twelve for Rejection Pattern

Score, six for Primary-Secondary Pattern Score, and five for Total Stand-

ard Score. Thus the two best criterion variables, from the viewpoint of

which correlated most frequently with the predictors in the moderator

subgroups, were Letter Range Score and Rejection Pattern Score. This

conclusion is similar to the one noted in the examination of the canon-

ical coefficients of the first roots of the two samples. In both s&'n-

ples the coefficient of Letter Range Score was larger than those of

Primary-Secondary Pattern Score and Total Standard Score. One can also

note in Tables 5 and 7 that Rejection Pattern Score appears to have a

combined importance in the two samples greater than that of the latter

two variables also. In this study, then, the conclusion seems inescap-

able that the best criterion of profile height was the Letter Range Score.

Yet another point should be noted in respect to the criteria. It

will be remembered from the discussion of Table 10 that Letter Range

Score and Rejection Pattern Score had the smallest mutual correlation of

all the criterion variable pairs. All of the other possible pairings of

the criterion variables were significantly correlated in both the sam-

ples. Thus it would appear that the two criteria most strongly related

to the predictors were also the two most independent among the criteria.

Even though Letter Range Score correlated around .70 with both Primary-

Secondary Pattern Score and Total Standard Score, the latter two were

nowhere nearly as effective in "capturing" the variance of the predictor

variables in the moderator groups as was the former. Thus the bulk of

the correlations in Table 12 do not appear to be simply repetitions of

correlations between the predictors and some common factor among the

criterion variables.


It is not clear why Primary-Secondary Pattern Score was a poorer

criterion measure than Letter Range Score, except possibly for the fact

that Letter Range Score was defined in a way which included more of the

variability of the SVIB profiles than did Primary-Secondary Pattern

Score. Why Total Standard Score, with its even greater variability than

Letter Range Score, was not as effective as Letter Range Score in relat-

ing to the predictor variables may be due to the possibility that Total

Standard Score contains contradictory variances. An examination of a

standard SVIB profile sheet reveals that Primary-Secondary Pattern Score

was defined in such a way that its entire variance represents changes

in positive differentiation on the SVIB. That is, only those occupa-

tional scores which are greater than the average score on each scale of

the men-in-general group go into the production of the Primary-Secondary
Pattern Score. From the way it was defined, most of the variance in

Letter Range Score is also attributable to scores greater than the aver-

ages of men-in-general, so that Letter Range Score measures mostly how -

the subjects differentiated themselves in a positive way from men-in-

general. Now, as Table 10 shows, Total Standard Score correlated pos-

itively wi-h Letter Range Score and Primary-Secondary Pattern Score in

the two samples, so that all three, then, were measuring at least par-

tially a kind of positive differentiation of the SVIB profile. However,

Total Standard Score, from the way it was defined, includes variance not

only from occupational scores higher than the averages of men-in-general

but also from occupational scores lower than the averages of these men.

'his men-in-general group was the second group so named, as dis-
cussed in footnote 4 in Chapter I. The average score of this group on
each scale is represented by the center of the shaded areas on a
standard SVIB profile sheet, on which the scores are reported.


The purest measure of scores lower than the averages of men-in-

general, or negative differentiation, was Rejection Pattern Score. From

the way it was defined, most of the SVIB occupational-scale scores which

contributed to it were lower than the averages of the scores which men-

in-general made on the scales. As was already pointed out, Rejection

Pattern Score correlated positively with Primary-Secondary Pattern Score

in the two samples with a correlation of about .23. Thus, extreme posi-

tive differentiation on the SVIB tended to be correlated with negative

differentiation, so that very high scores on some SVIB occupational

scales tended to be associated with the presence of very low scores on

other occupational scales. And therefore there also seems to be a gen-

eral factor of differentiation, as well as positive and negative differ-

entiation factors. This general factor was, however, apparently par-

tially cancelled out in the Letter Range Score and especially in the

Total Standard Score, for these two correlated approximately -.16 and

-.36, respectively, with Rejection Pattern Score. The correlation of

Letter Range Score with Rejection Pattern Score was not significantly

cross-validated, but the significant and larger negative correlation be-

tween Total Standard Score and Rejection Pattern Score seems to indicate

that the positive differentiation factor in Total Standard Score was

seriously confounded by the negative differentiation factor as repre-

sented by Rejection Pattern Score. That is, a very low score on a

particular occupational scale, while possibly increasing the Rejection

Pattern score of the individual, would also have lowered the Total

Standard Score of that individual. Since the general differentiation

factor indicates that there was a tendency for very high scores to be

associated with other very low scores, then in Total Standard Score the


result was for the extreme positive and negative differentiation factors

to cancel each other out. And while Letter Range Score also tended to be

negatively correlated with Rejection Pattern Score, the effect was not as

large and evidently not as damaging to Letter Range Score's ability to

relate to the predictor variables.

This probability of different common variances among the criterion

measures raised the possibility of simplifying the data in Table 12 by

arranging the relationships according to these possible common variances.

Table 13 is thus a rearrangement of Table 12, containing exactly the

same data. It was constructed according to these following rules. Since

Letter Range Score (LRS) and Rejection Pattern Score (RPS) appeared to be

the most effective criterion variables, any relationships which included

them were considered to be of a primary nature. An attempt was made to

treat the relationships involving Primary-Secondary Pattern Score (PSPS)

and Total Standard Score (TSS) as secondary or supportive. This was

done by examining Table 12 from the viewpoint of the different moderator

conditions and noting whether any relationship within each particular

condition which involved PSPS or TSS and a particular predictor varials

were also accompanied by a relationship between either LRS or RPS and

the same particular predictor variable. The restriction on this proce-

dure was that PSPS was not made subordinate to RPS, only to LRS, since

it is assumed that PSPS and RPS are respectively loaded on positive and

negative differentiation. TSS, however, could be placed under either

LRS or RPS since it seems to have variances in common with both positive

and negative differentiation. One final exception occurred when a sub-

group contained relationships involving both LRS and RPS when their

correlations with a common predictor were in the same direction. Here



Initial Validation
Relationships Sample Sample Moderating Conditions

Complexity of Verbal Apparatus--PSPS


Field Independence--LRS

Complexity of Verbal
Complexity of Verbal
(c'&, --P~S)
Complexity of Verbal

Category Width--LRS




(. 35***





( .27*

Conceptual Differentiation--LRS






High Conceptual Compartmentalization
(High CC)

Low Social Desirability

Low Complexity of Verbal Apparatum
(Low CVA)

High Conceptual Differentiation
(High CD)
High Internal Control
(High IC)
(High IC)
Low Social Desirability
(Low SD)
(Low SD)

High Test Anxiety
(ifigar TA)
(High TA)

High Test Anxiety
(High TA)
(High TA)

Internal Control--LRS
Occupational Level-LRS
Occupational Level-LRS
Occupational Level--LRS
Occupational Level--LRS
Occupational Level--LRS
Occupational. Level--LRS
Occupational Level--LRS
Occupational Level--LRS
Occupational Level-LRS

Specialization Level--LRS
Specialization Level-LRS
Specialization Level--LRS
Specialization Level-LRS

Specialization Level--TSS

Field Independence-RPS

Complexity of Verbal Apparatus--RPS
Complexity of Verbal Apparatus--RPS
Complexity of Verbal Apparatus--R1s


(. 31**


-. 36**
-. 35**





High Test Anxiety

High Field Independence
(High FI)
Medium Category Width
(Medium CW)
Low Category Width
Medium Conceptual Differentiation
(Medium CD)
High Conceptual Compartmentalization
(High CC)
(High CC)
Medium Dogmatism
Medium Conservatism-Riskiness
Medium Test Anxiety
Low Social Desirability
(Low SD)

Low Complexity of Verbal Apparatus
Medium Conceptual Compartmentalization
(Medium CC)
(Medium CC)
High Test Anxiety
Low Occupational Level
(Low OL)

Low Internal Control

High Complexity of Verbal Apparatus

Low Field Independence
(Low FI)
Medium Verbal Equivalence Range
Medium Dogmatism

TABLE 13 (continued)

Initial Validation
Relationships Sample Sample Moderating Conditions

Internal Control--RPS .25* .29 High Verbal Equivalence Range
Internal Control--RPS *31** .21 High Conceptual Compartmentalization
Internal Control--RPS .37*** .23 Low Test Anxiety
Dogmatism--RPS -.23* -.33 High Internal Control
Dogmatism--RPS -.27* -.27 High Test Anxiety

Conservatism-Riskiness-RPS -. 30** -. 31** High Social Desirability

Specialization Level--RPS .31** 943** Low Field Independence
Specialization Level--RPS .29** .41** Medium Complexity of Verbal Apparatus
Specialization Level.--RPS .41*** .31** High Verbal Equivalence Range
Specialization Level--RPS .43*** .18 High Conceptual Differentiation
(SL--TSS) (-.27* -.29) (High CD)
Specialization Level--RPS .41*** .23 High Internal Control
Specialization Level--RPS .38*** .20 Medium Social Desirability
Specialization Level--RPS .27** .24 High Occupational Level

*p< .10
**P< .05
***p< .01


Primary-Secondary Pattern Score
Letter Range Score
Total Standard Score
Rejection Pattern Score


it would seem that the third, or general differentiation, factor was

at work, and the RPS relationship was placed under the LRS relationship

to display this possibility. In those subgroups where these accom-

paniments existed, the PSPS and TSS (and some RPS) relationships were

placed in parentheses under the respective primary relationships involving

LRS or RPS, even where their correlations were larger than those of LRS

or RPS in the same moderator subgroup.

Table 13 shows the interesting result of this procedure. All but

one of the TSS and two of the PSPS relationships were placed in sup-

porting roles. Eight of the TSS were placed under LRS relationships,

while two of them were placed under RPS relationships. In respect to

these latter relationships, the opposite correlation signs are existing,

presumably because of the negative correlation between RPS and TSS due

to their being scored in opposite ways, as mentioned. It appears, then,

that TSS served little function in this study independent of LRS or RPS,

while PSPS mostly echoed the relationships of LRS.

The primary question addressed to the moderator analyses of the

two samples was, of course, whether these analyses would reveal relation-

ships, which would otherwise not have been apparent, between the cognitive

differentiation variables and the criterion variables. The groupings of

the moderated relationships in Table 13 form the basis for this examina-


Positive Differentiation

Neither of the two relationships between PSPS and variables used

as predictors, which were not mirrored by relationships involving LRS,

was cross-validated significantly. The relationship with Complexity of

Verbal Apparatus in the condition of High Conceptual Compartmentalization