DUKE
UNIVERSITY
LIBRARY
DUKE UNIVERSITY LIBRARY
MANUSCRIPT
THESES
This volume nay be oonsalted freely, but the literary rights of the
author must be respeoted.
No passage may be oopied or olosely paraphrased
without the previous written oonsent of the author.
If the reader obtains
assistance from this volume he must give oredit in his own work.
This thesis by
has been used by the
Gary Bruce Cox
following persons, whose signatures attest their aooeptanoe of the above
remsbriot ions.
t A library borrowing this thesis for use by one
of its patrons should seoure the signature of the user]
NAME
ADDRESS
Digitized
by the Internet Archive
in
2016
https://archive.org/details/cognitivestructuOOcoxg
Duke University Library
The
use of this thesis
is
subject to the usual
govern the use of manuscript material.
of the text
is
restrictions that
Reproduction or quotation
permitted only upon written authorization from the
author of the thesis and from the academic department by which
was accepted.
Proper acknowledgment must be given
references or quotations.
FORM 411
1
M
II -49
in all
it
printed
COGNITIVE STRUCTURE: A COMPARISON OF TWO THEORIES
AND MEASURES OF INTEGRATIVE COMPLEXITY
by
Gary Bruce Cox
Department of Psychology
Duke University
!
I
j
A
dissertation submitted in partial fulfillment of
the requirements for the degree of Doctor of
Department of Psychology
the Graduate School of Arts and
Sciences of Duke University
Philosophy
in
in the
1969
ABSTRACT
(Psychology -Social)
COGNITIVE STRUCTURE: A COMPARISON OF TWO THEORIES
AND MEASURES OF INTEGRATIVE COMPLEXITY
by
Gary Bruce Cox
Department of Psychology
Duke University
An
abstract of a dissertation submitted in partial
fulfillment of the requirements for the degree
of Doctor of Philosophy in the Department of
Psychology in the Graduate School of Arts
and Sciences of Duke University
1969
-
7Z.ZP.
CS 776
jf
7&
*7
ABSTRACT
COGNITIVE STRUCTURE: A COMPARISON OF
TWO THEORIES AND MEASURES OF
INTEGRATIVE COMPLEXITY
by
Gary Bruce Cox
This study was intended to assess the generality of a particular type
of cognitive structure characteristic,
suant of this, the theories of H.
that of integrative complexity.
M. Schroder and O.
J.
Pur-
Harvey were com-
pared, and their respective measures administered to 440 Ss of both sexes
from three different southern schools.
The theoretical analysis suggested
truly structural in nature, and
and obviously applicable
position
domain.
which
is
is
to a
is
that Schroder's theory is
more powerful
broad range
much more firmly grounded
in that
of cognitive
it is
more
domains.
more
easily
Harvey's
on the content of the interpersonal
Both theories claim that the characteristic of cognitive structure
most important
in
determining cognitive complexity
(ill)
is
not differ
or an increase in the dimensionality, of the cognitive domain, but
entiation,
the subsequent integration of the differentiated
neither theorist
from
is
Unfortunately,
able to define integration so as to clearly distinguish
a dimensional position.
seems
components.
to be the stronger,
it
Here again, however, Schroder's theory
since
it
is at
of difficulty can be precisely identified.
least explicit enough that the locus
Further, even
if
Schroder
is
un-
able to define adequately the integration concept, his theorizing suggests
the importance of the possibility of super- and sub-ordinate relationships
among dimensions.
Results of the testing were as follows:
tive
measures
each other,
of cognitive integration
(b)
(a)
As expected,
the respec-
were nonsignificantly correlated with
Both measures of integration were significantly correlated
with such measures of intelligence as vocabulary, abstract thinking, and
SAT
verbal and mathematical scores,
may
exist, although the pattern is not clear,
istics
may
(c)
On
not ipso facto inferior to Caucasians, even
point advantage on
Northern whites,
the other hand,
when
when both are
measure and unsatisfactory
scores
is
Negroes are
the latter enjoy a 100-
SAT averages. Nor are Southern whites
at least
scores
Other sample character-
(d)
inferior to
of superior intellectual ability.
Reliability, as estimated by coefficient alpha,
vey's
in the
e.g. large intelligence differ-
affect the distribution of scores,
ences, socioeconomic differences, etc.
(e)
Sex differences
for Schroder's,
is
satisfactory for
(f)
Har-
The distribution
of
such that for both measures complex Ss are rare, so pools of Ss
(iv)
must be tested
more
a
in
problem
order
in
to obtain
adequate numbers of complex Ss.
This
is
attempting to apply Schroder than Harvey, largely be-
cause Schroder has often not bothered to study middle-range Ss, so their
characteristics are unknown.
tially inapplicable to
Schroder's variable (especially)
scores.
essen-
an unscreened group of subjects.
Overall, Schroder's theory
Suggestions were
is
made
for
seems more promising than Harvey's.
improving the reliability and distribution
of
Additionally, a translation of Schroder's theory into dimensional
terminology was attempted, and some important implications of his position for the dimensional orientation
were discussed.
(v)
.
ACKNOWLEDGMENTS
I
would like
to
for the support and
my
committee chairman, Dr. Michael Wallach,
encouragement he has offered throughout
my work
at
Dr. John Altrocchi has also consistently shown a willingness to
Duke.
talk,
thank
which
I
have appreciated, along with the efforts of the other
members
of
my
A
special note of thanks should go to Dr. J. R. Law, an unusually coopera-
committee, Drs. Clifford Wing, Edward Jones, and Donald Burdick.
tive person,
I
am
for the time he has
made
also grateful to the people on
tion has fallen
my
whom
secondmost heavily, namely
Johnson C. Smith University, none
ing
available to me.
of
the
my
whom
burden of
wife and
my
this disserta-
students at
expressed open irritation dur-
periods of preoccupation and neglect.
Both H. M. Schroder and O.
materials.
would like
J,
Harvey supplied me with useful
Dr. Schroder was especially helpful, and in addition to
to thank
him
Gary Gardiner, Jacquie Phares and Jack Carrington
I
at
Princeton for their hospitality as well as their suggestions.
In
Durham, Bob and Joanna Morris frequently opened
refrigerator to me, and Edna Bissette did
my
typing, all of
their
which
home and
I
deeply
;
I
appreciate
i
(vi)
Finally,
I
would
like to thank
Rudolph Cox, who made
by occasionally insisting- on a break.
G. B. C.
(vii)
life
bearable
CONTENTS
ABSTRACT
iii
ACKNOWLEDGMENTS
vi
OF TABLES
x
INTRODUCTION
2
LIST
I.
Components of Complexity
The Generality of Cognitive Complexity
Summary
H.
III.
14
CONCEPTUAL SYSTEMS: THEORY
17
The Theory of Conceptual Systems: Harvey, Hunt
and Schroder
17
Conceptual Systems:
25
Theoretical Modifications
CONCEPTUAL SYSTEMS: OPERATIONS
Schroder's Paragraph Completion Test
Streufert, 1967)
(Schroder, Driver
Harvey's "This I Believe" Test (Harvey, 1966)
A Comparison of the Two Positions
The Present Study
IV.
3
7
METHOD
32
32
39
47
49
51
Subjects
51
52
54
Measures and Procedure
Scoring
(viii)
V.
RESULTS
Preliminary Tests
Complexity, Age, College Class, Sex and
Residence
Complexity, Schools and Intelligence
PCT, TIB and Intelligence
Reliability
VI.
DISCUSSION
Summary of Results
Comment and Implications
APPENDIX
REFERENCES
(ix)
LIST
OF TABLES
Table
1.
Means, Standard Deviations and N's
60
2.
Complexity Score Intercorr elations and N's
61
3.
Preliminary Tests
63
4.
UNC-C Sample Homogeneity
64
5.
Sex x Schools
70
6c
Sex x Schools
71
7.
Correlations Between Intelligence Measures and the
Paragraph Completion Test (PCT)
74
Correlations Between Intelligence Measures and the
"This I Believe" Test (TIB)
75
Alpha Coefficients
79
8.
9.
(x)
COGNITIVE STRUCTURE: A COMPARISON OF TWO THEORIES
AND MEASURES OF INTEGRATIVE COMPLEXITY
Chapter
I
INTRODUCTION
In 1961 O.
J.
Harvey and Ho M. Schroder collaborated, with D. E
0
Hunt, in writing the book Conceptual Systems and Personality Organization
(Harvey, Hunt
Schroder,
1
96
1
) 0
Since that time both Schroder and Har-
vey have actively elaborated the subject.
their theories and
to
measures
This paper
is a
comparison
of cognitive complexity (or conceptual
of
systems
use Harvey's term, or conceptual structure to use Schroder's).
The need
that the
which
is
two
for this
men
are
comparison results from the common assumption
still
dealing with the
strengthened by their tendencies
work as support
same
theory, an impression
to refer to
and
cite
each other's
for their own.
Both theories are concerned with the topic of cognitive complexity or
cognitive structure, a research area which
deciding what
it
means, operationally, by
is
its
having considerable trouble
own terms.
There
is
some
j
|
i
question as to whether the various measures of cognitive complexity are
j
i
-
related to each other.
This concern has taken the form
of a
number
of
i
studies testing the "generality of cognitive complexity,
"
that is,
whether
'
'
3
various measures produce comparable results.
j
The present study may be
j
viewed as an example
of
such work.
Accordingly,
it is
important to ask
I
]
what evidence already exists concerning generality and how
relates to the
it
i
specific question of generality at hand.
to
As
a preliminary
understand something about the theoretical components
it
is
necessary
of cognitive
comi
These
plexity.
will be the two functions of this chapter.
Components
of
Complexity
The cognitive structure theorists share the opinion that
to explain
pattern
of,
and predict behavior
it
in
attempting
will be fruitful to look at the structure or
or relationships among, cognitive elements (opinions, beliefs,
'
•
attitudes,
ideas, etc.) in addition to the content of the elements.
I
Beyond
|
this basic orientation
structure
is
agreement diminishes.
regarded as being
Generally, complexity and/or
a function of differentiation, articulation or
!
discrimination, and integration (Bieri,
1961; Schroder, Driver & Streufert,
i
Other writers (e.g. Scott, 1969) have culled additional structural
1967).
concepts from the literature ("centrality,
11
etc.),
I
but these will be coni
I
sidered here only very briefly.
j
Differentiation
may
be defined either in terms of dimensions (Bieri,
I
1955) or categories (Gardner,
number
1953),
and
in either
case represents the
of stable, unique and important units (dimensions or categories) in
4
realm (content area).
a particular cognitive
measured with
usually
Kelly,
Dimensional differentiation
(REP)
the Kelly Role Repertory Test
(Bieri,
1955) or with multidimensional scaling (Schroder et al.,
is
1961;
1967).
The
semantic differential, with or without factor analysis, might also be used,
Categorical differentiation
may
be assessed using object-sorting tasks,
either without replacement (an object
Clayton
&:
Jackson, 1961: Gardner,
may appear
many
in as
1953) or with replacement (an object
for that matter,
Messick & Kogan,
one category--
1962).
no obvious relationship between the two types
is
between the two ways
Gardner’s scoring procedure
categorical differentiation.
(see
in only
categories as S wishes --Scott,
Unfortunately, there
of differentiation, nor,
may appear
is
and the researcher interested
1963),
of
measuring
unsatisfactory
in object sort-
ing would probably be better off with an adaption of Scott's (1962) technique.
Since categories
may
be based on either different dimensions or different
locations along a single dimension,
some
further
measurement
steps would
be required to relate categories and dimensions.
The
justification for the various dimensional
straightforward.
The
REP
is
measures
ways
that a separate chapter of discussion
explore the subject.
based on the
REP
As
a result, one
unless the
fairly
probably the most commonly used instrument
for assessing cognitive complexity/structure and has been
different
is
way
Multidimensional scaling
is
was used
in so
many
would be necessary
must be chary
the test
used
is
to
of interpreting results
also known.
related to dimensionality in that one
5
learns the number of dimensions necessary to support the relationships re-
ported by the subject.
Due
ing and factor analysis,
models underlying scal-
to the differences in the
it is
not appropriate to
name
the dimensions for
the former, although this is typically the goal in the latter.
The second property
of cognitive structures is articulation (Bieri,
1961, and Schroder et al.,
means
the
number
Articulation
is
is
1967, prefer the
of distinctions one
term "discrimination"), which
makes along
a particular dimension.
expected to vary from one dimension to another.
Also,
it
a dimensional and not a categorical characteristic.
Perhaps the best measure
(in
terms
ency with dimensional techniques) of articulation
Carr also prefers the term "discrimination";
Degree
measure).
mantic differential and
Schroder
et_
ad.
Gorlow
(Sloane,
some
of articulation is to
in
most versions
and
of its flexibility
is
its
consist-
that of Carr's (1965,
Scott,
1969, uses a similar
extent controlled in the se-
of the
REP.
Both Scott (1967) and
(1967) suggest that variables such as equivalence range
&:
1963) and category width (Pettigrew,
Jackson,
Scott leaves
be related to articulation.
it
1958)
with that, but Schroder et
may
al.
attempt to incorporate the findings that abstract Ss have broad "category
width" scores (e.g. Clayton
retical explanation for this
many narrow dimensions
will "„
uli
.
.
&:
Jackson,
phenomenon
1961; Pettigrew,
is that,
1958).
The theo-
given the choice between
or a few broad ones, the S with abstract schemata
lean toward greater integration- -that
is,
per dimension- -and, hence, to fewer dimensions
greater range of stim-
..."
(Schroder
et al.
,
6
1967, p.
to
why
175)
.
No
integration
real explanation or theoretical derivation
is
recognize situations
expected, nor
in
it
more and
will appear as
how
to
finer gradations.
What,
should we expect in the Scott (1962, 1969) sorting task which
number
restricts neither the size nor
(1963) have
there any specification of
offered as
which discrimination will be evidenced by integra-
tion vs, those in which
for example,
is
is
shown
of
that the Clayton and
groups?
Also, Messick and Kogan
Jackson type
of sorting results
are
due more to the failure to categorize than to the number of categories
formed.
(two or
to
So far as
I
know, relationships between size of true categories
more items) with
other variables, a
which Scott's procedure
is
more relevant question and one
applicable, has not been investigated.
pos this topic, Wallach^ has suggested that
it is
important
situations in which Ss are given a category and asked to
Pettigrew, 1958) from those in which S
is
he
is
supposed
may
to distinguish
cases to
it
(as
permitted to provide his own
categories or dimension (as Clayton & Jackson,
important variable
fit
Apro-
More
1961).
broadly, the
be whether S construes the situation as one in which
to find similarities or one in
which he
is
supposed
to find
differences, the complex S presumably being better at either.
The third characteristic
we
will consider,
is
of cognitive structures,
integration.
other theorists (e.g. Scott,
1969),
Although integration
Harvey
et ad.
^M. A. Wallach, personal communication.
and the last which
is
mentioned by
(1961) stand alone in the
7
degree of emphasis they put upon
complexity/abstractness
o
it
as the most important determinant of
Scott's attempts (1966,
measure
1969) to
inte-
gration differ radically from those of either Harvey or Schroder and will
not be discussed at length,
viewpoint
may
(Similarly for Zajonc,
be dealing partially with integration.)
approaches will be dealt with more extensively
it
I960, who,
in
from
Scott's
The Schroder/Harvey
Chapter
III.
In passing,
should be noted that the differences in emphasis on integration are not a
logical consequence of theoretical differences.
for
James
Bieri,
structs (Kelly,
who has roots
1955),
more systematic
than,
which
is
particularly true
George Kelly's theory
of personal con-
strikingly similar to, and in
is
Harvey
in
This
et al.
The Generality
some respects
(1961).
of Cognitive
Complexity
Questions about the generality of cognitive complexity have two distinguishable components, related respectively to theory and operations.
Essentially all theorists agree that complexity will vary across cognitive
domains or content areas
1962) or within the
(Bieri,
1961;
Gardner
&:
Schoen,
1962; Scott,
same content area across time (Zajonc,
function of situational variables (Schroder et_al.,
1967).
ever, is the methodological question of whether various
I960) or as a
Beyond
this,
measures
of
how
com-
plexity at a particular time and with regard to a particular topic would
8
order Ss similarly.
Among measures
or discrimination,
of
some
complexity which emphasize dimensionality and/
findings favoring generality have been reported.
For example, Bieri and Blacker
the
REP
(1956) found complexity as
and by Rorschach perceptions
test
son (1963) administered three different
moderately high levels,
formance on
Hess
(1966)
to
REP
be related.
tests,
measured by
Allard and Carl**
which correlated
at
compared REP complexity and per-
a sorting task and reported that the two
Todd
were related,
and Rappoport (1964) discovered that two different models of cognitive
structure (factor analysis and Hays' implication model) produced highly
similar interstimulus distances which, in turn, were related to sorting
performance.
Since these findings are somewhat tangential to our major
interests, no particular effort will be
More immediately
made
to evaluate
them.
relevant results which favor generality are re-
ported by Carr (1965) who states that conceptual system level
discrimination, but not to
REP
is
related to
fluency, and Harvey's (1966) and Wolfe's
(1963) findings of relationships between
REP
scores (not fluency) and con-
ceptual level.
I
know
of three studies of generality
scope than the above.
to the
which are more extensive
Of these, Scott's (1969)
is
in
not particularly relevant
immediate purpose here, because his theoretical orientation
is
dimensional, and he includes neither Schroder's nor Harvey's measure of
integration.
In addition to dimensionality
and articulation, Scott does
r
i
9
measure what he defines as
integrative characteristics (Centrality, Affec-
j
tive Balance,
and Affective -Evaluative Consistency), but each of these im-
plies a reduction in the dimensionality of the cognitive space, a position very
different
from
that of
Schroder or Harvey,
Scott
measured each
of the five
characteristics above in several different ways and found modest intra-
The characteristics
characteristic correlations across the various scores.
correlated with each other in the predicted pattern, but this evidence for
generality within Scott's model does not bear on the question of the generality of
Schroder's and Harvey's measures of integrative complexity.
The best-known study
on his dissertation (1964).
tion Test (PCT), but not
of generality is
Vannoy's (1965) article based
Vannoy includes Schroder's Paragraph Comple-
Harvey's "This
I
Believe" (TIB) Test (both the
;
PCT
and the TIB are intended
to
measure integrative complexity), plus
a
I
i
number
plexity.
of other tests of varying
He concluded
and sometimes uncertain relation to com-
that complexity is not unitary but
comprises two or
|
three uncorrelated components.
serious flaws.
No one claims
Unfortunately, the study contains several
that complexity is unitary.
most theorists hypothesize components
complexity will vary with content.
ality issue
would have
failure to keep
them
in
to be
of complexity,
I
have noted,
and most suggest that
Any adequate examination
of the
gener-
organized around these points, and Vannoy's
mind has resulted
in
an interpretive jungle.
not all of the tests he used are primarily intended to
A
As
First,
measure complexity.
case can be made that the scores might be related to complexity, but how
'
10
much and
in
what ways
is
For some measures,
not clear.
viz., the "social
distance" questionnaire, the scores as used (variance components) are absolutely ambiguous.
the extent that there
Second, the cognitive domain varies
is
a consistent domain,
it
among
To
tests.
would be interpersonal, but
even this rubric incorporates a broad range of types of stimuli, varying
from four adjective descriptions
nated by the Ss to
fill
of hypothetical people to real
particular roles.
Clearly,
some
of the
persons nomi-
measures
cluded should not have been, and some important procedures were
Of most immediate interest here
is
tor by any other complexity
no other high loading
noy does
the
PCT
not,
is
measure (SCAT -Q
itself.
It
PCT
not accompanied on that facis
second highest) and shows
would seem natural
apparently because he did not attend
to notice,
to
although Van-
such distinctions, that
the only test in the battery specifically designed for
is
left out.
the finding that Schroder's
loads highly on the last of Yannoy's factors,
in-
measuring
integrative complexity.
A much
better study
is
that of
Gardiner (1968).
sensitivity to "integrative complexity,
"
Gardiner's greater
as espoused by Schroder and
Har-
*
vey, leads
him
to include a total of six
measures
of this variable.
Scott's
(1962) sorting task is included, but is applied to nations (as Scott used
so
its failure to
relate to other tests
plexity between domains as
ality
true
(if
sorting
measure
much
may
be due to independence of
it),
com-
as to independence between dimension-
measures dimensionality) and integrative complexity.
of dimensionality is included (e. g. Carr,
1965).
The
No
REP
11
appears only
number
in a
check
of variables
complexity.
In
list
"version.
"
Also included
in the
battery are a
which are expected on theoretical grounds to relate
most cases, these derivations are
to
quite straightforward:
generally, the theorists themselves have stated that such-and-such a characteristic is related to complexity, and Gardiner has chosen a
commonly
accepted (or acceptable) measure of the characteristic.
The major findings are,
plexity (the TIB,
PCT
first,
that the
measures
of integrative
and Interpersonal Topical Inventory, or
ITI,
com-
a
forced-choice objective version of the PCT, two scores being derived from
each of these tests) define the first factor
that the third factor,
appears
to
in a factor analysis; and,
which Gardiner pays less attention
approximate the idea
loading on this factor are the
of a general
PCT, TIB,
thinking.
tor,
to than the second,
complexity factor.
Tests
Scott's sorting task, and
bedded Figures Test, plus three other tests requiring
second,
flexible,
Em-
analytic
In fact, the label "Analytic Thinking" could be applied to the fac-
an interpretation which would seem consistent with Schroder's (1969)
description of complex thinking.
Gardiner suggests that the factor relates
to dimensionality or cognitive differentiation, noting that the
PCT
relate to dimensionality, while Scott's nation sorting task
"specifically
is
and TIB
dimensional." In that case, the loadings belie a strictly dimensional interpretation,
since both scoring versions of the TIB, as well as the
Figures, load
more
Embedded
highly than does the Scott task.
The third major finding
is that,
according
to
Gardiner's original
12
factor analysis, a
to
number
of characteristics
covary with complexity do
which have been hypothesized
including two
not,
measures
of flexibility.
Associations IV and Gestalt Transformations; two tests involving acquiescence, Social Acquiescence and Seeing Deficiencies; a scale of Internal-
External Causation; and two measures based on tolerance for inconsistency
in attributed
may
descriptions (references for these tests
However, there
Gardiner, 1968).
is
some question
as to the appropriate-
ness of testing such hypotheses with factor analysis.
analysis
is
that
it
be found in
The virtue
compresses or organizes large quantities
the process, finer details
may
of factor
of data.
In
(Additionally, of course, there is
be lost.
the basic arbitrariness of the location of the dimensions and thus of the
Thus, whenever possible, hypotheses should be tested
factor loadings.)
without first subjecting data to the leveling effect of factor analysis.
This
more
direct type of testing
ses in the study;
(A) that there will be
(B) that firstborn Ss will be less
is
used
in
examining three hypothe-
no sex differences
in
complexity,
complex than laterborn, and
(c)
that in-
discriminantly pro-religious Ss will be less complex than others.
hypothesis as stated
is
based on theories about complexity.
Each
There should
be no sex differences because no one has ever given grounds for any, or
suggested that there would be.
Anticipated differences between first and
laterborn are based on widespread findings (reported
that firstborn are
in
Gardiner,
more dependent, acquiescent and conforming
1968)
and,
more
importantly, that these differences are based on childrearing practices
13
which differ
in
such a way, according to the theorizing of Harvey
(1961), as to justify the current prediction,,
et al.
The relationship between re-
ligious conviction and complexity is based largely on Harvey's (1966) state-
ments and scoring criteria
Each hypothesis was tested by
for the* TIB.
dividing Ss into the two appropriate groups and using
t
tests to
compare
scores on the other variables.
In
each case, the hypothesis was not supported.
complex than males.
some
Gardiner explains
Females were more
this rather plausibly
by invoking
clas s -related differences in parental treatment of boys and girls.
The failure
to find birth
order effects
is
not well handled, however.
Since
he argues that the sex differences are due to differences in child-rearing
practices, and that the anticipated birth order differences would have been
due to child-rearing practices,
example,
if
a disproportionate
it
may
number
be that the two are connected.
of firstborn
For
were also females,
the
hypothesized birth order effects could have been washed out by the hypothesized sex effects.
Finally, indiscriminantly pro-religious (IPR) Ss
more complex than
tempts to explain
were found
other Ss, a result opposite that expected.
this first
sically oriented than the
to be
Gardiner at-
by stating that the IPR Ss were more intrin-
non-IPR Ss
(a
"complex" characteristic), and,
second, by pointing out that the Ss were young (15-16 years old) and thus
not yet fully differentiated regarding religion; i.e. that the
IPR group
in-
cluded both the genuine IPR plus intrinsically religious Ss who are currently
14
also global but will not be in a few years.
it
also avoids
some
still
which complexity
are more complex.
the
measures
of
is
if
religion
measured.
This
not unreasonable,
is
were
a distinct
In the present case,
domain from
lief
It
that
particularly unexpected in light of the fact that
complexity which reach significance are based on the TIB
I
function-
would seem that the hypothesized relationship between religious be-
and integrative complexity
On
this basis, then,
it
is
not present.
would seem that the
PCT
and the TIB are re-
even though Gardiner does not report the correlation between the
lated,
two.
At
however, IPR Ss
scoring criteria which state that IPR behavior represents Level
ing.
is
expect relatively complex- Ss to be non-IPR.
worst they might be unrelated,
in
is
Ss are all of an age, thus differentiation
issues.
One would
relative.
While this
Unfortunately, there
dicts this finding.
is
Schroder
one further piece of evidence which contraet_
al
.
(1967, p.
115)
comment
virtually en
passant that the two tests bear only a "low-order relationship."
picture
is
So the
confused.
Summary
Both Schroder and Harvey share the position that integrative complexity is the
makes
most important component
their position distinguishable
of cognitive complexity.
from
that of the
This
more dimensionally
15
oriented theorists (Bieri,
immediate question
is
1961; Scott,
1969).
It
also suggests that the most
whether Schroder and Harvey agree on their mea-
sures of integrative complexity, rather than whether integration
is
related
to dimensionality.
Regarding the comparability
measures,
there
first,
stitute a separate,
its
own
factor, and
is
evidence that
common
factor:
factor analyses the two tests con-
Vannoy (1964) found
all
however, and Schroder
et al„
the
PCT
PCT
defining
and TIB (plus the
load highly on a factor by themselves.
The correlation between the measures
’'only a
in
Gardiner (1968) showed that the
ITT, a derivative of the PCT)
bear
Schroder and Harvey positions and
of the
(1967, p.
low -order relationship."
T
is
not reported in the Gardiner study,
115)
have stated that the two tests
Thus, the evidence
is conflicting.
Second, although they are not predicted, sex differences in complexity
have been found in at least one case (Gardiner, 1968).
ences were attributed to differences
sexes, but
it
in
child-rearing practices for the
not impossible that other, albeit
is
These differ-
more
subtle,
sex-related
differences in child-rearing practices and/or cultural role expectations
would affect complexity
of
at other
socioeconomic levels.
whether there are sex differences
There
here.
tween
is
complexity
is
not settled.
another area of ambiguity which will be briefly mentioned
Schroder
PCT
in
Thus, the question
et al.
(1967, p.
197) also state that the correlations
scores and intelligence have ranged between
various samples.
This
may remind
.
be-
12 and .45 for
us that the second highest loading on
16
Yannoy's
to
"PCT
factor” was
SCAT -Q, and perhaps
also of the analytic cast
Gardiner’s (1968) third "general complexity” factor.
(One of Gardiner's
greatest deficiencies was his failure to include an intelligence measure.)
The general question
is,
then,
what
is
the relationship between intelligence
and integrative complexity?
A somewhat
related question
is
that of
sample characteristics.
Not
only are Ss typically college students, but college students from fairly un-
representative schools.
The potential
effects of these biases have
appar-
ently never been fully considered.
Before further developing these points, some background discussion
of theory
seems appropriate.
Chapter
II
CONCEPTUAL SYSTEMS: THEORY
The
fullest description of this theoretical position is found in
Hunt and Schroder's (1961) book.
The basic ideas from
this
Harvey and Schroder
(1963),
Harvey,
source appear
I
in briefer,
more readable form
in
Harvey
I
(1966), and to
some degree
in
Schroder and Harvey (1963).
In general,
it
j
*
i
appears that Harvey continues
of the 1961 theory,
to
espouse an essentially unchanged version
whereas Schroder harbors some revisionist tendencies,
I
hints of which
may
be noted in the Schroder and Harvey (1963) paper.
i
'
The Theory
Conceptual Systems:
Harvey, Hunt and Schroder
of
Conceptual systems are composed of concepts and their interrelations.
!
r
The
definition of
"concept" varies somewhat from place
to place,
at one
i
point being defined so as to imply a dimensional characteristic
system
of ordering that serves as the
(
17 )
(".
.
.a
mediating link between the input side
-
18
[stimulus] and the output side [response]" p„
1
),
in
another case suggest-
ing that concepts are points along dimensions (concepts represent "a cate-
gory of varying definitiveness and breadth along some specifiable dimension.
o
p.
.
10),
and
minimally a comparison
in still an'other situation saying that a
of
two points on a dimension, and
concept
may
be a point
or points in a space of infinite dimensionality ("The simplest concept
involving two points on a single dimension, and the
is
is
is
one
most complex concept
one embodying an infinity of points on an infinity of dimensions." p, 62).
These differences are probably not
The basic idea
ent as well as overdrawn.
George Kelly's
(Kelly,
critical, although the last
is
seems
apparently quite similar to
1955); namely, that organization is
imposed on per-
ceptual experience, that "meaning" (or "understanding") derives
imposed organization, and
syncratic and arbitrary.
Concepts ("constructs" and "contrasts,
concepts are the content of conceptual systems.
to
from
the
that this organization is in varying degrees idio-
Kelly's terminology) are the terms in which the world
wish
differ-
is
" in
viewed; that
is,
Harvey, Hunt and Schroder
regard differentiation-integration as a process, and concreteness
abstraction as a result of the process.
dimensions
is
Consequently, neither of these
inherent in the stimuli, but rather in the conceptualization of
the stimuli.
An increasing number
"'‘In
1961
.
this section,
of concepts implies increasing differentiation
page numbers refer to Harvey, Hunt and Schroder,
19
on the part of the person, but greater abstractness
comitant of greater differentiation.
is
not an automatic con-
True abstractness requires integra-
tion of concepts as well as differentiation, and differentiation
without subsequent integration.
Also, although differentiation
may
occur
in
is,
some
sense, a logical prerequisite to integration, these authors argue that the
two can occur essentially simultaneously.
clearly
is to the
"Integration
is
Unfortunately, important as
theory, the nature of integration
the relating or hooking of
previous conceptual standards"
(p„
18).
.
A
.
.
is
it
never really made clear.
parts to each other and to
straightforward interpretation
would be that differentiation involves seeing differences and integration involves seeing similarities, both of which can be along dimensions.
How-
ever, the book consistently suggests, and there are explicit statements
elsewhere
(e. g.
Schroder
et al „
.
1967, p.
tive rules are qualitatively different.
concept
is a
near
fatal
weakness
7),
that
dimensions and integra-
This failure to clarify the integration
in the theory,
and one which persists
in
the current positions.
Given proper parental training practices
(to
be discussed below),
children will progress from concrete to abstract functioning.
If
conditions
are unfavorable, arrestation will occur at some point along the dimension,
the stopping point being determined by the nature of the unfavorableness.
Of course, the level
to
which a person progresses
is
likely to vary
from one
content domain to another, again as a function of varying training conditions.
Harvey, Hunt and Schroder have concentrated on the interpersonal domain,
20
which
is
appropriate, but
would generalize
to a
how and
the degree to
which their discussion
more content-oriented domain
"history")
(e. g„
is
not
always clear.
introduction into a new, unfamiliar realm of experi-
In the theory,
ence necessitates relatively concrete behavior, i.e. dependency on external
sources or authority.
With increasing familiarity the potential for increas-
The authors
ingly abstract behavior develops.
mental theorists
in
cite a
number
of develop-
support of their argument that children typically pass
through several concrete to abstract cycles, the concrete phases corresponding to the times at which children begin to talk (roughly age
social behavior (age
5),
enter peer -centered society (age
take on adult roles (apparently seen as about age 17).
gresses
level,
in the
10),
The
2),
begin
and begin to
first cycle
pro-
direction of abstraction only as far as the second stage or
or perhaps to the transition between the second and third levels.
Later cycles progress as far as Level IV.
The
first stage (Level
I)
is
labeled "unilateral dependence."
also be termed a "pro-authoritarian stage":
to be the
same (provided emphasis
is
the connotative baggage
^
which
is
might
seems
on style or process, not content).
differentiation and integration occur to a proper degree, Level
reached,
It
II
If
will be
described as a period of negative independence, and
in
of the elaborateness of the book, which is considerable, is due
to extensive consideration not only of the major stages but also of the transitions between stages. This material seems to have dropped out later and
Much
will not be considered here.
21
at least
some forms bears
The categoricalness
is
a strong
similar to Stage
levels is on the basis that at Level
from
of
to anti-authoritarianism.
The distinction between the
I.
the person is differentiating himself
II
whereas
the external authority,
terms
resemblance
at
Level
I
he
is
defining himself in
Exploring this differentiation takes the form of resistance to
it.
the authority.
Integration of the two possibilities, necessary for progress,
would make available
types of behavior.
to the
person some form of both attitudes and both
Fixation at this level, a failure to integrate, would leave
only a strong negativism.
Level
III is
characterized by conditional dependence and mutuality.
In this stage, hypothetical or conditional ("as -if") thinking
sible,
and a more objective, information-oriented attitude
Socially the person at this level of abstraction
the world.
becomes posis
taken toward
is
able to see
another's point of view and to understand the other person from that point
of
view.
Additionally, the third stage permits the "holding of alternative
views of the
self,
of events,
and
of
concern for ambiguity"
of
dependency and/or a diffuseness
101).
(p.
defined as the stage at which
simultaneously with a
of others
"...
By implication there
in the
is still
minimum
an element
self-concept, because Level IV
mutuality and autonomy are integrated
so that neither interferes with the other and yet both are important"
In science
and other noninterpersonal endeavors, Level
empirical orientation; so presumably IV
the analogy with mutuality and
is
is
III
(p. 106).
appears as an
the theoretical stage, although
autonomy does not appear
to be strong.
22
Level IV functioning
is
characterized by an increased sense of internal
causality as well as increasing problem-solving ability, tolerance of anxiety,
ambiguity and inconsistency, greater resistance
to stress,
and a
stronger orientation toward informational solutions to questions or prob-
lems
.
If
training conditions are unfavorable, then development will be ar-
rested at some point along the scale, the point of arrestation being deter-
mined by
Again, a super -articulated
the nature of the training practices.
theory will be greatly simplified, for two reasons:
to
conserve space, and
because the theorists themselves have apparently paid no serious attention
to the elaborations in their
Training
of these,
ing are
is of
in turn,
(1)
own research.
two major types, unilateral and interdependent.
dependent, or
(4)
Thus, the four major types of train-
has two subtypes.
reliable unilateral,
unreliable unilateral,
(2)
informational interdependent.
arrestation of the trainee at Levels
Each
I
protective inter
(3)
These dispose toward
through IV, respectively.
Reliable unilateral training implies an authoritarian orientation in
the trainer.
nal,
That
is,
the criteria for
rewards and punishments are exter-
"objective, " rigid, evaluative and absolute.
Good behavior, as de-
fined by the criteria, is consistently rewarded, and bad
punished.
There are few,
if
is
consistently
any, lapses in either direction.
of the criteria is not permitted.
Questioning
Requirements, however, must,
sense, be within the capacity of the trainee.
in
Training of this sort
is
some
hypo-
'
.
23
thesized to lead to Level
I
functioning.
Unreliable unilateral training (defined from the trainee's point of
I
view) occurs primarily either
external criteria
(e. g„
when
the trainee is unable to discover the
when feedback
is
too erratic) or
when he
unable
is
I
to
conform
to
it
(as
when
the
demands are beyond
his ability).
In these in-
stances, the trainee either sees no pattern of reward and punishment, or
A
builds up an expectation of failure,
enced as unreliable occurs
if
may
be experi-
parents are indifferent toward and neglectful
Fixation at Level
of their children.
third condition which
likely to follow
JJ is
from any
of these
circumstances
Protective interdependent training
for instrumental behavior,
where
is
based on systematic rewards
the goal of the behavior is not evaluated
against external criteria as in the reliable unilateral condition.
case, support
ior.
is
used both as
This support
is
a
reward and as
likely to be so
a
extreme as
protect the trainee, insofar as the agent
is
means
to
In this
of directing
become
behav-
a tendency to
capable, from failure experi-
ences,
Thus, although relativism and an emphasis on interaction exist,
there
also a restriction as to the kinds of experiences the trainee
is
mitted, and progress beyond Level
III
is
per-
will probably not occur.
Under informational interdependent
training, the trainee is permitted
i
to
learn from the consequences of his own experiences, including failure,
with the important proviso that the environment be so structured that the
subject
is
not pushed beyond his capacities.
So long as this
is the
case,
24
the trainer feels, protection
as success.
It is
is
unnecessary and failure
is
this informational aspect of experience
sized and valued by the trainer.
as informative
which
is
empha-
Such an approach permits the development
j
of the abstract skills required for Level
For present purposes,
IV functioning.
this is an adequate
summary
of the 1961 posii
Modifications in the theory will be considered below; however,
tion.
it
should be emphasized that what Harvey, Hunt and Schroder set out to do was
i
to
describe a cognitive variable which
is
independent of content but which
i
would predict certain styles
of behavior.
cessful at a descriptive level at least.
Seemingly, they have been suc-
For example, theoretically the
i
authoritarianism of Level
I
may
be of the
left or right,
pro- or anti-
religious, behavioristic, humanistic, etc., and supposedly the
true for each of the other levels.
ward a closer relationship with
As we
will see,
same
is
Harvey has moved to-
content, and Schroder has
moved
further
i
i
away
c
Before discussing such changes, however, a problem inherent
in the
|
development and training arguments should be mentioned.
Basic
to the
!
problem
is
the fact that by either
measurement technique
below) complex Ss will comprise only
different
the
meanings --Level IV
way between
II
and
III
for
4%
to
10%
of a
(to
be discussed
sample ("complex" has
Harvey and approximately a quarter
for Schroder).
That
is,
in
of
young adult (college)
i
populations, complex people are quite rare.
This would seem consistent
j
with the description of the training agency hypothetically required to proi
i
<
25
duce abstract thinking.
mental cycles,
in
which
It is
not consistent with the discussion of develop-
it is
stated that the
havior for nine -year -olds
is at
That
"
is,
at
age
(i.e.
most common) be-
the level of transition into the fourth stage,
and that Level IV functioning "is
development.
modal
rriore likely in the
9 or 10 the
adolescent cycle of
most common functioning
very
is
nearly the most abstract possible, but by age 20 this same abstract behavior has
become
the least
were carried away
common
found.
My
suspicion
is that
in their desire to integrate a variety of
als, although there is at least one alternative explanation;
the theorists
source materi-
namely, that
parents are relatively interdependent and schools relatively unilateral
trainers, and that by college the increased proportion of time spent in
school, plus selection, have done their job.
likely,
This hypothesis seems un-
however.
Conceptual Systems:
No
Theoretical Modifications
significant theoretical revisions are apparent in Harvey's
recent writings.
The most recent exposition
a concise,
of
convenient
which
I
am aware
summary
more
(for
ex-
heavily based on
ample, Harvey, 1966)
is
Harvey
(The tendency toward closer ties to content will be
et al.
discussed
in
(1961).
conjunction with measurement.)
over the term "integration,
"
There
but this results not
is
reduced confusion
from elaboration
of a
con-
26
cept which
is
different
the concept until
it
from dimensionality, but from
becomes
itself essentially
a simplification of
dimensional.
Integration
is
said to be "interrelating, " which can easily be given a dimensional interpretation.
This, in turn,
means
that there has
become
little distinction
tween Harvey's theoretical position and a strictly dimensional definition
beof
complexity (e.g. Bieri, 1966).
Schroder (1969) stands
of
in
some contrast
to
Harvey, since a number
changes have been introduced into his theorizing, most of them having
to
do with the nature of the concrete -abstract dimension.
First, specification of the dimension has even less to do with content
The dimension
or with behavioral styles than in earlier material.
ceived in
more
truly structural terms.
con-
is
Schroder no longer specifies the
nature of the poles of the developing dimensions, and the specification of
the behavioral characteristics
of the cognition as
the
model
is far
opposed
more
is
more obviously derived from
to the nature of the
the structure
dimensional poles.
Further,
easily generalizable to domains other than inter-
personal, and the progression through the stages
iy.
and more easily seen as developmental.
the
model enormously.
is
more
logical and order-
All of this has strengthened
Second, the ideas of differentiation and integration have been amplified,
and
in the
process definitions have changed.
Finally, there is a
strong tendency toward specification of three instead of four levels.
is
This
probably for practical reasons, as virtually no one operates even at
.
27
Level
on the device used by Schroder to measure conceptual level.
III
More concretely
(if
that can be permitted)
current position (Schroder,
fined as "the
a given
number
domain
viewing stimuli.
number
number
of scales or
to
is
now defrom
of unique principles of stimulus
dimensions used by the person
in
These scale or dimensional values are the informational
and for a given stimulus the amount
person
Differentiation
of categories or kinds of information extracted
of stimuli, " or the
ordering, or the
input,
1969) is as follows:
and briefly, Schroder's
person with the degree
of
information will vary from
of differentiation.
It is
hypothesized that
people learn to combine these values in various ways according to learned
rules.
The integrative complexity
person's conceptual structure
of the
is
determined by the number of these combinatory rules and by the "degree
connectedness" between the rules.
A
combinatory rule
weights assigned to the scale values, so
if
is
of
a pattern of
stimuli are scaled along three
dimensions, and the values along the scales are weighed 50
,
30
,
and 20
,
then the resultant values have been arrived at by a combinatory rule (alternately termed a "perspective").
Level
functioning occurs
I
such combinatory rule, or set of weights,
scales.
In simplest terms, this
means
is
when only one
available for a given set of
that a concrete
person
is
one who
can see only one alternative or one conclusion in a particular situation.
When two
or
more
alternatives, patterns of weights, understandings, per-
spectives, interpretations, judgments, or whatever, are available, simul-
taneously and for the same information, then the person
is
operating at
.
28
C onnectedness between these two or more perspectives plays
Level
II.
same
role as integration or integrative rules in Harvey, Hunt and Schroder
necessary condition for Level
(1961), and is the
Schroder considers„
"integrative rules,
"
"Connectedness"
which
is
the
this the concept is
every
'functioning, the highest
bit as
poorly defined as
most obvious and immediate criticism
To connect perspectives
the theory.
is
III
the
is to
simply not developed.
of
relate perspectives, and beyond
My
best guess
is that
connecting
perspectives means to compare them against some superordinate, relative
construct, like goals, values, or motives.
In spite of the terrific flaw resulting
from Schroder's
ability to specify the nature of connectedness
tion intuitively and aesthetically
more
more
fully,
I
failure or infind his posi-
attractive than Harvey's.
This
is
not to say that there are not other problems, or that there are not viable
alternatives
It
seems clearly
binatorial rules,
stimuli in a
necessary
may
to be the
case that perspectives, the results of
be regarded as dimensions.
new way, which
to locate the
is the definition of
com-
Each perspective orders
a dimension.
new dimension/per spectives
is to find
All that
is
an axis such
that the cosine of its angle with each of the dimensional axes is equal to the
weight given ratings on that dimension, expressed as a decimal fraction.
There may be cases where
suspect
I
it
this
would not be possible,
but,
in general,
I
could be managed.
can imagine two objections Schroder might raise.
First, the
new
29
factor would be oblique, and he specifies that informational dimensions be
mutually orthogonal.
The urgency
requirement
of this
inasmuch as orthogonal dimensions are probably rare
is
me,
not clear to
at best,
and
may
be
j
psychometric inventions.
more common.
least,
Oblique relationships are probably, at the very
(Schroder uses "friendliness” as an example of a
t
dimension:
to
what
is
dimension orthogonal?)
the friendliness
It is
not
I
I
j
even clear that research reported later
quiremento
In one study,
dimensions:
in the
stimulus persons were described along three
orderliness, and reciprocity, which Schroder re-
creativity,
The sense
ports as being independent.
in
which they are independent
it
may
clear (statistically or cognitively), and
would
feel
chapter conforms to this re-
moderate dissonance
if a-
target
well be that at least
is
some
not
Ss
were described as very creative
and also very orderly.
I
The second objection
to
viewing perspectives as dimensions might be
the "kinds of information ex-
based on the statement that differentiation
is
tracted from a given domain of stimuli."
The implication
is that
informa|
tion is
somehow
whereas perspectives are the re-
intrinsic in the stimuli,
sult of cognitive processing.
I
regard
this as
an untenable distinction.
;
Examples
of
dimensions used by Schroder have been mentioned above
,
(friendliness, creativity, orderliness, reciprocity).
The only case
in
which
i
|
these are "given 11 occurs when an experimenter tells a subject that a stim-
ulus person is "friendly,
of this information will
" etc.,
and even
vary among £s
.
in this situation the
connotations
Normally, making judgments along
30
dimensions will involve processes like those involved
differences,
if
any, will have to do with the ease and/or speed with which
the process operates.
parable
In general, perspectives
and dimensions seem com-
.
Because the nature
cult to
The
perspectives.
in
of
connectedness
so unclear,
is
attempt a dimensional reinterpretation.
ness" means
’’relating, ” this
suggests to
me
it is
However,
if
more
diffi-
"connected-
the finding of similarities and
differences, which, in turn, suggests dimensions or concepts (equivalent
in
Schroder's terminology).
or motives (one's
If
connectedness
is in
terms
own or others, simultaneously or one
of goals, values
at a time),
I
sus-
pect these could be represented as idealized dimensions or points in the
space, against which comparisons could be made.
This predilection toward a dimensional model demonstrated above
has two bases;
first,
a reaction against the Schroder and
Harvey claim
that
they are dealing in concepts which dimensional approaches cannot explain,
especially since the integration concept
an interest on
my
is
never specified, and, second,
part in pushing the dimensional model to
introducing additional constructs.
However, even
interpretation of conceptual levels
is
if
this
plausible, Schroder
portant point, and one which a dimensional approach
may
its
limits before
dimensional reis
raising an im-
have trouble re-
presenting; namely, the issue of super- and subordinate relationships
among dimensions
or perspectives.
such hierarchical arrangements, one
Kelly (1955) identified two types of
in
which the dimensions
lie,
as
it
'
31
were, parallel to each other (e.g. "intelligent -stupid"
is in
some degree
subordinate to "good-bad"), the other in which both poles of one dimension
are subordinate to one pole of the superordinate dimension (Kelly's ex-
ample:
"light-dark" are both desc-riptive on a "descriptive -evaluative"
dimension)
o
These examples could probably be handled by leaving empty
quadrants in the space, but this seems inelegant and the general issue could
quite possibly prove thorny.
Schroder's connective rules, then, would probably have to be viewed
as superordinate dimensions
functioning,
in
if
they are to be dimensions at
all.
Abstract
Schroder's sense, would be dependent upon the development
and use of such superordinate dimensions.
have argued that abstraction
is
Conceptual systems theorists
not related to dimensionality.
This would
be an agreeable statement so long as the dimensions considered are at a
given ordinal level.
In passing
commonly used
1955),
ity (e.g.
Bieri,
1966),
to
is
measure
it
a dimensional
form
REP
test (Kelly,
of cognitive
complex-
is
needed which will reveal such structure be-
can be determined whether the relationship proposed above exists.
A comparison
Harvey
should be noted that the
not used to distinguish super- and subordination.
Some dimensional measure
fore
it
of the current theoretical positions of
will be delayed until their respective
Schroder and
measures have been consid-
Chapter
III
CONCEPTUAL SYSTEMS: OPERATIONS
Although a number of procedures are viewed as yielding scores rei
levant to structural complexity, in practice both Schroder and Harvey have
j
evolved particular tests which they regard as criterion measures.
The two
i
tests are both similar and different and will be discussed in this chapter.
Schroder's Paragraph Completion Test
(Schroder, Driver & Streufert, 1967)
The Paragraph Completion Test (PCT, Appendix
ject with a
word
I)
presents the sub-
or phrase which is supposed to be the beginning of a sen-
tence, and requests S to finish the sentence and write two or three further
sentences on the same topic.
nonscored item.
Each item
There are
is
five
such items, plus a warm-up,
timed, the limit ranging from 100 to 130
seconds
Items are rated by trained judges, scores being given on a
(
32
)
7
-point
j
'
W'MT-
33
scale, the odd
sum)
numbers representing
of the two highest
item scores
the four
is
The mean
major levels.
the preferred overall score.
(or
Interi
i
rater reliabilities for this score reportedly range from .80 to .95, with
inter-item correlations mostly
No
ranging between
in the .40s,
.
39 and
62.
.
test-retest reliabilities are given.
Scoring criteria are as follows (based on Schroder
Appendix
A
II):
score of
1
et al.,
("Low Integration Index" or Level
I)
1967,
is
given for
a response generated by a single rule, where alternatives are neither present nor considered, and
tions.
Additionally,
(a)
where thinking
conflict is
guity as indicating weakness;
(b)
categorical and without condi-
is
viewed as bad, and uncertainty or ambi-
S seeks fast,
ternally consistent, balanced processes;
(c)
unambiguous closure via in-
a specific rule or guide is of-
fered which leads to or implies an absolute, categorical solution;
compartmentalizes or categorizes facts and effects and tends
(d)
S
to treat only
one side of an issue.
Score
2:
An
Score
3:
(Medium-Low
absolute score
is
qualified, but no clear alternative is
given.
Integration Index, Level
II):
Two
or
more
"alternative rule structures" (or perspectives or interpretations) for viewing 'the event are simultaneously present, plus possibly a conditional rule
for selecting
or
more
among
the alternatives.
Indicators include
different interpretations of events;
ferences without considering relationships;
(a)
mentioning two
(b)
listing similarities or dif-
(c)
giving conditional rules for
34
choosing among ways of categorizing;
(d)
reacting against absolutism and/
or avoiding dependence on external forces.
Score
4;
Alternatives are offered, with the implication of interaction,
but the "interaction is expressed as a qualification rather than as the
emer-
gence of comparison rules."
Score
5;
(Medium-High Integration Index, Level
terpretations are given and interact in
cally,
some specified way.
the alternatives are related in a
(a)
guity or conflict;
(b)
(d)
Alternative in-
More
way which maintains
specifi-
the
ambi-
the existence of the alternatives is itself regarded as
having various meanings;
consideration;
III):
(c)
other people's points of view are taken into
indication that one's behavior
is
influenced by others' be-
havior and/or that social interactions should be based on mutual responsiseeing alternative reasons for similarities and differences in per-
bility; (e)
spectives.
Criteria for scores of 6 and
IV) are described,
7 (7 is
the High Integration Index, Level
but such responses are exceedingly rare and will not be
considered here.
It
should not be difficult to achieve the interrater reliabilities
Schroder reports.
The criteria
for scores
above
5
are extremely abstract,
but since such responses are rarely encountered there
One
difficulty
(Score
1)
would be
in
is
no great problem.
discriminating between "polarity responses"
and true alternatives (Score
3).
The former are alternative de-
cisions based on a single rule (e.g. "good or bad"), whereas the latter are
;
35
based on different rules (something might be good and bad simultaneously).
Once these can be separated, rater
reliability should not be
hard
to
estab-
lish.
There
however.
is
some
Schroder
potential confusion about the distribution of scores,
et ad
.
(1967, p.
195) indicate that,
for the
most con-
crete and most abstract of their samples to date, approximately
12%
of the Ss (respectively) obtained scores of 7 on a 7 -point scale, i.e.
obtained the highest possible score.
the
4% and
mean
of all his
Current practice
a score,
At that time a S's score was either
item scores or the mean of the two highest item scores.
is to
add the two highest item scores together
so these are distributed between 2 and 14
(a
to decide
13-point scale).
The
confusion lies in the fact that the scores on this 13 -point scale are seldom
higher than
7
and that this distribution of scores on the lower half of the
13 -point scale (the lower half of the complexity dimension)
taken for the distribution reported
in
Schroder
et al.
may
be mis-
for the entire 7 -point
scale (the whole dimension).
Schroder explains
been a change
in
this shift in distributions
^
by stating that there has
scoring procedures (the changes appear to be
in the appli-
cation of the scoring criteria, not in the criteria per se ), so that
now raters
are being especially careful to avoid making liberal assumptions about or
interpretations of responses, since these were leading to unreliable dis-
*H.
M. Schroder, personal communication.
36
tinctions in the upper range of scores,,
The implications
more nearly normal
of the shift are unclear.
than the "old" one, i.e.
it is
center and has few scores at eithe'r extreme.
much
flatter,
crete score.
and
in
some samples,
et_
aL
but,
mode was
highest item scores
was required
to constitute
II in
is
Harvey
"Medium-Low
one for
who
integrate interpretations.
even less
whom
in
the
Integration Index,
Schroder
sum
"
of his
is,
et al.
two
very
or very slightly
termi-
Put differently, a complex per-
son, by Schroder's current definition, is one
is
of
It is
(1966) or Harvey, Hunt and Schroder (1961)
nology (ignoring theoretical differences).
tations to events, and
complexity
least 7 (3.5 on the 7-point scale); that
is at
slightly above Schroder's
defined
most con-
the lowest,
Thus, the relationship between complex Ss as defined
by present standards, a complex S
above Level
The old distribution was
(1967) and as they are used today is not certain.
not clear what score
is
sharply peaked in the
The new scores are obviously not merely a scaling-down
the old distribution.
by Schroder
the
The ’’new" distribution
who sees
alternative interpre-
additionally shows the tiniest wisp of ability to
It
common
seems
likely that,
if
anything, complexity so
than complexity as defined by Schroder et
al.
(1967).
Obviously there
is
nothing wrong with being interested in
uncommon
characteristics, but such an interest does introduce certain problems.
the present case, the theory
occurring groups
(a
becomes
In
virtually inapplicable to naturally
problem exacerbated by the restricted variance
of the
'
37
scale and by Schroder’s habit of using extreme samples and ignoring the
middle group).
portant.
If
Given
done.
Further, the reliability of the test becomes extremely im-
the test misidentifies subjects,
this fact,
considerable mischief can be
especially odd that Schroder does not report
it is
test-retest reliabilities.
Insofar as validity
portant distinction
is
among
concerned, the test does appear to make an im-
Ss regarding their information-processing and
decision-making capacities.
Schroder's typical procedure
is to test a
large enough pool of Ss to obtain a given number of "complex" persons.
"Simple" Ss are then chosen who match the complex Ss on important control
variables (intelligence, in particular, but sometimes other variables also).
Groups
of Ss so defined differ in the
utilize in decision-making,
and
amount
of information they are able to
in the quality of the
decisions made, the
complex groups being more coordinated, interrelated and self-directed.
The tasks
to
which the groups apply themselves vary, but are character-
ized as being long
term (usually several hours
overall),
relatively realistic
and interactive such that what a group does affects what happens later.
There
is
an interesting anomaly here.
measures interpersonal complexity, and he
Schroder feels that his
PCT
exhibits the usual caution about
generalizing to other domains.
Yet the games or tasks which the groups
are given are not interpersonal
in
*1
me.
am
nature, thus there would not appear to be
indebted to Lawrence Strieker for pointing this problem out to
.
38
any basis for predicting differences between the performances of interpersonally simple and complex groups
arguing that the content of the game
cisions
must be made
and the
PCT
subjects.
in a group,
is
this objection
by
not so important as the fact that de-
so interpersonal dynamics are important,
again becomes relevant.
In this case,
Schroder obviates
»
One study has been run using single
complex Ss were
still
superior to simple, but here
the task had to do with learning about a culture and solving a
problem based
on one's knowledge about and understanding of that culture.
In this case,
Schroder argues that the task
is
analogous to interpersonal situations,
that similar skills are required for success in both situations.
in
So far as
I
know, individuals have never been run on the noninterpersonal games.
In
summary,
tive complexity.
It
the
PCT
is a
sentence completion measure of integra-
has satisfactory interrater reliabilities, and no obvious
serious problems in the rating process.
Test-retest reliability
established; however, extreme groups, matched for intelligence
late of
PCT
(a
corre-
scores) do differ systematically in predicted ways on informa-
tion-processing and decision-making tasks.
that
not
is
One problem with
the test is
complex persons are rare, which emphasizes the importance
of the
missing test-retest correlation, and which limits the applicability of the
measure
39
Harvey's "This
The "This
I
I
Believe" Test (Harvey,
Believe" Test (TIB)
is
1966)
similar in format to the
are presented with a series of sentences reading, "This
.
.
.
life,
,
"
the blank being filled in by such referents as the
marriage, religion, friendship,
etc,
I
PCT.
Ss
believe about
American way
of
Items are
(see the Appendix)*
timed, and Ss are requested to write at least three sentences.
Generally,
about 10 items are used*
Trained raters assign scores
items are not scored.
score.
An attempt
levels.
is
to the protocols.
Rather, the entire test
made
to place
is
given one, overall global,
each S at one or another
Some protocols are mixed, however, and
one level
•
is
indicated.
However, individual
In these cases,
two
(or
of the four
functioning at
more
even three) scores
than
may
be
assigned.
Harvey
states (1966, p. 47) that the "inter-judge reliability for three
and four trained judges
in classifying subjects into
systems has been .90 or above for
take this to
mean
that,
12 different
for the Ss for
whom
one of the four main
samples
of subjects."
I
each judge gives a single as-
signment, interrater reliabilities equal or exceed .90, and that Ss given
more than one
that,
rating do not appear in these correlations.
out of 1400 TIBT’s,
7% System
No
4,
30% were System
1,
15% System
Harvey reports
2,
20% System
and the rest (28%) were mixed.
test-retest reliability
is
reported, so far as
I
have been able
to
3,
.
40
discover
Scoring criteria can best be described as messy.
They are complex,
rather content-oriented (theory to the contrary notwithstanding), and multi-
dimensional
in nature.
For example, Harvey describes some
of the
more
important criteria as,
the absolutism of his expressed beliefs, consideration of contingencies or modifying circumstances, dependency on external authorities, especially God and/or religion, frequency of trite and normative
statements, degree of ethnocentrism, acceptance of socially approved
modes of behavior, concern with interpersonal relationships, and the
apparent simplicity-complexity of the interpretations of the world
.
o
.
(Harvey,
More
and closed
1966, p. 47).
System
specifically (Harvey, undated),
in their thoughts
ventional, to
show
and beliefs,
requirements, etc.
Ss tend to be absolute
to be highly evaluative
a strong positive attitude
to be highly ethnocentric,
1
and con-
toward traditional authorities,
and to identify closely with social roles, status
These Ss appear
to
have internal standards, but these
derive from reliance on external authority.
This orientation
may
be char-
acterized as pro-authoritarian.
System
ity
2 Ss
are anti -authoritarian.
They
dislike and distrust author-
and automatically rebel against norms and social pressures.
as rigid, simple and consistent as System
tion.
They add
analytic,
hostility,
1
Ss,
They are
but in the opposite direc-
rejection and negativity to the concrete, non-
categorical thinking of System
1.
Harvey also
states that, while
these Ss reject conventional standards, they are simultaneously without
other stable guidelines.
They lack personally derived standards, and
41
Harvey does not mention
norms
„
System
3 S_s
appear
own resources
to be high in
They tend
oriented and friendly.
their
subgroup
(but implicitly rejects) the possibility of
to be
need
dependent on others rather than on
in the solution of their
They
The hypothesized increase
in
feel that
They are
problems.
quiescent in a socially accommodative sort of way.
be flexible and superficial.
They are socially
affiliation.
quite ac«
Their opinions tend
everyone needs
complexity over Systems
1
lots of friends.
and
2 is
based on
the assumption that these Ss have learned more, and
more
techniques for dealing with other people (System
after all, are
pulative --neces sary
goal-oriented and
Unless
Ss.
it is
if
more
"inner -directed,
more
no familiar label exists for System 4
"
to be independent,
They
flexible,
like novelty,
tolerate ambiguity and/or dissonance.
say,
mani =
flexible.
nonjudgmental, and relative.
is to
sophisticated,
they are to be dependent), and are therefore
They are supposed
which
3's,
to
and are relatively able
They are "information-oriented,
rational, analytical, and empirical.
and juxtapose diverse elements
in their
open, nonevaluative,
They tend
to
to
"
present
responses, and to evaluate these
noncategorically and relativistically.
These criteria are, categorically,
difficult to apply,
for the
reason
that the criteria within a
system are not necessarily highly correlated.
This means that
uncommon
to get
it is
not
to get
some indicators
but not others,
an indicator that does not distinguish between two systems, or worse,
.
42
from more than one system,
to get indicators
ent sees no value in (condemns) the
For example,
American Way
if
a respond-
of Life (Level 2) but
simultaneously argues that people should concentrate on learning how
communicate better with each other
value system
is
A number
(not
System
2,
to
because an alternative
proposed), should the score be 2 or 4?
of other rules exists for scoring,
having to
do, for
example,
with the overall tone (pessimistic “optimistic) of the response, the attitude
toward friends,
this rater at least,
and
2's
tegy,
if
The
etc.
in
effect of these rules is to lead to difficulty,
discriminating between
and 4's on the other.
one
is
If
If
Harvey
is
in
my
institutions constitutes
who
is
is
is
assignations jump
is this
tendency, then
considerable dependence on con-
strong positive attitude toward religion and/or other
System
1,
negative attitudes System
feels that a lot of friends are important is a
unreliable he
If
systems seems questionable.
should be noted that there
A
to give both
moderate negative correlations be-
experience there
the unidimensionality of Harvey's four
tent in the scoring.
is
talking about a dimension, confusion should oc-
about as described above, and
it
on the one hand,
On such ambiguous protocols
cur between adjacent categories, not alternate ones.
A1 so,
3's,
attempting to establish rater reliability,
not particularly difficult to obtain
tween raters.
and
the decision is at all close, the best stra-
scores and thus remove the S from the pool.
it is
I's
for
System
2.
System
3,
2.
but
Examples could be easily added.
Someone
if
friendship
In practice,
only with System 4 criteria do structural considerations (how something
is
43
said rather than what
is
that for the first three
said) begin to outweigh content.
would suggest
I
systems the TIB primarily measures content rather
than structure, and, in fact, measures structure only insofar as
it is
re-
lated to content.
Of course, Harvey does see a relationship between content and structure, and the relationship is
based on theory.
In the
process he reveals
his theoretical conservatism, because the derivation of the relationship can
be found (perhaps by implication for some specifics) in Harvey, Hunt and
The original theory specifies a universal sequence
Schroder (1961).
least in this culture) in which interpersonal experience
integrated, including in a general
sion at each stage.
way
is
(at
differentiated and
the nature of the developing
dimen-
This, in turn, implies something about the content of
attitudes and behavior, in the sense of the focus of interest and the general
style of
approach
to the topic.
for his scoring criteria.
It is
on this specification that Harvey draws
The circularity
of this
process requires that the
validating evidence be particularly sound.
Parenthetically,
sults as correlations.
at various levels.
of his variable.
it
should be noted that Harvey does not report re-
Like Schroder, he picks groups of Ss who operate
Unlike Schroder, his Ss are chosen from the
full
range
Instead of "simple" and "complex" groups, Harvey usually
uses equal numbers of Ss from each of the four systems (levels).
These
groups are then compared either using _t_ tests or analyses of variance.
(Of course, the latter would be preferred, but
Harvey does
not mention
44
which
is
used.)
Typically, nonlinear results are found; e.g. Systems 2 and
4 are higher or lower than
Harvey's rejection
1
and
3.
The
effect of this is,
first,
to
support
group comparisons (which
of correlation in favor of
should be by analysis of variance); and second, to again raise the question
of
whether the four systems should be viewed as
falling along a single
con-
tinuum.
Harvey
(1966)
date the TIB.
Most
summarizes
a great deal of evidence intended to vali-
of these reports fall into one of
two groups.
The
first
group constitute trivial findings, theoretically speaking.
At best these
could be regarded as validating the scoring criteria, and
it
more appropriate
to
seems much
regard these as hetero-method reliabilities.
These
relationships should obtain because they are built into the rating criteria.
I
include here the findings on authoritarianism, dogmatism, need for affi-
liation,
most
of the religion material,
"identification with the
American
EPPS,
motif.
"
If
Scott's Scale of Values, and
these had not proved agreeable,
then the application of the rating criteria would be under question.
There
a small subgroup
is
validity nature.
First, the
which does seem
to
be of a construct
WAIS has been administered, and
differences
between the groups are found only on Verbal Intelligence and Vocabulary,
Systems 4 and
2
being significantly higher, although Harvey claims that the
systems do not vary
in
vocabulary usage on the TIB.
sirable that the TIB and
PCT
Of course,
it is
de-
be independent of intelligence; however, this
finding is at variance with Schroder on two counts.
First, Schroder re-
45
ports that intelligence typically correlates with integrative complexity at
some
level between
components
tive
5,
.
a linear, as
is
of intelligence the one
vocabulary, because
processing
A
and
2
opposed
to
Harvey's nonlinear,
Second, Schroder has argued (Schroder et'al.,
finding,,
plexity
.
at all is
1967) that of the
which should be least related
it is
based largely on
to
memory and
com-
no cogni-
required for high scores.
second validating relationship
between cognitive complexity as
is
I
I
measured by "a modification" of Kelly’s
REP
test.
In this case, a linear
relationship was found between complexity and the four systems, System 4
being most complex.
that Harvey's is
Note that this substantiates the position taken earlier
now essentially
any argument he might make
beyond dimensionality.
It
to the effect that his test
sure
is
is
weakens
measures something
very nearly the only variable
validity relationship is with rigidity.
to do so.
This
based on the Gough and Sanford Scale on Rigidity, with which
not acquainted, and the reference for which
I
it
also serves the valuable function of ordering the
four systems sequentially, and
The third construct
a dimensional theory, and that
is
unpublished.
meaI
am
Consequently,
can only report that this scale also ordered the systems sequentially,
System
1
being the most rigid.
The second major group
of validating studies is
Here anticipated behavioral correlates
of
more impressive.
complexity were examined.
The
I
reports are exceedingly brief and cannot really be adequately evaluated.
Tasks intending
to
measure
a variety of learning or concept formation or
46
cue utilization skills, ability to change
creativity (novelty plus appro-
set,
priateness), admission and tolerance of deviant inputs, role playing, etc.,
were involved.
and
In general,
varying back and forth
3
Systems 4 and
in the
studies are not independent, the
or the
same
mention
is
1
middle.
same
behave appropriately, with
There are some hints
been included.
of controls for intelligence,
If
that the
Ss being used on various tests, and/
test-data being analyzed in several different ways.
made
2
a fuller report of the
Also, no
which definitely should have
work met these
objections,
how-
support would accrue for the theoretical position.
ever,
To summarize,
the
TIB relies on scoring criteria which are not
fully
satisfactory, partly because they are too heavily dependent on content in-
dicators for one to be convinced that the test measures structural properties,
and partly because the criteria defining a particular system do not
seem
to constitute
an especially cohesive characterization.
The
latter
leads to problems in assigning Ss to systems, and causes a fairly high disqualification rate (between one -fourth and one -third of the Ss).
evidence
is spotty.
from
some cases
positive results are built into the scor-
Other predictive validity findings are more impressive, but
ing criteria.
suffer
In
the brevity of description and
from
the implied possibility of
lack of independence and control for intelligence.
apparently
is
related
to,
Beyond
more
this,
the
but not correlated with, intelligence, and
related with Bieri's type of cognitive complexity (Bieri,
variables,
Validational
1966).
TIB
is
cor-
On other
often than not, there is a reversal in the sequence such
47
that Level Z is
more similar
and Level
to 4,
scores being sequentially ordered.
Harvey
3 to
Level
1,
rather than
superior to Schroder in that
is
his scores are
more evenly spread along
his dimension, and that he con-
siders the
dimension
rather than the extremes.
full
in his re-search
A Comparison
At a theoretical
represent Level
I
of the
Two
Harvey's
level, both of
Positions
first
two systems seem to
For example,
functioning for Schroder.
sponse might be "Law and order must be maintained
Each
a System Z S might state, "Cops stink.
"
perspective, and so would be scored
by Schroder.
a
System
at all costs, "
of these
1
re-
whereas
represents a single
I
sounds a good
bit like
1
Schroder's score of
characteristic for both
is
3
(Level
Harvey's System 4
II),
in that a critical
the ability to see alternative points of view (al-
though Harvey seems to emphasize more the development of internal standards).
It is
interesting to note that the incidence of Harvey's System 4
(7%) is roughly equivalent to that of Schroder's
where "complex"
a score of
3
is
defined as very slightly above a score of
represents Level
II).
Harvey's System
spondence with Schroder's scoring system, but
scored higher than
If
complex Ss (4%
Z,
i.e.
it
would
still
it
3
3,
to 1Z%,
and where
has no obvious corre-
would not be likely
to be
be simple in Schroder's thinking.
reasonable, this comparison suggests that the two
men
are dealing
48
with different portions of the simple -complex continuum, Harvey's dimension being a portion of Schroder's.
making
out his segment, i.e.
Schroder
Any such comparison, however, must be
o
is
finer discriminations'within
the scoring criteria for the two
A
Harvey, presumably,
it,
then spreading
than does
in spite of the fact that
measures are radically
different.
i
second important difference between the theorists seems
have
to
to
do with their respective visions of the purpose or value of their variables,
or with the
way
Harvey wants
in
which they manifest themselves
to identify a characteristic
in behavior.
I
think
with broad implications, one which
affects a good bit of the range of daily behavior, a variable which,
when
a
i
person's standing on
his typical style.
it is
known, will convey considerable knowledge as
Schroder, on the other hand,
information processing and decision making
kind of complexity with which he
is
primarily attending
complex
in
concerned
is
is
to
to
The
situations.
not engaged in typical be-
1
havior, and he goes to considerable effort to devise tasks which are
ing enough that complexity as he views
The
last
major difference we
where differences are
distinctions
it
will begin to play a role.
will consider is that of scoring criteria,
so striking as to scarcely need reiteration.
among Harvey's
demand-
first three
systems are based on content,
whereas Schroder can legitimately claim
This criterion issue reflects on theory,
In brief,
to be
in that
oriented toward structure.
it
results
from Harvey's
reliance on a theory which essentially specifies content, while Schroder's
restricts itself to structural considerations.
This, in turn,
means
that
49
Schroder's position
aesthetically
is
more
is
more
easily adapted to
new circumstances and
attractive in adaption than Harvey's,
since the latter's
so intimately associated with the interpersonal domain that one
either refer back to the interpersonal content
who
(in
the
from
affiliative,
that is Level
3,
it (in
must
domain "History,
relies on the writings of famous historians for his opinions
or attempt to analogize
is
is
"
one
System
1)
"History, " what corresponds to need
behavior?
) 0
Apparently, then, Schroder and Harvey vary in their definition of
complexity and the dimension of complexity; they vary
in their
research
interests and thus in their choice of dependent variables; and they vary in
both the definition of scoring criteria for their tests and in the nature of the
theories underlying their positions
.
The Present Study
The purposes
A.
of the
present study are as follows;
To compare TIB and PCT scores across a broad range
The implication
of the
discussion above
is that the
of
S_s
.
two sets of scores should
be nearly independent.
B.
To compare both scores with measures
of intelligence.
Harvey
reports that TIB scores are related to Vocabulary and to Verbal IQ.
Schroder states that the
PCT
should not be correlated with vocabulary, but
50
should be correlated with components of intelligence involving complex
A
problem solving-
test of the former, and possibly of the latter,
is
pos-
sible with the data at hand.
C.
To compare males and females on complexity.
discusses sex differences, and, so far as
looked for such differences-
I
Neither theory
know, neither researcher has
Gardiner (1968) found girls
to be
more com"
plex than boys, which he explained in terms of child-rearing practices.
D.
T o compare samples drawn from different populations.
Schroder's work
is
The present study
Most
of
done at Princeton, and much of Harvey's at Colorado.
will
sample from populations rather different from those
typically used by Schroder and Harvey.
Chapter IV
METHOD
Subjects
Approximately 75 female and 75 male subjects were sought from
each of three schools.
Negro college
Johnson C. Smith University (Smith)
of about 1200 students
drawn largely from
scores typically average around 350.
and 76 men.
branch
The University
of the state university
of
Data were collected from 72
North Carolina
at Charlotte
UNC-C sample
tested during the spring semester and also during the
SAT
from Catawba College nearby, thought
scores.
mer
UNC-C
women
(UNC-C)
is a
Due
to unantici-
consists of subjects
summer
to be
Information was not available, but
expected to be around 500 for both schools.
lected during the
SAT
system, nominally integrated but preponder-
pated difficulties in collecting data, the
garding
private
the South.
antly white, currently operated on a nonresidential basis.
plus subjects
is a
session,
comparable re-
SAT scores were
Data for 38 females were col-
spring semester, for 30 females during the
session, and for 4 further females during Catawba's
(
51 )
summer
sum-
session,
"
'•X
52
a total of 72.
UNC-C's spring semester yielded
sion 25 more, and Catawba's
summer
17
summer
males, the
session the final 27
(total 69).
ses-
Stu-
dents at both institutions were mostly southern, but the proportion from
northern states appeared much higher for Catawba than for UNC-C.
The third school, Duke University,
nantly white.
Duke
is
is
also integrated, but predomi-
private, with about 4000 undergraduates.
draws
It
as heavily on the North for students as on the South, and entrance requirev.
ments, as largely defined by SAT scores, are high. Freshman averages
being generally around 625.
Seventy-six
women and
75
men from Duke
participated.
Subjects at Smith and
services.
ward
Ss at Duke
UNC-C
were given
(and Catawba)
were paid $1.00
for their
the choice of $1.00 or one hour credit to-
the six required in their Introductory Psychology course.
Virtually
everyone chose the hour credit.
Measures and Procedure
Three tests were administered
in the
experimental session, the
Shipley-Hartford Test of Vocabulary and Abstract Thinking (SH), Schroder's
Paragraph Completion Test (PCT), and Harvey's "This
(TIB).
task.
The PCT and TIB were combined so as
Additionally, Ss signed a
to
I
Believe" Test
appear to be a single
form giving E permission
to obtain
SAT
53
scores from the school's records, and gave such information as age, class
and home address.
in college,
Ss,
especially those in
summer
SAT scores was
portion of
SAT's proved
school.
to
be unavailable for
Not surprisingly, the highest pro-
Duke samples.
for the
Number
Actual testing occupied something under an hour.
session varied between
3
and about 35.
these conversations were prolonged.
first,
5
All tests
were timed.
The SH was
minutes being given for the 40 multiple -choice vo-
former are standard for the type test involved.
of ability to abstract consist of a series of
ranged according
to
raw
"
hip
,
in the
sequence.
in the
Appendix.
sequence
in
The respective sets
is
the
PCT
(the first of
is
item "tarn tan
is hit.
a
rib rid
)
appeared together, and the
One item appeared on
In the present case, the
There were
which
deduce the principle
Instructions and items appear
toward the upper end
gested by Schroder and Harvey.
Items for the test
response
which they were offered was alternated.
was 130 seconds, which
from
is to
(In the
of items
each page, and each item was timed.
6
S
for example, the correct
The PCT-TIB combination followed.
Instructions
numbers, letters or words ar-
an organizing principle.
and complete the last item
rat
per
some cases
in
cabulary items, then 10 minutes for the 20 abstraction items.
for the
of Ss
At the conclusion of the testing,
an explanation was offered those who were interested, and
administered
many
of the
time limit
range of limits sug-
16 items; 10
from
the TIB,
warm-up, nonscored item).
The
-
54
PCT
items are fairly well standardized.
lection of
More
variation exists in the se-
TIB items.
Scoring
The SH
scored against a key and produces four measures:
is
ulary (SH~V), abstraction (SH-A), total (SH-T, the
and the "Conceptual Quotient" (SH-CQ) which
SH-A (CQ
=
(SH-A/SH-V) X
that the test be
100).
The
CQ
is
sum
vocab-
SH-V and SH-A),
of
SH-V and
the ratio between
reflects the original intention
used as a rough screening device for detecting the type
of
cognitive impairment likely to be experienced by mental patients, based on
the
assumption that vocabulary, a function
of
memory,
remain rela-
will
tively stable, while the active skill involved in abstracting will be an early
loss.
Thus, a low
CQ
results
from
a relatively high
indicate a loss in abstract functioning.
lated to analytic thinking,
it is
To
SH-V score and may
the extent that
of interest in the
CQ may
be re-
present case because of
the possibility, based on Gardiner and Vannoy, of a relationship between
this kind of thinking
and complexity.
The PCT's were scored by Schroder’s research assistants
laboratory
in
Princeton.
liabilities for his raters
ed a score from
1
to 7 as
Schroder reports (Schroder
range between .80 and .95.
described previously.
et al
.
,
at his
1967) that re-
Each item was award-
The score used
in the
55
analysis was the
tween
and
2
14.
a score above
able, in
4,
sum
of the
two highest item scores, which could range be-
In practice,
out of the total 440 subjects, no item received
and no
score exceeded
final
which case they are given a zero.
the protocol, no final score is given.
7.
If
Some items are unscor-
two or more zeros occur
in
This occurred on approximately 6%
i
Two
of the protocols.
sets of
PCT
scores were used, the first derived as
i
just described.
scores.
The second
In this case,
may
set
subjects
be termed the Forced
PCT (PCT -F)
who had previously received zero scores
I
were given scores based on whatever number
these scores being, again, the
sum
Two previously inexperienced
follows:
of the
of items
were
available,
two highest item scores.
raters proceeded with the TIB's as
Basing procedures and criteria on mimeographed material sent
from Colorado by Harvey, approximately 90
used for training purposes.
of the protocols at
The number was so large because
exceedingly difficult to establish reliable criteria.
were then rerated
at a later time.
rated a sample of 107 protocols.
hand were
it
proved
The protocols involved
After training, the judges independently
Of this number, one or the other or both
judges gave zeros to 30 Ss, a 28% unscorable rate, comparable to Harvey's.
A
contingency table approximation of chi-square for agreeing or not agree-
ing to score or not score protocols produced a value of 15.95,
at well
beyond the .001
the judges as to
tion
level,
significant
suggesting that there was agreement between
which subjects were and were not scorable.
between the judges on the remaining 77
S_s
was
.83.
The correla-
Since this was
56
regarded as an adequate indication
remaining protocols, and her ratings were used
Throughout
one judge rated
of reliability,
all the
in the analysis.
process there was one major and intentional devia-
this
t
tion
from Harvey's recommendations.
He advises
that a single global
score be given the subject instead of individual item scores.
Since item
scores do not preclude global scores, and since they would seem to permit
greater flexibility in scoring as well as tighter control over biases, item
scores were assigned throughout.
Global scores, then, were based on the
frequency of the item scores, ameliorated
seldom as possible, by the judgment
if it
seemed appropriate, but
of the rater.
as
This was the Global
(TIB-G) TIB score.
Three further scores were baaed on the TIB.
test are rather heterogeneous in nature.
seems
of
ly,
to feel that he is
some items
5
of course,
some
(I)
matters
is
According-
not always obvious.
in the
Appendix) were considered separately, and
score given on the basis of the item scores.
of these
a zero value
was
determine a single rating
though distasteful), but
As with
G,
were unscorable.
The two remaining scores are forced versions
wherever
Harvey
items which seemed most immediately interpersonal
(marked with an asterisk
an Interpersonal
In spite of the fact that
measuring interpersonal complexity, the relevancy
to interpersonal
the scores on the
First, the items in the
given, the judge
in addition.
more
of
G and
was required
In general, this
difficult for the
I
than the
to
I.
That
is,
attempt to
proved possible
G scores.
(al-
57
These scores, then, plus SAT scores from the college records
fices for those students for
whom
they were on
file,
and the information con-
cerning age, college class, and home (residence) address, were used
analyses
to follow,,
of-
in the
j
Chapter V
RESULTS
The
fact that n's are typically
uneven makes especially relevant the
Although the F max test for homo-
question of homogeneity of variance.
geneity
intended for equal n's, Winer (1962) states that
is
applied to unequal n’s
if
F max may
be
one uses the n of the largest sample in the analyi
sis.
of
In the case of
thumb
(Box,
nonhomogeneous variances plus unequal
1954a,
1954b)
seems
to be that,
if
n's,
the rule
the larger variances be-
long to the larger sample, the probability levels for the F-ratios will be
conservative, whereas
if
the larger variances are based on the smaller
samples, the probabilities will be overstated.
volved
is
The size
not so important as the ratios of their sizes.
appear
ferences
sample variances exist along with unequal
in
become rather complex and "true"
by.
Where important
it
Thus, approximate-
n's,
some
will be noted.
of the
dif-
relationships
probabilities extremely difficult to
Unfortunately, this situation exists in
ses below.
samples in-
However, when large
some robustness.
ly equal n's
to retain
of the
come
preliminary analy-
The two-way analyses
of
variance have been computed using the least squares procedure for unequal
(58)
.
59
n's outlined
by Winer (1952,
p.
291
ff
.
)
where one factor has only two
levels
Means and standard
Table
deviations for the variables are presented in
Protocols were 94% scorable for the PCT, 85% for the TIB-G,
1„
and 81% for the TIB-I.
(The "unscorable" rate for the TIB
is
lower here
than Harvey reports, and lower than in the preliminary reliability check
This
here.
is
because scores for only one rater are used.
"Unscorable"
decisions, although significantly related, do not overlap perfectly.
the
more
raters, the
scores from
the
2
more
through
7
TIB-G, percentages
the TIB-1,
31,
12,
31,
rejects.)
For the PCT, the percentages
were, respectively,
for scores
and
1
Thus,
to 4
9,
39 s
28,
12,
and
36,
15,
26,
and
8,
3,
were
bers of a pair occur when an S received a zero score
and these are removed from the correlation.
rived from the TIB, however, four pairings
from
.
7
1
to .87,
2).
For
and for
their corre-
sponding nonforced values, since the only discrepancies between the
combined sample (Table
3.
7.
The various forced scales correlated perfectly with
were calculated separately
of
The range
nonforced case,
Since four scores are de-
may
for each of the six
*
in the
mem-
deviate from 1.0.
samples and also
for the
for these 28 correlations
with 22 of them equalling or exceeding .80.
These
was
In short,
^These and the other analyses herein reported were run at the Duke
University Computer Center, using programs written by the author, and
funds provided by the University.
8
3
5
:
2
60
Table
1
Means, Standard Deviations and N's
Shipley-Hartford
Smith Females
Smith Males
UNC-C Females
Age
Vocab.
Abs.
T otal
CQ
X
19. 5
28. 2
29. 2
s
57.4
7.42
99.0
1.
70
Duke Females
Combined
50
5. 11
72
70
70
70
X
19. 8
27.
29. 3
56. 8
s
n
1.90
76
3.47
76
6.00
76
76
X
22. 6
31.5
33. 6
65.
6.
79
3.
32
3. 57
7.
4. 79
72
72
X
31. 6
33.
64. 9
s
21.4
3. 84
n
69
69
69
69
X
18.7
34. 2
36. 3
70. 6
2.46
2.49
n
74
76
X
18. 7
34. 3
s
1.
n
75
X
20.
3.
s
18
2.
1
74
2.
75
75
31.3
33.
438
64
5.
438
3.
32
86
1
100.8
10.0
102.
69
1
7. 87
76
70.
102. 0
4. 10
75
22
10.7
72
69
76
36. 5
28
4. 12
438
n
2.
76
7.
1
102. 2
1
72
4. 77
13.
76
72
3.77
10. 6
100.9
98
n
s
Duke Males
3.
n
s
UNC-C Males
15
6.
88
j
I
t
75
I
101.
64. 3
8. 27
438
10.
1
438
N's vary because of missing data or unscorable protocols.
j
Abbreviations
SH-V
SH -A
SH-T
SH-CQ
SAT-V
SAT-M
PCT
PCT-F
TIB-G
TIB-GF
TIB -I
TIB-1F
Shipley -Hartford Vocabulary
M
"
"
"
Abstraction
Total (V + A)
"
"
Conceptual Quotient ((A
Scholastic Aptitude Test--Verbal
"
"
"
--Mathematical
V) x 100)
Paragraph Completion Test--Score
"
"
--Forced Score
"This
I
"
"
"
"
Believe" Test--Global Score
"
--Forced Global Score
"
"
--Interpersonal Score
"
"
--Forced Interpersonal Score
Table
1
(continued)
PCT
SAT
TIB
y
M
PCT -F
PCT
G-F
386
372
4. 15
4. 16
1.94
83
54
75
54
86
.98
374
86
48
400
474
477
4. 50
71
50
81
50
.89
468
509
97
80
49
49
614
75
621
74
75
615
74
663
73
71
71
505
130
347
524
134
347
88
83
48
.
85
72
.
68
4. 37
95
.96
o
76
1.
4. 52
.
88
67
4.41
19
4. 36
1.
20
66
69
.
88
76
68
72
1.86
1.88
1.01
69
4. 58
4. 63
2.40
1.06
76
1.05
1.
4. 88
1.
16
75
4.48
1. 04
440
76
71
4. 88
1.
10
15
72
4.49
1. 04
412
35
1.01
75
2.
05
1.03
439
2.
22
.99
98
.99
2.
1.73
85
.
1.75
87
64
.
78
1.
1.03
64
1.
1.
1.
97
1.02
59
70
08
1.03
64
2.
2.
1
2.
39
2.
1.
04
.
06
1.04
373
2.
50
1.09
7
66
96
06
71
2.40
1.11
I
29
.99
2.
56
60
59
1.86
1.05
72
I-F
1.
55
71
4. 34
G
39
92
70
2. 15
1.03
400
1.75
.89
55
1.
98
1.06
62
09
1.06
2.
53
52
1.11
66
2.
39
.93
2.
65
2. 18
1.
357
04
I
I
I
i I
(1
CD
U
O
O
CO
cd
0)
CD
?H
(D
o
u
a
0
U
0
u
o
o
?H
co
co
fn
— CO
—
CO
CD
i
ft
»
ctf
tj
CD
<D
!h
o
0
U
?H
o
rO
cd
U
o
fl
CD
O
-l->
fl
CO
CO
i—
f-i
r
d
CD
ft
Jh
o
—° o
0)
S-c
H->
pH
l—
4->
co
CD
_
“
CD
o
1
0
I
O
fl
Ph
I
!
t
!
I
I
|
|
4->
_
CO
-
0)
H
ft
CD
a
>
0
o
U
r—
•
r—
<d
ft
CP
cd
!h
bO r
co
cd
• r-i
fn
rS
H
cd
Ph
CO
fl
O
Ph
Ph
H H
CP CP CP
U O CP
— — i— —
Ph Ph H H H H
i
•fi
d
i
'>
CD
Jh
rO
rO
<
Ph
O O
r
i
i
i
i
i
t
I
H
i
[
i
i
ii
!
I
61
-x-
vO
un
in
l>
o
in
o
co
m
<b
co
r-H
m
o
o m
in
rH
lO
rH
—
i
CO
vO
1
o
i—
•
CD
I>
m
b>
00 vD
«
•
co
o
m
•
i—
nj
2
o
r4->
•
rH
xH
o m
g
i
r-
o
CO
in
o
o
in
—
i
i
m
CO
H
i
1
o
vO
in 00
r~-
vO
o
j
ctf
N's
CM
and
o
o
1
—
1
—
1
CO
sO
Intercorrelations
1
vO
v£>
o
1
xh
P"\
CM
m
cb vO
r^-
i_ps
in
co
O
o
vO
in
—
1
i
ji
Score
co
—o
in
,
o o
o
M'
1
O
vO
in
o o
CO vO
in
00
o
H
1
•
•
oj
1°
protocols.
a
<u
Complexity
ft
•
m
m
vO
p-H
.d
—
0
CM
r-H
CM
in
0
0
i
r-H
5
c
I
O m
o m
m
unscorable
!
CO
o
i-H
rH
r-
r-
o
r-
sD
of
CM
o
o
co
vO
because
rH
d
d
d
r
d
n
r
n
r
n
vary
ft
•GF
H
H
O
O
ft
#
r—
1
0
CQ
CQ
i—
i—
ft
H
Eh
0
•
.
(M
CO
M
1
IF
l-H
ft
HH
CQ
hh
H
m
•
H
o
vO
v
ft
O
i
(i
1
i
%
vO
O
—
i
i
CO
un
—
i
r-H
i
LO
CO
r- LO
CO
00
6la
O
o
o
CO
in
r-H
w
0
iTi
CO
—
1
r-H
1
sO
o
o
r-H
P-i
sO
0- vO
0
LO 00
00 IT)
#
*
nj
s
-
0
o
1
00
O'-
i-h
in
CO t"
r-H
LO
£
o o
o LD
t—
tD
CO
co
00
o
o
vO
o
o
sO
vO
O
O
V.
vO
r-H
(continued)
vO
00
00
r-H
r-H
vO
ro
rLD
iD 00
00 'O
I
00
00
in
o
o
oo
no
r-H
2
Table
Cfi
0
0
r-H
jj
nJ
d
R
0
in
CO
—
1
1
}),
LO vO
O'
sO
r-
r-H
r-H
o
•
i-H
r- 'O
•
•
h
O
U
£
1
ro
—
'O
LO
CO
r-H
in
o
0
o
o O
0
r-H
ID
CO
O0
i—
oo
r-
O
o
vO
vO 0r-H vO
r-
r-H
Jh
R
u
R
u
R
v
R
Vi
R
r
0
0
•IF
hH
eq
cq
CQ
CQ
i—
1—
i
Ph
H
H
00
CO
U-t
i
H
U
&
H
O
r-H
—
H
m
n
H
vD
n
{
i
I
i
1
l
i
I
OO
in
n
o OO
c-
in
c- co
r—H
vO
r—
CO
o
[-
c-
r-
o o
i—1
sO
00
C-
OO CO
OO
o
O
oo
vD
r-H
t"
61b
r~
oo
oo
in
r00
o
o
O
OO
o
o
oo
oo
u
b
in
oo
CO
o
r~l
ni
c- oo
s
X
b
Q
O
OO
r-1
LT)
O
<u
CO
CO
o
o
o
r-
O
o
CO
r-
co
r-
(continued)
OO
OO
00
2
co
O
>—
co vo
OO
oo oo
oo
oo
o
--i
oo
r-
CO i—
00 OO
Table
CO
<u
r—
in
O
\n
»—
r-l
C'-
o
O'
O
in in
r-
o
oo
CO oo
r-l [>
o
—
—i
u
b
b
a
CO O0
co -n
<D
P
o
<D
M
b
Q
co
co
O
o
r-
u
b
i
i
o o
o r-
O
u
b
f-<
Pc
0
6-i
H
U
P
H
0
b
u
Pi
i—i
i-i
CQ
CQ
CQ
CQ
|
i
t
O
P
t—
1—1
i—
r-l
H
H
H
H
co
CO
in
oo
i1
i
i
!
6 lc
(continued)
]i
sO
2
co
I
s-
o m
CO
O
!—
CM
f—
00
o
CO
CO
00
CO
1
I
s-
uo
CO
in oa
oo CM
CO
o
er
e'-
1
LCD
.
CO
u
d
Table
in
00
o
o o
0
O
o o
o
s-
00
3
CO
•
O'
CO-
o
o
I
CO
00 in
•
CO
Ti
CD
00 CO
o
d
0
s-
I
co
CD M.
CO
1
—
CO
e'-
en
1
d
d
O
U
CO
in
o
O'
CO
°
<M
o
o
vO
i—l
0
^
r
n
O
r-l
CM
r-H
r—~i
u
h
H
d
0
i
O
&
L
u
H
O
1
PQ
i—
Ph
H
cm
CO
d
u
0
1
CQ
HH
d
fn
Ph
i—i
i
m
CQ
—
>
H
H
un
vO
d
62
they are highly comparable.
in
nearly
all
Parallel analyses were run on all six scores
cases, but with rare exceptions, parallel results were found
for scores deriving
from the same
test.
Accordingly, unless the others
are of special interest, only the results for the
(i„e. the
All
t
PCT
and TIB-G scores
scores as used by their respective creators) will be considered.
tests and correlations are two-tailed.
Preliminary Tests
The
initial
question asked of the data was whether the sequence in
which the Ss took the
PCT
and TIB affected their scores,
according to which test they had taken
first,
on each of the complexity scores by Jt^ tests.
significant,
The sequence
in
which the
PCT
Ss
were divided
and the two groups compared
These were uniformly nonand TIB were taken did not
affect the scores (see Table 3),
The second issue was whether the UNC-C sample, drawn as
from several sources, was homogeneous enough
sample.
to
17,
25,
one-way analyses
27),
was
be regarded as a single
Since the males were relatively evenly distributed
three sources (n's were
it
of
among
the
variance were
calculated on the complexity scores and on each of the six intelligence
scores (Table
4).
These tests were
stacle to regarding the
all
UNC-C males
nonsignificant,
so there
is
no ob-
as representing a single population.
•
H
'
1
1I
0
d
O
o
CO
0
d
0
1)
<
d
o
o
co
co
r—H
O
CO
cd
X)
0
d
O
O
f——
1
0 £) d
d o d
O — o
£ o
CO
o co o d
°
r
X) 0
CO Id
d 0o Pd
d
,o
+->
0
Q d
co 4
0
r—
d
0 “
Id hH
0
d
°
0
^
t
1
i—
1
I
I
1
)
->
I
0
H
d
o
•
i
1
1
1
d
d
0
0
d
0
Pd
d
0
d
1—
+->
X)
0
o
d
O
fd
1
1
+->
co
p~
-
-
—
0
a,
0
S
o
>
0
O
——
• I
.
0
rd
Ph
PQ
d
d
DC
co
d
d
d
•
r~t
X
^
“
H
Ph
CC
(X|
.
g
,
H
O O H Id
%5 r
n
U\r,
U pq
>
0
)
d
0
r
<
Ph
Ph
hh
h PQ
H H H H
pq
HH
ii
—
HH
i
h
H
i
I
63
ffl
h—t
H
o
CO
o
—
LO
LO
CO
1
1
vO
e
o CO CO LO
oo o o
CO
r-H
i—
o
1
CO
o
x
„
r-
o
i-H
l
o
CO
O
CO
_0
0
o
r-
o
r—
oo
r—-i
LO
-<
CO
•
o
<
X
i—
t
r-
co
o
i—H
i-H
*
n
6
-i
t—
i
ffl
CO
o
°
t—
co
co Cl
00
cr-
O
o
O
i
W
—
H
co
co
.
i
o io
O O —
co
X
o-
<o
H
w
0
On
—
LO
o
oco
O C
X
X
r-H
co
'
6
•
C-
r-l
O
co
„
.
00
0-
h CO
—
o
1
CO
r-
O
o
a
CO
CO CO
-
(O CO
co
r-H
O
LO LO CO
-
o-1
00
oo
CO
co
o x
o
co c—
co
o
o
t
co
CO
I—H
CO
•»
CO
H->
•
rd
Ui
0
a m
H
H
o
O
X
0
co
r-
n
co
l
.
r<
c
n
+->
co
CO
0
r-H
X
d
H
H
O
0
C6
rH
h
H
co
CO
X
o
CO
NO
oo o o —
O
O O' N O
X
o
d
o
X
vD CO
X
o
vO O'
X
r-H
O
r—
lD
CO
x
d
CO
CO
o
o o
vO
r—
CO
LD
00
CO
1.0
CO
LO CO
'—I
CO
r-
o
CO
o
H
d
d
U
0
0
oo
CO
oo
X
O O LO
X o o
CO
•
C
CO
Ph
'
•
i-h
oo
X —i o X oo o oo
o x O h N ^
oo eo
I-H
•
O
c
1
1
CO
1
in
Ph
w
•
+j
<-W
o
m
X
•
1
0
o
iH
^
r
C^J
64
u
CO
ro
CO
CD
cd
t-H
O
CO
CO
id,
ro
X
0
CO
X
<dH
U
TJ
d
t-H
U
X
0
CO
X
^H
U
T5
„
CO
X
0
X
Hh
CO
A
"d
X
<
>
C/J
k-t
0
A
0
O
n
I
co
A0
^2
X
Sz
lD
4->
•rH
X
0
A
a
o
O
w
o
CO
g
^
d
_
^
^
0
o
a
0
d
-u
0)
CO
o
<D
0)
rg
mh
w
5
-x
d
0
T,
rH
Oj
4-3
'dH
CO
d
0
X
4-3
0
oo
o
>n
m
o
(
1)
4-3
_
d
0
u
!h
0
0
+->
•ri
i—H
0
o
xi
CO
VH
CO
5
d
<
>
O
£
<!
r-H
1
1
CO
X
X
>H
qp
CQ
or
d
O ^
CO
CO
d
63
ra
pill
4—3
•
rH
a
CO
s>
d
+j
z
w
6
>
6
s>
?
o
d
<
>
0
B
1-0
<
u
d
0 £
OjO
0 o
r-H
r-H
£
O +->
Q
O
>•
rn
Vi
A
(C-I
1
6
o
X
X <
0 >>
o d
d
0
X
0 rH 0
Ard 0 +j£
•
0
>
rH
£
LO
O
Vi
t
•
Q A
W
A
iC
i
1
o
o
1
0
p
o
o
in
>
-
!
0
P 0
o p
o
o o
in
0
p
o
u
-
<
ft
P
0
• rft
4->
o
P
< a
i
+ p
p
—P
P
>
C-
P
0
ft
ft
U
t>
^
&
P
O
O
5
o
g
o
ft
O
o
> <
in
ft
i
i
P
i
i
^
+->
ft
I
!
I
I
0
in
0
p
o
o
tj
s
O
CO
0
4->
CO
ft
0 0
p o
° o
r
ft
r
p
o
ft
CO
p
0
in
in
+->
P
P
o
p
ft.
p
0
P
o P
o
ft
o
co
ft
fH
p
P —
0 ft
« ft 0
o
H p
0 p
a 4-5 O
*
ft
I
1
6
I
S
I
4->
CO
0
0
ft
ft
'
P
ft
P
0
4->
m
4->
+->
<
ft
1
4->
Cti
ft
ft
c
P
p
ft
in
O
in
ft
*•
> < H O
1
1
1
i
ft
r-H
0
ft
'
ft
<
<
in in
w
_
-
T—l
•
ft -
H
ft
a > £
'
r-i
•
M
ft
r
H ft ft ft ft
> in in in in
0
p
+->
rH
r
0
•
ft
P<
P
P
P
•
rH
u
P
—
o
>
0
O
r
Ps r
•
a
o
CO
ft
0
*0
-
Oh
P
P
P
o
Ph
_
•r—
ft
i
a a
ft
1—
H
ft
O O
PQ
ft
pq
cq
ft
ft
ft
ft
ft
ft
I
E
E
64
Homogeneity
Sample
UNC-C
65
F m ax
significantly large for the three groups in the analysis of
scores, and the larger of the variances
though this combination
result (Box,
is
1954a), the
is
based on the smaller
SAT -V
n.
Even
likely to lead to an unrealistically improbable
ANOVA F
is
nonsignificant.
Since only four female Ss were from Catawba, these were combined
with the
UNC-C
female
pare this group with the
summer
UNC-C
session Ss, and
were used
female spring semester Ss.
the difference between the two groups on their
at a probability level
tests
t
PCT
4).
com-
In this case
scores was significant
between .01 and .02, the summer session
having the higher average (Table
to
women
The difference between the two groups
on their TIB-G scores has an associated probability between .06 and .07,
which
is
generally considered nonsignificant, but
is of
interest here for two
First, the two groups had different variances on
reasons:
TIB-G, and
since the larger variance was based on the larger n, the probability of the
difference
may
be slightly conservative (the n's are not greatly different).
Second, the direction of the difference
is
opposite that on the
PCT,
i.e. the
spring semester students have the higher average.
The same two groups were compared on the intelligence measures.
The two groups differed significantly
sure, the difference favoring the
Thus, while the
(.01 < p < .02) only on the
summer
school students.
UNC-C male sample may
be treated as a single unit,
reason to doubt the homogeneity of the female group.
there
is
which
this
SH-V mea-
Analyses
might prove important were generally run three times, once
in
.
„
66
with the combined
UNC-C
female sample, and once with each of the sub
groups
Complexity, Age, College Class, Sex and Residence
Since complexity
is
UNC-C sample
since roughly 20% of the
were older than normal
often discussed in developmental terms, and
(combining males and females)
for undergraduates, a series of
complexity scores to check for age effects.
was
n old"
(Table
p.
3,
As
and under 23 "young,
11
t
tests
In this situation,
These tests were
all
were run on
23 or over
nonsignificant
63).
a second rough check for age effects (rough since clearly in this
case, as well as the one above, selection factors other than age are also
operating) Sex x Cias s -in~College two-way analyses
for the Smith and
UNC-C
man and Sophomore)
performed on
For
the
the
were run separately
samples, class being considered lower (Fresh-
vs. upper (Junior and Senior)
„
No such analysis was
Duke sample because there were too few upperclassmen.
UNC-C
sample, a class effect existed such that upperclass-
men were more complex
sures, including the
PCT
at the .05 level for the
than lower on four of the six complexity
and TIB-G (Table
PCT
3)„
F max
is
mea-
nearly significant
analysis, but again large variances are based
on large n's, so the results are probably conservative.
In this case,
older
.
)
67
students, as defined by college experience, are
more complex than younger,
a finding perhaps supportive of a developmental positiom
(Apropos the
previous finding of nonsimilarity between the two female groups, the spring
Ss had 9 upperclassmen, the
pressed as proportions,
summer group
not large,,
is
sistent with the pair of results for the
findings
13.
Such as
PCT
The difference, even exthe difference
it is,
.
however (Table
3,
p.
63)
In this case, there
.
Smith sample,
in the
were no class
effects
what-
This analysis, however, constitutes the sole instance in which the
TIB-G and TIB-I provided
ing way.
con-
and contradictory for the TIB-G
The college class difference was not replicated
soever,,
is
The sex
results different
from each other
in
an interest-
effects are highly significant for the TIB-I and not at all
so for the TIB-G, this despite their correlations of .83 and .84 for the two
samples from Smith.
the higher average,
The difference between
seems
to result
from
the sexes, females having
the fact that a fairly large pro-
portion of the Smith males express severe doubts about the permanence and
such distrust and cynicism leading to scores of
reliability of friendship,
or 2 in Harvey's system.
Sex analysis (nor
in the
This finding did not appear in the
Duke ANOVA, below), so
minority group phenomenon
(not
it
UNC-C
1
Class x
represents either a
an implausible hypothesis) or sampling er-
ror.
Duke's selection practices permit a test
South or North affects complexity scores.
of
whether residence
The question
is
in the
relevant because
)
68
Southern education
x Sex
ANOVA
sults (Table
is
often regarded as inferior.
was run.
3,
becomes so when
Neither sex nor residence produced significant reApparently, either residence
63),
p.
Accordingly, a Residence
intelligence
is
high.
(No check was
is
unimportant, or
made on
ability of intelligence scores for the two residence groups.
women, several
the
of the intelligence
the
compar-
For the Duke
scores did correlate significantly with
TIB -G,
Complexity, Schools and Intelligence
College x Sex
ANOVA
1
s
were run
Since the heterogeneity of the
scores.
for both intelligence
UNC-C
and complexity
female (UF) sample might
be important here, each analysis was done three times, once each with the
combined
UF
sample, the UF-spring subsample, and the
sample.
For
the
combined
UF
sample, n's
The subsamples make the
proximately equal.
no serious problem for the complexity
homogeneous.
cant for the
For
PCT,
in the
the
UF-summer
analyses tend
to
n's fairly unequal.
ANOVA's because variances
UF-spring subsample, F max
is
sub-
be ap-
There
is
are
borderline signifi-
but the large variance and large n's go together, mitigat-
ing the problem.
The situation
For
all
three
UF
is
otherwise for the intelligence score comparisons.
samples
all
variances based on the Shipley-Hartford are
69
SAT -V
significantly (or nearly so) different, and in at least one of the cases
is
also nonhomogeneous
„
For two
of the variations either
roughly equal (combined UF), or there
is a
size and variance magnitude (UF-'spring)
for the
UF»summer
,
favorable combination of sample
Not even that much can be said
However, the pattern
sa.mple„
sample sizes are
of
ANOVA
results is
identical in all three cases, and is in accord with expectations and selec-
There
tion criteria:
is
an enormous intelligence effect across schools,
F's range from approximately 100 to approximately 350,
gence-based measure not related
to schools is the
The only
SH Conceptual Quotient
(Abstraction divided by Vocabulary and multiplied by 100),
there
is
main
a sex
higher (Table
6),
SAT -M, males,
as would be expected, being
for complexity are similar but not so
All complexity scores for all three analyses
school effects.
significant.
Additionally,
5),
The results
(Table
effect on
intelli-
For
the
combined
UF
overwhelming
show significant
sample, only the school effects are
For the subsamples, additional significant results occur
pattern which follows logically from the differences in their scores:
the
UF-spring group was used (who were lower on
the
TIB than were the UF-summer
males averaging higher on the
the
UF-summer
actions:
On
the
PCT
Ss),
the
PCT
in a
When
and higher on
significant sex effects appear, with
and females higher on the TIB,
With
group, sex effects disappear, and are replaced with inter-
PCT, Smith males, UNC-C females and Duke males have
higher means than the opposite sexes at their respective schools.
The
HH
HH
H
H
H
1
H
H
H
1
70
f
2
f*H
ft
i—
r
-
i
c
CO
r^.
oo
H
H
o
.
co
,
>
cO
or-H
1
.
CO CO
1
1
r
-x-
-X-
i—
r—
H
r—
00
X
r-H
00
r-H
oo
CO
00
LO
CO
•
,
vO
CO
CO
r-H
co
00
00
X
„
X
CO
C
a
o
o
pH
H->
• rH
•
X
c
o
CJ
X
CD
PC.
>
-X-
-X-
CD
P
O
•5C
|
M
00
OO
<1
.
r
.
*—
^
CO
'
00
t
co
r—
.
f—
,
H
CO
CO 00
r-H
,
ro
CO
H
r—
r-H
#
.
00
r-
co
oo
r-H
00
r-H
00
CO
X
„
X
00
(X
co
a
<D
rd
<D
co
£>j0
<D
nJ
a
0
P
,P
+->
a
u
X
r—
o
H
X
oo
CO
•
i
.
0s
.
p—
H
o
a
X CO
o o
.
CO
«%
rx
00
r-H
CO
00
CO CO
00 00
rH
.
•
r-H
00
co
X
r.
00
r-H
M-l
U
u
£
p
CO
1
f-l
o
X
o
pH
H-H
•
nj
P
P
H
X
i
CO
r—
•
oo
co
H
r-H
o
00
CO
oo
CO
•
.
H
o o
e
00
r—
.
.
O'
«
00
o-
X X
CO
r\
•%
CO
a
00
i—
X
rv
X
0-
r-H
r—H
00
CO
r-H
CD
?n
o
o
O
U
x
rP
Oj
0
w
Oj
r-H
H
o
£-1
a
CD
to
nJ
Eh
(/)
0
O
P
0
CuO
0
co
w
0
rH
o
a
h-t
o
00
CO
ID
r-H
•
.
.
00
co
H
r-H
00
o
p-
oo
„
CO
.
•
o o
oo
co
o
rH
r—i
00
1
o
o
—
r-H
oo
co
X
X
r—
o
o
co
>c
<u
o
£
0
CUD
rH
r—
r-H
•
•>P
>
X
i
rOO
.
H
00
oo
•
o
oo
CO
H
00 00
o
0-
r-H
CO
o
»
o o
.
00
rv
CO
ft
vO
00
CO
r-H
X
X
f)
00
LT)
r-H
0
4->
oo
CO
OO
X
n
X
P
•
pH
<+H
o
$
o
£
<
—CO
0
i
o
p
p
o
CO
X
0
<hh
rP
•
u
CO XJ
co
r—
1
o
o
CO
o
o
CO
rp
o
X
P
.
XI
CO
CO
x
CO
X
CO
X
0
•
Uj
M-l
>-.
a
i
o
£
H
0
0
+->
T
0
0
rH
i—
nj
cti
“
I
a
•
a
o
u
<HH
o
o
p
£ p
1
H
e
0
Sh
'HC
0
r-H
Ph
a
M
aj
co
P u
H 'Z
£ P
-l_>
^
a,
a
nj
co
t
o o
IT)
CO
.
.
N N O
CO
.
«
I
1
o-
co
VO
U
00
1
co
CO
—
o
.
co
1
i
CO
O
CO
r-l
.
.
Tf
O
CO
00
co
CM
o
00
H
sO 'F CO
CO rf
vO CO
.
ro
CO
.
I
.
1
CO
.V
1
00
o
o
CO
Xf<
'vf
CO
.
CO
CO
CO
—
.
xf
1
1
[
^
.
CO
CO
>
xO
0-
00
*
I
<<
(continued)
O
-F
1
5
CO
i
—
t
CO
•
-F
^F
«
CO
O
a
C rH
oo
CO
00
Jo
Table
u
—
>
I
P
.
t'-
00
CO
CO
i
i
rQ
CD
co
ctf
O
o
P
O
H
O
> c
CO
!>
P
+->
+->
o
LD
CD
Ttf
P
Ftf
00
ITl
CD
t-P
P<
CJ
F
CO
ctf
rQ
p
«
ctf
ctf
vO 'F
CO CO
r-4
o
P
CO
p
m0
P,
-t->
C
P
ctf
X
w
o
o
CO
X m*
rP
X
.
V
CO 'V
CO
CD
4-1
1
CO
>>
r—
rtf
CD
r*H
O
Ph
<+H
rH
rP
rP
CO
co
•
CO TD
1
i
—
“
o
o
o
co
V
P
CD
'-H
P
o
rH
4->
•
i
•
u
u
£
D
I
^
° > 3
> < H U
H H
>
CD
P
a
rP
n
c
'
i
i
rH
P,
ctf
i
P
c
co
co
co co
co
co
(
b
H
H
H
H
in
CO
—
00
Lfl
CO in
ro
.
on
ro
,
iii
i
t
H
I
f
71
i
pq
i—
H
i
.
m
ro
rO
ro
00
ro
in
i—
rv
H
oo
o
o
0
Pm
0
>—
ro
ro
O- vO OO
•
ro
.
in r-H
00
i—
V.
r
TD
P
O
V
i
i—
•
H
oo
aJ
CO
o H
CO
•
.
$£
1
CO
vO
c
co
.
ro
r-H
vO
r-H
ro vD
.
ro
ro
.
00
s-
I
ro
r—
rH
r-H
r-H
»>
ro
r-H
CO
0
0
rd
co
r—
O
CO
<h
a,
rO
Hi.
o
PP
S
0
")r
i—
rT
O
-x-
-X-
Jh
0
i
biD
0
Q
P
/“N
oq
MM
Eh
JI
CO
n-
o X
•
:o:
(NJ
i—
ro
•
ro
r-H
r-H
oo
ro
ID
in
CO oCO vO
co
.
1
in
o.
CO
ro
ro
<NJ
LO
•
00
LO co
•
co
1
O'
1
ro
ro
r-H
(D
MM
o
o
z
p
CO
0
rH
4->
•
cd
eJ
o
1
X
i
pq
—
.
i
1
Jm
h
0
Pm
mm
H
H
1
r—
i—
.
I
i
rH
oo
OO
.
CO
CO
co
CO
00
CM
ro
ro
co
**
—
'
OO
ro
.
CM
(\3
t
CD
Jm
^0
o
CD
vO
o
o
CO
rH
o
aJ
Ph
CO
0
co
H
X
CD
—
i
Ph
X
co
<d
CD
co
Jm
O
CD
qr
+->
•
d
H
>o
?m
rp
CD
*
nj
H
U
CO vO
CO
ro
ro
o
.
Ph
H
1
,,
CO
in
.
.
CO
rH
rH
a
o
X
o
H
1
O
vn in
—
i
i
•
rcn
.
•
CO
ro
ro
0- in
rH 0-
m
„
ro
r-M
U
CO
Pm
•
rH
X
CD
rH
cu
i
H
U
Pp
X
H
CO m
ro
.
.
1
00
1
H
cn
— H
ro
c
1
•
•
1
O
O O
ro
•
H
n-
co
1
,
o
m o
ro
.
co
i
rH
£
o
o
1
1
tt
rH
ro
H
ro
Mm
0
2
O
z
<
Jm
P
O
CO
w
o
0
CO
rP
X
CO
—
o
rH
CD
CD
X
p
CD
•
co
T3
CD
^
o
CO
1
o
CO
'O
X
MM
,P
X
0)
co x’
co
x
CO CO "O
o
>>
d
£
1
o
£
H
CD
r
— 2>h
t
I
Cl,
(H
CO
CD
r—H
Ph
£
rd
CO
fn
g>
0
^
8 u
*p u
£ £
—
cd
CD
•H
S
CD
r—
Ph
£
cd
co
£
P
u
'+*
Ph
u
u
z
£
co
I
rj
Ph
a
rd
co
p
o
«
m
cf—
en
co
.
r-H
.
—
H
1
CO
O
e'-
.
co
^
c-
1
II
1i
i
I
co
r— vo
—
<
l_n
»
CO
O' CO
O)
o
h m
O
.
.
co
~
-cf
CO
CO
O
.
C
CO
CO
CO
O' CO
00 CO
.
co
—
i
CO
CO
co
co
i
p
fH
in
o
NO lo u~>
CO O'
o
p
CO
CO
N
co
.
.
CO
r—
r—
P
u P p
o Jh o
o o co
CO a ?h
— CO P
CO
I
cci
<D
*h
O
o
(continued)
CO
&
1
TO
6
o o
sp
.
p V
U o
o ^
o O
m o
h CO N
co
cco
.
.
co
co
CO
t
rQ
"
o
O
r—
£ °
—
1
O'
n
.
.
C-
CO
I
p
o
vO
U
CO T3
CO
•
P
p
4->
CO TO
V
1
£
Ph
Ph
nj
co
H
Ph
O
O
r—
o
CO
p
o
•
a
rP
cci
V
I
V
Ph
rd
|
•
I
rP
_
—
W
u
x m’
O
u
o
I
bO -
IT)
co
IF
t—
cci
s
g
p
—
p
>
p
u
CO
p
£
p
I
cci
CO
r*H
rP
I
I
Ph
HP
a,
.
1
i
P z
u
<+H
Ph
i
i
H
CO
a
o
X
p
£
Ph
1
CO
O'
CO
o
o
i
O
H
O
P
U
U
°
P
T
—0 ^
o
i
CD
co
^
TO
in
CO
.
P
H
Ph
I
I
CO
?H
p
0
u
P
+H
OD vO
cci
CO
Ph
I
CO
Table
cci
0)
Ph
-JP
ri
>
P
Ph
i
H H
U U
Ph
Ph
a
GF
i
1
1
i
i
cq
cq
CP
CQ
i
H
1—
i
1
—
—
H H H H
Jh
-X-
;
P
3
<
72
direction of the interaction
significance for the
UNC-C males
is
reversed on the TIB, and does not quite reach
TIB-G scores,
For TIB-I, however. Smith females,
and Duke females have the higher means, and the interaction
is significant.
All three complexity scores (PCT, TIB-G, TIB-I) provided signifi-
cant school effects, and in each case the
that for
Duke highest.
Smith and
on
for
Smith was lowest and
For the two TIB scores the differences between
UNC-C were extremely
nearly in the middle.
mean
In
view
small; but on the
of the difference
SAT scores (approximately
PCT, UNC-C
between Smith and
is
more
UNC-C
100 points) and race, with all the implica-
tions associated with the latter,
it is
of particular interest to
know which
!
:
groups differed from each other.
individual comparisons (Winer,
the
ANOVA
using the combined
suits is precisely the
same
Accordingly, for each complexity score
1962)
|
were made
UNC-C
for all pairs of schools in
female sample.
The pattern
UNC-C
Smith and
for all three scores:
of re-
are not
i
different
from each
other, but both are different
from Duke.
This pattern
holds whether one selects critical values according to "risky" or "con-
servative" principles.
PCT, TIB and
Intelligence
Results regarding the relationships between the
PCT
and TIB (see
.
Table
2,
61)
0
No
scores.
TIB-G
p
may
significant correlations
or the
PCT
and
and probably
II,
for either the
The correlations ranged between
Nor was there support
I
were found
PCT
and
and TIB-I, for any of the six samples or for the entire
combined sample.
terns
The two tests yield orthogonal
be succinctly stated:
07 and +.21.
for the earlier suggestion that
III,
were discriminations
Harvey's Sys-
of Schroder’s
simple
A
scores, whereas Harvey's IV's were like Schroder's 6's (and above).
chi-square testing this hypothesis (TIB 1-3 vs.
and
7)
PCT
PCT
relations with the intelligence
In
my
7
and
measures
8).
is
Further, the pattern of cor-
basically the
combined samples.
scores, and an
is
The
never very large
PCT
(It
SH-V and SH-A.)
so at the
of course,
typically ranges between 300 and 400.
lated with SH-Vocabulary,
SAT -Mathematics.
set of these correlations,
is for
the
This provides a particularly broad range of intelligence
N which
find that both the
with
for both
typically quite small (though statistically significant).
The most important
we
same
opinion, there is no great cause for concern, however.
proportion of variance involved in these correlations
is
6
.
and TIB-G correlate more strongly with intelligence
than they do with each other (Tables
and
less than 6 vs
was nonsignificant.
Both the
tests.
4,
most about 8%
and the TIB-G are significantly
< .01) corre-
SH -Abstraction, SH-Total, SAT -Verbal and
should be noted that
In
(p
In this case,
SH-T
is
positively correlated
magnitude the correlations range from
of the variance is involved.
A
.
14 to
.
27,
few of the correla-
V
.
74
Table
7
Correlations Between Intelligence Measures and the
Paragraph Completion Test (PCT)
SAT
Shipley-Hartford
Vocab.
Smith Females
r
.
n
Smith Males
r
.
n
UNC-C Females
r
Duke Females
Duke Males
r
.
.
71
Samples
n
.
.
n
.
18
26***
410
08
66
05
71
.
72
.
09
68
18
-.01
r
.
.01
67
r
r
19
T otal
.
66
67
66
Combined
-
30*
n
n
*p
19
.
68
n
UNC-C Males
23
66
Abs
05
72
16**
410
.
.
.
.
24
66
CQ
.
68
04
68
22
67
67
15
-.
18
05
66
14
66
-.04
.
-.
04
71
71
13
-.07
72
72
23***
410
.
02
66
04
410
Verbal
.
02
Math.
.
.
07
.
10
.
47
.
.
.
06
47
03
47
-.01
47
04
70
.
14
70
12
.
68
18**
328
.
07
46
46
.
08
50
50
16
68
20***
328
.
05
**p =.01
***p
=
.
001
Abbreviations:
SHSH -A
SH-T
SH-CQ
SAT-V
SAT -M
Shipley-Hartford Vocabulary
"
"
Abstraction
"
Total (V + A)
"
"
Conceptual Quotient ((A
Scholastic Aptitude Test--Verbal
"
” --Mathematical
"
•«
*—
V) x 100)
75
Table
8
Correlations Between Intelligence Measures and the
"This I Believe” Test (TIB)
SAT
Shipl-ey-Hartford
Smith Females
r
Vocab.
Abs.
Total
07
53
03
53
.01
53
33*
.
.
n
Smith Males
06
59
r
n
UNC“C Females
r
.
n
UNC-C Males
r
.
n
Duke Females
r
.
n
Duke Males
r
.
26*
64
“.
.01
66
23**'!
Samples
n
371
02
24*
70
11
.
08
64
59
23
70
r
.
.
19
59
Combined
~
59
66
n
*p
.
=
14**
371
.
.
22
CQ
08
53
.40**
59
59
23
64
64
15
.09
59
».
.
02
“.
35**
70
07
“.07
66
20***
371
o
05
371
17
Math
.
.
10
.
38
.
17
.
44
.
10
.
24*
.
69
.
26
08
41
02
69
23
62
.
27***
295
.
26
38
44
41
.
13
41
41
59
70
66
.
15
Verbal
03
62
23***
295
.
05
**p “.01
***p “.001
Abbreviations:
SH-V
SH-A
SH“T
SH“CQ
SAT-V
SAT-M
Shipley-Hartford Vocabulary
"
”
Abstraction
”
"
Total (V + A)
"
Conceptual Quotient ((A
Scholastic Aptitude Test“-Verbal
"
"
-“Mathematical
.
11
11
V) x 100)
76
samples are larger, but even the largest
tions for the individual
Within the samples no impelling pattern seems to exists
is
In
only .40.
two cases,
Smith males and Duke females, SH -Abstraction and SH-Conceptual Quotient
correlated significantly with TIB-G.
while in the former they are positive.
relationship between
SH-A and
SH-A
This weak relationship
garding his
PCT
is
that
if
is
at least
consistent with Schroder's prediction re-
His expectation was based on
The SH-A tasks are not interpersonal and they are not
to fully
engage complex thinking.
PCT
contradicted by the present findings.
anything the
my
this skill.
prediction of nonrelationship between the
scores
makes unlikely
based on analytic thinking,
and the Abstraction score.
two considerations:
complex enough
is
SH-CQ measure
or
This ambiguity, plus the weak
the complexity scores,
earlier suggestion that complexity
insofar as
not in-
However, for the latter group the two correlations are nega-
dependent.)
tive,
SH-A and SH-CQ are
(Again,
PCT
will be related to
On
the other hand, his
and vocabulary and/or. verbal
That
Schroder argues
is,
complex information-processing
or decision-making aspects of intelligence, but not to verbal skills per se,
and particularly not
both the
PCT
This finding
memory
skills (e. g. vocabulary).
and the TIB are related
is
Finally,
to
to
vocabulary and
In this study,
to
verbal skills.
consistent with Harvey's report (1966).
since Harvey typically analyzes his data using group
means
instead of correlations, in each of the six samples Ss were grouped by each
complexity score (separately) and one-way
ANOVA's run
for each of the
77
intelligence scores and for age (four groups
scores of 2-3,
4,
5,
6-7).
No surprising
PCT
also:
were discovered,,
In a
were used
results
for the
few cases (four), significant correlations between complexity and intelli-
gence had been found, but the present
presumably due
to greater
linear hypothesis, plus
its
power on
F
ratios
were
This
nonsignificant,.
is
the part of the correlation in testing a
lack of concern for unequal n's, etc.
instances the correlations and F's were both significant
significant F/nonsignificant r combinations
„
Three
In three
There were six
of these
were
for
groups as defined by TIB-I scores, for which correlations had not been calculated,,
the
No
TIB there
to be alike,,
particular pattern appears for these results, except that on
is
some tendency
for Levels
II
and IV
to be alike
and
I
and
III
This latter pattern, but not the linear findings, are consistent
with Harvey (1966)„
Reliability
Of course, the best way
to learn about the test-retest reliability of
these tests would be to devise alternate forms and administer both versions
to
some
Ssc
(R ozeboom,
Lacking
this,
an estimate of the reliability, coefficient alpha
1966), can be calculated on the basis of the variances of the
items and of the total scores.
The scoring procedures used above, how-
ever, do not lead to values which can be used to calculate alpha.
Accord-
-
1
78
ingly, the
PCT
and TIB item scores were used
patible with the application of alpha.,
scored items was found, and
items
(5
and
10,
to derive total
each case* the average of the
In
this value multiplied
The variance
respectively).
scores com-
by the total number of
of these
scores was used in
finding the alpha coefficients.
In general, these scores
Summed
scores.
(SumPCT and SumTIB) behaved
scores generally correlated
Also, the
summed
based on the other
Smith females,
scores tended not
SumPCT
„
in the ,70's).
correlate significantly with scores
Here, however, there were exceptions.
test.
both significant at the
to
with the corre-
in the ,80's
sponding regular scores (SumTIB with TIB-I being mostly
as the other
correlated
.
235 with
SumTIB and
241 with
.
05 level; and in the Duke male sample,
barely misses significance with TIB-I
(r
= .231),
For the
TIB-G
SumPCT
With the combined
samples, none of the intertest correlations approach statistical significance.
For the combined samples, the estimates
are .406 for the
ues for the
for the
PCT
PCT
range between
TIB (Table
able, and the
and .776 for the TIB.
PCT
9).
On
cannot.
.
294 and
.
of test-retest reliability
Within the samples, the val-
515,
and between
.
605 and
.
814
these grounds the TIB can be regarded as reli-
'
I
1
I
i
C
1
i
79
ft
•
0
d
—
i
r-H
1
r-
-sf
00
o-
6
o
o
o
ft
*
O
0
ra
ft
pH
0
rO
d
cd
xT
Q
1
0~
r-
S
to
0
0
r—
ft
cd
r-H
CO
Q
d
d
0
—
i
00
•
P>4
U
o
£
CO
s
1
CD
—
i
a
CO
0
CM
r-H
IT)
cd
•
ID
2
0
CD
r-H
a
CM
r-
cd
O'
0
—
m
m
0
t
ft
0
rd
U
H
CO
CO
1
o
£
d
d
0
\p
Ph
cd
ft
ft
cd
00
CM
MO
•
•
r~
—
i
<J
ft
4->
•
I—
a
CD
0
r-H
r-H
ft*
ft
o
CM
ft
cd
CO
s
x
4->
(D
r-H
rH
cd
d
d
d
d
0
CO
•
CO
H
U
Ph
ft
PQ
l—
H
R
u
mO
m0
cd
cd
ft
ft
ft
ft
r-H
r-H
<
<
i
Chapter VI
DISCUSSION
Summary
The major purposes
ter
III.
A.
solved:
of this study
were outlined
The main point
Schroder’s
TIB scoring
of interest
PCT
it
is
seems
Chap-
to
have been unequivocally re-
and Harvey's TIB are uncorrelated with each
revised to
measuring
make
the
same
variable.
should be emphasized that even the global score
The current data cannot
the failure of one test to
measures
is
Even when
more obviously interpersonal
it
be interpersonal), no relationship with the
the
at the close of
The relevant results are as follows:
other, and are therefore not
tion (and
of Results
tell
PCT
is
it is
"better" than the other.
orienta-
assumed
to
exists.
us whether this independence
measure what
in
the
supposed
to,
is
based on
or whether one of
As has been shown,
the scoring
criteria are vastly different for the two, but each has a basis in the re-
spective theory.
The suggestion
that the
(
80
)
two theories might be dealing with
81
was not substantiated.
different portions of the dimension
The most obvious consequence
of this is that the theories
ciated research are irrelevant to each other.
in
background, language, terminology, etc,,
combine the findings
in
and asso-
Irrespective of similarities
it is
simply not appropriate to
attempting to define or understand "integrative
com-
plexity, "
A
B,
The pattern
telligence.
plexity
second major question
is
the
same
significantly related to
is
is
the relationship of complexity with in-
for both the
measures
of
PCT
and the TIB:
simple vocabulary and ab-
straction ability, and to verbal and mathematical skills as
SAT.
These findings essentially replicate those
much weaker
.40's
when
than the Schroder et
al.
com-
of
measured by
Harvey
the
(1966), but are
(1967) report of correlations in the
The results con-
the variance of intelligence scores is great.
tradict Schroder's prediction that complexity should not correlate with verbal skills, and especially should not correlate with
cabulary)
memory
,
The correlations with
intelligence, while statistically significant,
were not large, and were certainly
not large enough to threaten the value
or standing of either test on this basis alone, provided
either
functions (vo-
measure
of
complexity has more
with the other complexity measure.
in
common
it
is
understood that
with intelligence than
The relationship with intelligence
strong enough that research using either measure
unless controls for intelligence are incorporated.
is
is
of questionable value
Schroder routinely
82
matches Ss for intelligence
(e.g.
Schroder
Harvey, 1966) often does not mention
do
it,
et al.
f
1967);
Harvey
(e.g.
and therefore presumably does not
it.
C.
As
to the existence of
results are not clear-cut.
sex differences in the complexity scores,
Several alternative answers are available de-
pending on which sample and which analysis one wishes to rely upon.
ANOVA
broadest answer, based on the Sex x Schools
UNC-C
female sample,
is
female subsamples did lead either
At least one
ANOVA
to
result which
the TIB.
was done independently
minority groups are likely
male has
In this case,
makes some sense
to be
in playing his role).
in
UNC-C
sex or interaction effects.
and that was for the TIB-I score; that
measure based on
However, the
the two distinguishable
sex differences could have been demonstrated.
effect,
with the complete
that there are no sex effects.
same analysis performed separately with
The
terms
for each school in
which
Only Smith showed a sex
is,
a strongly interpersonal
females were more complex, a
of the kinds of stresses to
which
exposed (e.g. the greater difficulty the
Since students from Smith are probably also
the lowest of the three groups socioeconomically, this finding partially re-
plicates Gardiner's (1968), that
among children from lower
grounds females were more complex than males.
ferences, then,
is still
open.
Whether
every situation
in
The question
of sex dif-
a difference will exist apparently
depends upon population characteristics.
in
class back-
A
check should probably be made
which a difference might be important.
Differences
D.
from any
of
many
complexity scores among schools could result
in
most obvious being
factors, the
socioeconomic background, these being
race and
intelligence,
all interrelated.
The mean com-
plexity scores for the three schools did fall in the expected sequence, but
individual comparisons
show
between Smith and
that the differences
are nonsignificant, and that both differ significantly from Duke.
UNC-C
This means
that neither race per se, nor fairly large intelligence differences (approxi-
mately
sigma on the SAT), are sufficient
1
Of course, this could be attributed
ences.
scores, but that problem
is
PCT
ably low for the
to a floor in the
(.41),
former ranged between
and were mostly
Schroder
but reasonably high for the
PCT were
et al.
in the
.
.
05 and
and the latter between
17,
unaccept-
(.78).
The
.
15
(the
and .47,
and were much lower than those reported
20' s),
(1967) (range
.
TIB
is
lower than those for the TIB
.
39 to
.
in
62).
should be kept in mind that the scores used to calculate coefficient
alpha are not those regularly used in research.
difference
the
complexity
intrinsic in the tests and theories.
inter -item correlations for the
is
complexity differ-
Test reliability, as estimated by coefficient alpha,
E.
It
to indicate
sum
is
probably not important:
of the
necessary
is
presumably
(but adjusting for
still
case of the PCT, the
Although the score currently
two highest item scores, the mean of
has been used and
item scores
In the
acceptable.
all the
in favor
item scores
Using the sum of the
nonscored items) instead
of the
mean
is
for calculating alpha, but does not change score relationships.
84
The
set of scores derived
from
TIB and used
the
alpha bears
in finding
no such close relationship to regular TIB scores, for the simple reason that
Harvey does
to be a
(the
weakness
number
sible
In
<,
not assign individual item scores.
it
somehow bear
requires the scorer to
sources of information, which
of items)
my
rater keeps
in that
This practice seems to
in
me
mind
10
essentially not pos-
is
opinion, the kernel of this objection holds even though the
some
sort of mental cumulative tally:
In the present case
we
separately for each item,
did,
obtain the Interpersonal score, and later to
estimate reliability.
Why
In both cases, the
sum
summed
not write
down?
it
This enabled us to
the item scores so as to
scores correlated highly
with the regular scores.
The low
reliability of the
PCT
is
something of a threat
cially in view of the fact that high scores are so rare:
tions which result
from low
reliability are a serious
to
it,
espe-
The misclassifica-
problem.
That
Schroder’s research results conform so nicely to theory suggests,
of this
low
that his cutoff score
reliability,
necessary for a more reliable measure.
(and reverting to
sure.
Of course,
mean
it
may
be higher than would be
Increasing the number of items
scores) would probably help to stabilize the
would be valuable
to
is the
person best equipped
and provide authoritative ratings.
liability of the
instrument
may
mea-
have an actual test-retest coeffi-
cient instead of an estimate, and responsibility for this
Schroder’s, since he
in light
is
basically
to devise alternative
items
Until such an effort is mounted, the re-
be regarded as unsatisfactory.
.
The
situation of the
TIB
Although coefficient alpha
is different.
is
acceptable, the onus for a true test-retest reliability correlation for the
TIB rests on Harvey,
just as
it
does on Schroder for the PCT, in spite of
the present satisfactory value of alpha.
This
is
so because, beyond the re-
sponsibility of the test creator to provide adequate information about his
the set of
test,
TIB scores used
to calculate alpha
bear a less certain re-
was
PCT.
lationship to the usual TIB scores than
turn
is
partly due to the fact that confusion in the TIB scoring
between a score
it is
not
retest)
the case for the
of 2 or 4 on the one hand,
uncommon
would lead
and
1
or
3
is
This in
typically
on the other.
Thus,
that the switch of the score on a single item (as on a
change of two points
to a
in the global
would be scarcely noticeable on the sum-of-items score.
score, while
it
In short, al-
though the correlation between TIB-G and the summed-items scores
is
high, the categorization of Ss into four levels based on a retest might not
be so
much
in
agreement.
I
would not be surprised
if
alpha overestimates
the reliability for the TIB, but the issue could be resolved empirically
more
quickly and easily than by debate.
Comment and
Given a choice,
for two reasons.
I
Implications
personally prefer Schroder's position to Harvey's,
First, the theory
is
more
elegant, in that
it is
simpler,
86
more
easily applied to a broad range of cognitive domains, and permits be-
more
havioral predictions which are
more powerful than Harvey's, although Harvey seems
criteria, the theory is
regard his theory as relating tb a greater variety
to
Schroder, so
in this
By these
clearly related to the theory,,
sense Harvey's
The second reason
I
may
be the
of
more powerful.
prefer Schroder's position
gives the impression of greater rigor
behavior than
is
that his research
The most obvious case
0
in point is
the difference in approach to the scoring criteria for the respective tests.
Beyond
this is the looseness in
what Harvey
is
willing to call validational
evidence, plus his seeming failure to control for intelligence in these validational studies
Naturally, the best comparison of the two tests would be
o
with regard to predictive efficacy, but this was not possible
It
would be
difficult to
compare
there seems to be very
little,
if
in this study.
the tests using existing literature,
since
any, overlap in the actual dependent vari-
ables which have been investigated.
Although Schroder's seems the stronger position, he
problems.
As indicated
earlier,
if
integrative complexity
distinguished from a dimensional position,
of connectedness,
or
how
not without
At the theoretical level he has yet to elucidate the "connected-
ness" concept.
is,
is
it is
yet
it is
by coefficient alpha).
much lower than
be
will have to be on the basis
never made clear what this connectedness really
represented
The second problem
it
is to
is
in the theoretical
model.
the low reliability of the
The PCT
PCT
(as
estimated
inter -item correlations found here
those reported in Schroder
et ad.
(1967).
were
Coefficient al-’
87
pha will become larger as the inter-item correlations increase, but one
must assume
that the scores (ratings) are reliably
be taken as an estimate of test-retest reliability.
determined
if
alpha
is to
Thus, both inter -item
and inter -rater correlations must be adequate.
It is
interesting to speculate as to whether the change in the applica-
tion of scoring criteria (see Chapter
III),
since Schroder et al„ (1967)
lated to the reduced item intercorrelations
shift in distribution of
PCT
scores
may
.
is
re-
Schroder's explanation for the
be interpreted as implying that the
i
1967 inter-item correlations were inflated by the interpretations
made by
These interpretations, being unsystematic, should reduce inter-
the raters.
rater reliabilities, but the rater reliabilities reported in Schroder et al.
seem
fully adequate
(.
80-. 95).
correlations in Schroder
et_
ah,
The combination
of
item and rater inter-
implies a high value of alpha and
son to expect that alpha reflects reliability.
some rea-
The low inter-item correlai
j
tions in the present study virtually preclude high reliability, irrespective
of rater reliabilities.
changes
PCT
Since the
scoring procedure were introduced because of the difficulty in
in
distinguishing
the
The former situation seems preferable.
among scores
at the
complex end
format (two or three sentences
in
of the
dimension, and since
response to a sentence stem,
with a short time limit) restricts the expression of complex, abstract concepts, the
would be
to
draw
most obvious
to
first
attempt
to solve the
low reliability problem
arrange for stimuli and/or time requirements which would tend
(or at least
permit)
more complex responses.
Other issues become problems only within a context.
now, unless E
willing to screen large
is
few who are complex, and has reason
will be
rewarded
Smith and only
nearly 300
value.
UNC-C,
in
stand, or
is
to find the
were found
at
a complexity rate of less than 1% for a total of
of us are,
complexity (defined
This problem
information about middle complexity Ss.
them
stands
advance that his efforts
interested in studying naturally occurring or
either rare or nonexistent.
for
order
it
Schroder's theory, measure and research are of limited
is
most
of Ss, as
to expect in
of Ss in
(note that no Ss with scores of 7 or higher
2 at
Ss),
one
If
numbers
As
is
in
random groups
Schroder's terms) will be
compounded by
There appears
the scarcity of
to be little
concern
Schroder's work, consequently one does not know where they
where
to expect
them
to stand,
relative to the
extreme groups.
It
possible that changing the test as suggested above would improve the dis-
tribution of scores.
Learning about the behavioral characteristics
mid-complexity group
is a
matter of research.
of the
At least one, and prefer-
ably both, will be necessary before the variable will become useful in most
research.
Given these
to the
Bieri,
difficulties,
the interested researcher
dimensionally oriented theories and measures.
1966) is the leader in
terms
article mentioned earlier (Scott,
may wish
to turn
Of these, Bieri (e.g.
of published research,
but Scott's
1969) shows evidence of an unusual, and
laudable, degree of concern for preliminary, methodological questions.
Perhaps the major weakness
in Scott's
study
is that he,
too,
fails to control
89
for intelligence.
Beyond
this,
there seems to be considerable promise in
his work.
The decision
put one so
much
to
adopt a dimensional definition of complexity
at odds with the "-integrative
be thought at first.
As mentioned
dimensionality (Harvey,
level, will not
1966),
become clearly
earlier,
not
complexity" theorists as might
TIB scores correlate with
and Schroder,
different
may
REP
at least at the theoretical
from the dimensional viewpoint un-
less "connectedness" can be defined so as to achieve the distinction.
The correspondence between Schroder and the dimensionalists may
potentially be increased in two different, but related, ways.
complex thinking seems
to involve
two components;
first,
Schroder's
the ability to per-
ceive and tolerate alternative interpretations of a situation or event, this
being
of
It
in
some
may
a sense the first stage of complexity, and second, the development
sort of abstract basis for relating and evaluating the alternatives.
be that the first of these could be fruitfully conceptualized as a dif-
ferentiation of evaluative judgments (the development of several kinds of
"good" and "bad") and/or a reduction
of the correlations
between an "evalu-
ative dimension" and other nominally nonevaluative dimensions.
Either of
these could lead to judgments involving relative value, various kinds of
value, and simultaneous positive and negative value, as well as relatively
nonevaluahive judgments.
The second way
to increase the relationship
plexity and dimensional complexity would be to
between integrative com-
assume
that the connecting
90
or relating component of Schroder's theory represents superordinate di-
mensions, and that there
dimensions
than
it
is
therefore some hierarchical organization of
Such a conceptualization
,
would be with categories
,
is
more
difficult with
For example, one
of the
dimensions
few efforts to
investigate a hierarchical structure, that of Zajonc (I960), did not use di-
mensions.
It is
tempting to suggest that these hypothesized superordinate
dimensions are based on the person's philosophical values, but
be the case, distinctions must be
if
this is to
made between values which are and
ues which are not capable of preserving alternative viewpoints
,
val-
The struc-
tures would be quite different in the tv/o cases.
It
would seem, then, that the insights from the theory
of integrative
complexity could be applied to research on natural or random groups of Ss
either by
(a)
improving on Schroder's measuring techniques and/or study-
ing middle as well as
Schroder's position
in
extreme complexity
Ss,
or
(b)
by interpreting
dimensional terms, and looking for the differentia-
tion of evaluativeness or for a hierarchical organization of dimensions.
Either approach could provide valuable extensions of the theory of cognitive structure.
:
INSTRUCTIONS, ITEMS, AND A SAMPLE PAGE FOR THE
PARAGRAPH COMPLETION TEST AND THE
"THIS I BELIEVE" TEST
i
j
.
!
i
i
i
On
the following pages you will be asked to write a short paragraph
on your beliefs about or opinions of several different things.
On each page you
will find either the beginning of a sentence (a
word
j
or phrase followed by several dots, for example, "I like
sentence beginning with "This
I
.
.
believe (for example, "This
.
I
")
or a whole
believe about
j
human
nature")
„
i
•
J
j
If
I
H
the beginning sentence is unfinished, complete
f
it
and try
to
add
at
I
least two further sentences on the
same
topic,,
i
1
If
j
the beginning sentence is complete, try to write at least three sen-
.1
I
I
-
i
tences on the topic,
I
I
In other words, in either case try to write at least three sentences,
'
\
i
These answers cannot and
will not be graded right or wrong,
so
I
!
please write what you genuinely feel or believe.
Work
quickly, but be as
clear as possible.
i
J
You
will be given 130 seconds for each
answer.
After 110 seconds
I
!
(
92
)
will say "Finish your sentence. "
At 130 seconds
pleted.
I
will say
Make sure your
"Turn
last sentence is
to the next
page,"
com-
You should be-
gin on the next page immediately.
You should write on
to turn
each page until
not go back to
Do
I
the topics in the order of their appearance.
give the signal.
Once you have turned
it.
not turn this page until you are given the signal
Please raise your hand
if
you have any questions.
.
a page,
Wait
do
94
Items for the Paragraph Completion Test (PCT):
When preparing
for an
exam
--- (this
is a
non-scored,
warmup
Rules - - -
When
I
am
criticized---
When
I
am
in
When
others criticize
doubt- --
me
it
usually means---
Confusion---
Items for the "This
I
Believe" Test (TIB):
American way
This
I
believe about the
*This
I
believe about marriage.
*This
I
believe about people.
This
1
believe about religion.
*This
I
believe about compromise.
''This
I
believe about friendship.
*This
I
believe about sin.
This
I
believe about foreign aid.
This
I
believe about capital punishment.
This
I
believe about world government.
of life.
Items marked with an asterisk were those used
Interpersonal scores (TIB-I and TIB-IF).
to
determine the
item)
95
Sample Page
When preparing
Try
for an
exam,
to write at least 3 sentences,
Do not turn
this
page until you are given the signal.
(
96
)
1
REFERENCES
Allard, M.
& Carlson, E. R, The generality of cognitive complexity.
Journal of Social Psychology, 1963, 5_9 73-75.
,
,
Bieri, J. Cognitive complexity - simplicity and predictive behavior.
of Abnormal and Social Psychology 1955, 5
263-268.
Journal
,
,
Complexity- simplicity as a personality variable in cognitive and
preferential behavior. In D. W. Fiske &- S. R. Maddi (Eds. ), F unctions
of varied experience
Homewood, 111.
The Dorsey Press, 1961.
Pp. 355-379.
Bieri, J.
.
:
Bieri, J. Cognitive complexity and personality development. In O, J. FlarNew York: Springer.
vey (Ed. ), Experience, structure and adaptability
1966. Pp. 13-37.
.
Bieri, J.
& Blacker, E. The generality of cognitive complexity in the perception of people and inkblots. Journal of Abnormal and Social Psychol,
ogy
,
1956,
5_3,
112-117.
Box, G. E. P. Some theorems on quadratic forms applied in the study of
analysis of variance problems, I. Effect of inequality of variance in the
one-way classification. Annals of Mathematical Statistics 1954, 25
290-302. (a)
,
,
Box, G. E. P. Some theorems on quadratic forms applied in the study of
analysis of variance problems, II. Effects of inequality of variance and
of correlation between errors in the two-way classification. Annals of
Mathematical Statistics 1954, _25, 484-498. (b)
,
Cognitive complexity: Construct descriptive terms vs. cogniPsychological Reports 1965, j_6, 1 33-134.
tive processes.
Carr,
J.
E.
,
Clayton, M. B.
& Jackson, D. N. Equivalence range, acquiescence, and
overgeneralization. Educational and Psychological Measurement, 1961,
21, 371-382.
,
98
Gardiner, G. S. Some correlates of cognitive complexity.
master's thesis, University of Alberta, 1968.
Unpublished
Gardner, R. W. Cognitive styles in categorizing behavior.
Personality 1953, 22,, 214-233.
Journal of
,
Gardner, R. W.
& Schoen, R. A, Differentiation and abstraction in concept formation. Psychological
Monographs, 1962, _76, No. 41 (Whole
"
~
No. 5 60).
,
'
’
Harvey, O. J. Scoring manual for the This I Believe Test.
University of Colorado, undated (mimeo),
Harvey, O.
vey (Ed.
J.
),
Boulder:
System structure, flexibility and creativity. In O. J. HarExperience, structure and adaptability. New York: Springer,
Pp. 39-65.
1966.
Harvey, O.
Hunt, D. E.
J.
personality organization.
,
,
& Schroder, H. M.
New York:
Conceptual systems and
Wiley, 1961.
& Schroder, H. M. Cognitive aspects of self and motivaHarvey (Ed. ), Motivation and social interaction New
Ronald Press, 1963. Pp. 95-133.
Harvey, O.
J.
In O.
tion.
York:
,
J.
.
Hess, H. F. Additional evidence on the generality of cognitive complexity.
Perceptual and Motor Skills 1966, 23 857-858.
,
,
Kelly, G. A. The psychology of personal constructs
1955, 2 vols.
Messick,
S.
,
& Kogan, N.
object- sorting measures
1963, 16 47-51.
Skills
.
New
York:
Norton,
Differentiation and compartmentalization in
Perceptual and Motor
of categorizing style.
,
,
Pettigrew, T. F. The measurement and correlates of category width as a
cognitive variable. Journal of Personality 1958, 2_6, 532-544.
,
Rozeboom, W. W. Foundations of
The Dorsey Press, 1966.
111.
the theory of prediction.
Homewood,
:
Conceptual complexity and personality organization.
Princeton University, 1969 (mimeo).
J.
Schroder, H. M.
Princeton, N.
Schroder, H. M.
cessing.
New
:
,
Streufert, S. Human information proDriver, M. J.
York: Holt, Rinehart & Winston, 1967.
,
6
99
Schroder, H. M.
& Harvey, O. J. Conceptual organization and group
structure. In O. J. Harvey (Ed. ), Motivation and social interaction- cognitive determinants
New York: Ronald Press, 1963. Pp. 134-166.
,
.
Scott,
W. A.
1962,
_25,
Cognitive complexity and cognitive flexibility.
405-414.
Sociometry,
W. A. Brief report: Measures of cognitive structure- Multivariate
Behavioral Research, 1966, l_, 391-395.
Scott,
W. A.
The structure
Colorado, 1967 (mimeo).
Scott,
of natural cognitions.
W. A. Structure of natural cognitions.
Social Psychology 1969, J_2, 261-278.
Scott,
Boulder:
University of
Journal of Personality and
,
Sloane, H. N.
Gorlow, L.
& Jackson, D. N„ Cognitive styles in equiva389-404.
lence range. Perceptual and Motor Skills 1963, 1
,
,
,
Todd, F.
J.
,
A
& Rappoport, L.
cognitive structure approach to person
perception: A comparison of two models.
Social Psychology 1964, 68 469-478.
,
,
Journal of Abnormal and
,
A
study of the generality of cognitive complexity as a personality construct. (Doctoral dissertation, University of Illinois) Ann
Arbor, Mich.: University Microfilms, 1964. No. 65-3689.
Vannoy,
J.
S.
Generality of cognitive complexity- -Simplicity as a personJournal of Personality and Social Psychology 1965, _2,
ality construct.
385-396.
Vannoy,
J.
S.
,
Winer, B.
J.
New York:
Statistical principles in experimental design.
McGraw-Hill, 1962.
Wolfe, R.
The role
of conceptual
systems
in cognitive functioning at
Journal of Personality
ing levels of age and intelligence.
,
1963,
vary-
31
,
108-123.
The process of cognitive tuning in communication.
Social Psychology I960, 6_1_, 159-167.
mal
and
Abnor
Zajonc, R. B.
of
,
Journal
9
BIOGRAPHY
Name:
Gary Bruce Cox.
Born:
November
Married:
Carol Joan Hall, 1966.
Educational History:
Public Schools, Bend, Oregon,
6,
1
93
3
in
Bend, Oregon.
1945-1957.
Willamette University, Salem, Oregon,
1957-1959.
University of Illinois,
1959
—
1960.
University of Oregon, 1961-1962 (B.S.).
Duke University, 1963-1969.
Professional Experience:
Research Assistant, Mental Health Research
Institute,
Ft. Steilacoom, Washington,
1962-1963.
USPHS
VA
Traineeship, Duke University, 1963-64.
Traineeship, 1964-65.
USPHS
Traineeship,
USPHS Predoctoral
1965-66.
Fellowship,
1966-67.
Assistant Professor, Johnson C. Smith University, 1967-70.
(
100
)
t
D00553035