9 v— .—
g . ¢ ' I"—-‘.“ 0---...“ "s- .. wow—w”. .fl‘qu‘VM‘VC-c CW. 0
0 O > . ‘

 

 

 

. .

» GNITIVE COMPLEXITY AND' THE. STIMULUS _. . ‘ ' ; _- ~

, PERSON JUDGED: A 'R‘E-DMMINATI'ON or
~ VALENCE AND FREQUENCY or m

 

 

' , Thesis for the Degree'of M A. ' . -' '- . '- _ _
' MICHIGAN 8 TE UNIVERSITY- - ‘ .. ~

. IRENE IMANNé I ‘_ I _ - ..-

    

A .
.
.h"
d.
I
.
.
.
.
.
O
.
.
.
I
.
.

a!)

 

\
I

 

>
'.

I. 'l

O

I
~ .1

o

.

.
I
.
I
I
o

o

.
.. .
I’vll't
.X iv'
I
a ‘
.
.
I
I
I
. .
.
D
.
.
.
O
.
.
' I
.
.
. . .
O
I. O

‘~ ' _ _ . _, ‘.
- o

   

I
I
S
C

 

aqggsgsghfls’fi

"t0
0
.
.

n
l
o
I
0
c
I
l

  

.
. -
o
.
a
- .
M
I
.
\
o
\

 

 

_, I
.
o
.
.

 

 

0'1“‘O.'

 

.

{'tstigflsmh.

  
          

 

 

I'-
l
o
a
c
o
0..
u
o
-

Mi!!!

 

 

 

     

 

     

. _ . .. v
, . l‘ r.
.. -‘. '- - ’-. .u.
‘ . . - .p.~
.. t . . . . . , . . .--
un-v . . - ‘c . . _ ~""
. . . .- . v. . . . _ “'f"
.-o t .g , o . ' '
C 0-. o
. - 0.. ‘a ’
- .. _- r
. . . ' . _ ' .n —.
- . .. ’ ‘
. ' ..- -‘~.
. . ., . -. -—‘-’r.
' _ ' ’ I: o . _ ‘,- .
I - . _ — -- - A
v . . ‘ . ' cg... . . ... rook .- ,. ’
. . , - on . . n.n .4- ... .‘. u....-. .n“ . O-C-‘Has'k‘ \“- ,‘.'.l.'2._.‘.'-'.!;-vau"

M IIIIIIIIIIIIIIIIIIIIIIIIIIII

WWW \HWWHWMWWW \zmi. In 3.,

3 1293 10600 4116 .. ,.‘
n K 0’“: .’ \ :4 . ‘~ '1.
3 g , . é
.. ‘fn ‘ .
:3 z. «.1... ”J...- . It: "‘_ I.“- :3} ‘ [i
. 3
v.21,” 9": . a - 7‘

£8“??- " ‘ ' ' ' - .

’1‘? 3 '49.. I . .

 

 

 

A"??? 0 9 it: ;-:1

ABSTRACT
COGNITIVE COMPLEXITY AND THE STIMULUS PERSON JUDGED:
A RE-EXAMINATION OF AFFECTIVE VALENCE
AND FREQUENCY OF INTERACTION
BY

Irene T. Mann

This study re-examined research regarding complexity of response
as a function of the stimulus person judged. More specifically, it
dealt with the vigilance hypothesis, the frequency of interaction hypo-
thesis and the neutral affect hypothesis, which have made differential
predictions regarding complexity of response. The study attempted to
show that results supporting the three interpretations were related
to the measure of cognitive complexity employed. Accordingly, two
different measures of complexity were used and also two sets of role
figures, yielding four groups. It was hypothesized that the vigilance
effect would occur only when evaluative traits were used in adminis-
tering the Rep Test measure of complexity; in using nonevaluative
traits on the Rep Test and another measure of complexity, H/Hmax, no
such vigilance effect would be found. Assumptions regarding the
affective valences of the ten role figures conventionally presented
in the Rep Test were also examined. Finally, given that the vigilance
effect was found to be measure-specific, this study examined supportive

evidence for the frequency of interaction hypothesis and the neutral

Irene T. Mann

affect hypothesis. Four groups of 30 participants each were tested.
Participants in Groups 1 and 2 responded to Bieri's ten role figures,
and the Rep Test and the H/Hmax measure of complexity, respectively,
were administered. Participants in Groups 3 and 4 responded to six
role descriptions that varied in terms of affective valence and know-
ledge of the person; the Rep Test and the H/Hmax measure of complexity,
respectively, were administered. Results generally supported previous
assumptions regarding the affective valence of the ten role figures;
other predictions were also confirmed. Results for the Rep Test using
evaluative traits supported the vigilance hypothesis. For the Rep Test
using nonevaluative traits, no strong pattern of response emerged.
Results for the H/Hmax measure of complexity supported the frequency
of interaction hypothesis. Well known persons were more highly dif-
ferentiated. The data indicated no support for the neutral affect
interpretation. However, great variability in the strength and dir-
ection of relationships between complexity and the variables of main
interest, liking and knowledge of the person, for the separate role
figures and individual participants was noted; thus, overall trends
were quite weak. Results in this area of research depend on the mea-
sures used, masking to a great extent variance which is not accounted
for by the variables chosen for study. The domain specificity notion
was briefly discussed. However, this notion did not fully account for
a subsequent analysis which indicated low generality of complexity of
response for individual participants and across role figures. A more
exact definition of complexity of response in this context needs to be

developed.

COGNITIVE COMPLEXITY AND THE STIMULUS PERSON JUDGED:
A RE-EXAMINATION OF AFFECTIVE VALENCE

AND FREQUENCY OF INTERACTION

BY
h u“ U" 0‘
Irene T. Mann

A THESIS
Subndtted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
MASTER OF ARTS
Department of Psychology

1976

ACKNOWLEDGMENTS

I wish to gratefully acknowledge those persons who participated
in my study, for without them it would not have been possible.

I would also like to express my deep appreciation to Dr. Barbara
Riemer, Dr. James Phillips, and Dr. Ellen Strommen for their guidance
and support.

Special thanks also go to Bill Brown for computer assistance and
Sue Weesner for secretarial assistance.

Lastly, my mother, my brother Joe, and my friend Deb deserve

my deepest gratitude for their loving support.

ii

TABLE OF CONTENTS

LIST OF TABLES . . . . . . . . . . . . . .
LIST OF APPENDICES . . . . . . . . . . . .
CHAPTER I 0 INTRODUCTION 0 O O C O O O O 0

Two Measures of Complexity . . . . .
Comparison of the Two Measures . . .
Vigilance-Survival Hypothesis . . . .
Justification Hypothesis . . . . . .
Frequency of Interaction Hypothesis .
Neutral Affect Hypothesis . . . . . .
The Rep Test: Methodological Issues
Aim of the Present Study . . . . . .

CHAPTER II. METHOD . . . . . . . . . . .

Participants . . . . . . . . . . . .
Materials . . . . . . . . . . . . . .
Procedure . . . . . . . . . . . . . .
Data Preparation . . . . . . . . . .
Data Analysis . . . . . . . . . . . .

CHAPTER III. RESULTS . . . . . . . . . .

Group 1 . . . . . . . . . . . . . . .
Group 2 . . . . . . . . . . . . . . .
Group 3 . . . . . . . . . . . . . . .
Group 4 . . . . . . . .'. . . . . . .
Groups 3 and 4 Correlations . . . . .
Overall Comparison . . . . . . . . .

CHAPTER IV. DISCUSSION . . . . . . . . .
Interpretation . . . . . . . . . . .
Implications . . . . . . . . . . . .
smary o o o o o o u o o o o o o o 0

LIST OF REFERENCES . . . . . . . . . . . .

APPENDICES O O O O I O O O O O O O O O O 0

iii

Page

iv

10
ll
14
16
20

25

25
25
27
28
29

31

31
37
44
45
46
47

48
48
56
56
59

62

LIST OF TABLES

Table Page
1 Mean Ratings and Scores for Group l Role Figures . . . 33
2 Stepwise Regression Analysis for Group 1 . . . . . . . 34
3 Correlations Among Ratings and Complexity Scores for

Group 1 Role Figures . . . . . . . . . . . . . . . . . 38
4 Mean Ratings and Scores for Group 2 Role Figures . . . 4O
5 Stepwise Regression Analysis for Group 2 . . . . . . . 41
6 Correlations Among Ratings and Complexity Score .

for Group 2 Role Figures . . . . . . . . . . . . . . . 43
7 Group 3. Means and Standard Deviations of Complexity

Scores ' Evaluative Traits . . . . . . . . . . . . . . 45
8 Group 3. Means and Standard Deviations of Complexity

Scores - Nonevaluative Traits . . . . . . . . . . . . 46
9 Group 4. Means and Standard Deviations of H/Hmax

Scores . . . . . . . . . . . . . . . . . . . . . . . . 47

iv

LIST OF APPENDICES

Appendix Page
A PROVIDED BIPOLAR CONSTRUCTS ON THE REP TEST . . . . 62
B SAMPLE PAGES AND SCORING GRIDFORM FOR THE REP TEST

GROUPS 1 AND 3 O I O O O O O O O O O O O O I O O O 6 3
C SAMPLE SHEETS AND INSTRUCTIONS--GROUPS 2 AND 4 . . 66
D CORRELATIONS FOR INDIVIDUAL PARTICIPANTS IN GROUP 1 68
E ADDITIONAL DATA FOR GROUP 2 (BIERI'S TEN ROLE

FIGURES AND H/Hmax MEASURE) . . . . . . . . . . . . 69
F ADDITIONAL DATA FOR GROUP 3 (REICH'S SIX ROLE ,

FIGURES AND REP TEST) . . . . . . . . . . . . . . . 72
G ADDITIONAL DATA FOR GROUP 4 (REICH'S SIX ROLE

FIGURES AND H/Hmax MEASURE) . . . . . . . . . . . . 74

H OVERALL CORRELATIONS FOR EACH GROUP . . . . . . . . 78

CHAPTER I
INTRODUCTION

A great deal of human activity is spent in getting to know other
persons and thinking about them. The area of person perception or
social cognition analyzes such activity.

Research in this area has generally focused on the "processes
by which man comes to know and think about other persons, their
characteristics, qualities and inner states,“ [Tagiuri, 1969, p. 429]
and more specifically on the characteristics of the perceiver that
might affect such processes. One important characteristic of the per-
ceiver is degree of differentiation or the "tendency to make fine
distinctions among people and thus to perceive them as different
from one another" [Shrauger and Altrocchi, 1964, p. 292]. Bieri
[1955] defined differentiation as the number of constructs or attri-
butes constituting a given cognitive structure. The cognitive struc-
ture or domain of interest was the interpersonal construct system.

The cognitively complex person utilizes a more highly differentiated
construct system than a cognitively simple person. It has been shown
that cognitively complex persons and cognitively simple persons per-
form differently when given impression formation tasks [Nidorf and
Crockett,l965; Mayo and Crockett, 1964; Shrauger, 1967], in discrimin-
ating behavioral stimuli and making judgments [Bieri, Atkins, Leaman,
Miller and Tripodi, 1966], and in predicting how others will respond

[Shrauger and Altrocchi, 1964].

Complexity has not only received consideration as a personality
variable; researchers have also examined complexity of response to
various stimuli. Many researchers have focused on complexity of re-
sponse to stimulus persons differing in affective valence. This work
has been approached in the following ways: 1) in person perception, a
concern with the differentiation skills of the perceiver, and the per-
ceiver's use of personal constructs; 2) cognitive complexity as a
characteristic of cognitive structure and its generality across and
within domains; 3) perceptual defense and vigilance applied to the
perception of persons; 4) affective arousal and its effect on discrimin-
ation of persons. A fifth approach is information theory. Researchers
studying cognitive structure have adopted measures of amount of infor-
mation to assess dimensionality in terms of distinctions yielded by a
system of groupingscnrcategories [Scott, 1962, 1969].

The first approach was briefly introduced. In regard to the
second approach, generality of cognitive complexity has been the major
issue. Crockett [1965] reviewed research offering some support for
the following hypotheses: l) complexity in one domain does not neces-
sarily imply complexity in other domains; 2) persons for whom inter-
personal relations are functionally important have more complex systems
than individuals for whom interpersonal relations are less important;

3) an individual may show differential complexity with respect to dif-
ferent categories of persons depending on the extent of interaction with
them. The latter notion underlies the frequency of interaction hypothesis,
which indicates that one is more complex in responding to well known

and liked persons since these persons are socially close and more

familiar.

Crockettfisconclusions contradicted the findings of Miller and
Bieri [1965] and Irwin, Tripodi and Bieri [1967], who found that par-
ticipants responded to negative role figures in a more complex
manner. Their studies displayed a concern with perceptual effects
and discrimination. The vigilance hypothesis, called upon to explain
greater differentiation of negative role figures, was borrowed from
notions of perceptual defense and vigilance in the area of perception
[Erdelyi, 1974], and applied to person perception, offering a ready-
made and perhaps counter-intuitive explanation for obtained results.

Research representing the fourth approach has dealt with affective
arousal and how this affects judgments and discriminability of cues
[Bieri, 1967]. One view holds that an increase in affective arousal
decreases the individual's discrimination of stimuli. This view
served as the basis for yet another hypothesis, the neutral affect
hypothesis, which makes predictions contrary to both the vigilance
hypothesis and the frequency of interaction hypothesis. The neutral
affect hypothesis suggests that persons of neutral affect will be
perceived in a more complex manner. Liked or disliked persons are
more arousing affectively and are less well differentiated or
discriminated.

The three hypotheses mentioned, the vigilance hypothesis, the
frequency of interaction hypothesis and the neutral affect hypothesis,
posit different processes underlying responses to stimulus persons
when responses are examined in terms of degree of differentiation or
complexity. They also make differential predictions regarding which

stimulus persons will be more highly differentiated. In one case,

negative persons will be more highly differentiated, in another case
positive persons, and in another, persons of neutral valence. All
three hypotheses have found some support. The hypotheses were de—
rived from different approaches, as outlined above. There have also
been different approaches to the measurement and definition of come
plexity in this context.

This problem of defining and measuring complexity is a major
concern for the research area. Bieri [1961] pointed out that it is
unclear whether cognitive complexity is a differentiation concept
exclusively or relates to organizational properties of the cogni-
tive system as well. A factor analysis c>f several widely used mea-
sures of complexity in the interpersonal domain and several related.
measures by Vannoy [1965], revealed that cognitive complexity is not
a unitary trait as measured by the various paper and pencil tests.
vannoy contended that existing measures tap different variables or
aspects of complexity; these aspects have not as yet been clearly
defined.

Given this state of affairs, one must proceed with extreme caur
tion in this area; an imprecise definition of complexity permits dif-
ferent approaches to its measurement. The Role Construct Repetory
Test (RCRT or Rep Test), as outlined by Kelly [1955], and subsequently
modified by Tripodi and Bieri [1963] and Bieri, et a1. [1966], has
been one of the most widely used measures of complexity. In addition
to providing a method to assess an individual's cognitive complexity
in the interpersonal realm, the Rep Test has been used to measure

complexity of response to role figures differing in affective valence.

The work of Scott [1962] represents a slightly different approach to
the measurement of complexity than that of the Rep Test. The approach
was derived from information theory [Attneave, 1959]. The partici-
pant sorts a number of items; an index, based on the dispersion of
objects over a set of distinctions yielded by the category system, is
then computed. Reich [1969] utilized such a task in assessing come
plexity of response to persons differing in affective valence.

Results of studies in this area do not overwhelmingly confirm
any one of the hypotheses. It is extremely difficult to determine
the validity of the processes posited because researchers have not
maintained comparability of method and complexity measures. There
appears to be some relation between the measure and the results sup-
porting a particular hypothesis, such that certain measures consis-
tently yield certain results. The aim of the present study was to
confirm this. To this end, two different measures of complexity were
examined, Bieri's Role Construct Repetory Test and information theory's
index of relative entropy of H/Hmax. The two measures will be briefly
discussed; then research findings relevant to them as dependent mea-
sures in examining complexity of response to stimulus persons will be

presented.

Two Measures of Complexity

 

Rep Test
The Rep Test is based on Kelly's theory of personality [1955]
which assumed that every person has a number of personal constructs
to cognize and perceive others. A construct is defined as a dimension

for construing the way in which persons are alike and different.

Examples of bipolar constructs are "considerate-inconsiderate" and
"dominant—submissive." The Rep Test taps the individual's system of
constructs, and then requires the participant to judge a number of
persons on these dimensions. When constructs are provided for the
participant it is assumed that they are representative of the per-
son's own constructs [Bieri, et al., 1966]. Tripodi and Bieri [1963]
reported that providing constructs yielded results similar to results
found when the procedure involved use of the participants' own
constructs.

Bieri, et al. [1966] described the scoring of the Rep Test.
Participants are presented with a 10 X 10 grid. Each of the ten
columns is identified by a certain role figure. Ten rows of bipolar
constructs are provided. The participant uses a six-point Likert
scale to rate the ten role figures on the first construct, then rates
all ten on the second construct, and so on, for a total of 100 ratings.
Complexity is measured by comparing each rating in a row with the
rating directly below it for the same person and thereon for each
rating in the column, yielding 45 comparisons and for the entire grid,
450 comparisons. A score of one is given for exact agreement of rat-
ings on any one person. Thus, the highest possible score is 450,
indicating simplicity or minimal differentiation. The role figures
commonly used are: Self, Mother, Father, Friend of the Same Sex,
Friend of the Opposite Sex, Boss, Person You Dislike, Person You'd
Like to Help, Person With Whom You Feel Most Uncomfortable and Person

Difficult to Understand.

H/Hmax

This measure of complexity was derived from information theory
[Attneave, 1959]. Participants list and group adjectives describing
others. The number of adjectives written and the number of categories
created are counted. For example, one might write seven adjectives
(jealous, stingey, realistic, thoughtful, modest, kind, considerate)
and place the first three in one group and the last four in another
group. A measure of dispersion, H, of the items over the group com—
binations is calculated. The ratio, R, or relative entropy is obtained
by expressing H in ratio to the maximum H possible with a certain
number of categories (i.e., Hmax is obtained when all categories have
equal frequencies). The ratio R or H/Hmax varies between 0.00 and
+1.00, regardless of the number of categories [Scott, 1962]. As
H/Hmax approaches +1.00, complexity of response is greater. H/Hmax
corrects for verbal fluency; the measure is most useful when either
the number of stimuli or the number of categories varies across parti-
cipants. If a person creates a large number of categories and dis-
tributes the items equally over all categories, there is a high degree
of uncertainty and a low degree of structure, hence more ambiguity
and greater complexity of response.

H/Hmax is a measure of categorization. It assesses departure
from distributional equality and is not concerned with the content of
the response; it taps a purely structural property [Scott, 1962]. If
items are distributed equally over the categories, the amount of infor-
mation yielded is low, and it is uncertain, given any one item, which

category it belongs in. This uncertainty connotes greater ambiguity

and hence complexity of response. Glixman [1965] used H/Hmax as a
measure of the degree of structure in a domain. Scott [1966] spoke
of H as a measure of cognitive dimensionality; "the number of groups-
worth of information can be represented as the dispersion of the.
objects over the set of distinctions yielded by the category system"

[p. 408].

Comparison of the Two Measures

The two measures appear to require different types of responses
from1participants. The Rep Test requires the individual to use con-
struct dimensions to discriminate among persons. H/Hmax has little
to do with how construction dimensions are used to discriminate among.
others. With the H/Hmax measure, once the participant has written
adjectives to describe a certain person, the adjectives become the
stimuli. Although the individual may have the stimulus person in
mind when creating the groupings, it might also be the case that the
participant responds solely to the adjectives. The adjectives for
each role figure are grouped separately by the participant, so that
each role figure serves as a separate domain. This does not serve to
assess how well the individual is discriminating or differentiating
gmggq_others. Although Reich [1969] states that "discrimination in
cognitive research typically is regarded as being the process in whidh
the person uses a set of independent dimensions or characteristics to
respond to stimuli, responding differently to the different classes
but similarly to those stimuli within each class," [p. 107] it is not

exactly clear how H/Hmax measures this. Irwin, Tripodi and Bieri [1967]

pointed out that "such a task [written descriptions of persons] makes
no assessment of how functionally different the constructs may be
inasmuch as the subject is not required to actually use these dimen-
sions to discriminate between individuals," [p. 446] which is what
the Rep Test procedure purportedly accomplishes.

The lack of relationship between various measures of complexity
leads one to suspect that the measures are not convergent. Miller
[1969] found nonsignificant correlations of .009 for males and -.113
for females between Bieri's and Crockett's procedures. Wilkins and
Epting [1974] reported a correlation of -.03. Vannoy [1965] reported
a correlation of .06 between the Rep Test and number of categories
used to group 12 significant persons. The so-called "measures of
complexity" appear to tap different aspects of what might be called

complexity.

Vigilance-Survival Hypothesis

Miller and Bieri [1965], using the Rep Test, found that parti-
cipants were less complex in judging persons who were close to them
and toward whom they were expected to relate positively (Self, Mother
Father, Friend of the Same Sex, Friend of the Opposite Sex) than more
socially distant persons toward whom the participant was expected to
relate negatively (Persons You Dislike, Person You'd Like to Help,
Boss, Person With Whom You Feel Most Uncomfortable, Person Difficult
to Understand). A "vigilance" hypothesis was advanced to explain these
results, i.e., differentiation serves an adaptive function for antici-
pating and identifying the behavior of more remote and threatening

persons. Differentiation helps to clarify the nature of the threat.

1b

Irwin, Tripodi and Bieri [1967] obtained similar results using
the same technique to measure differentiation and two different sets
of stimuli. In Study I participants provided names for four liked
housemates (same sex), four housemates of neutral valence (same
sex), and four disliked housemates (same sex). In Study II partici-
pants provided names for eight role categories: closest friend (male
and female), person you admire (male and female), person you find
hard to like (male and female) and person with whom you feel most
uncomfortable (male and female). Positive figures were differentiated
significantly less well than negative figures. Also, positive figures
were differentiated less well than neutral figures and neutral figures
were less well differentiated than negative figures.

Results supporting the vigilance hypothesis were also reported by
Miller [1968] and Wilkins, Epting and VanDeRiet [1972]. In both
studies the stimuli were the ten role figures presented by Miller and
Bieri [1965]. Soucar and DuCette [1971] found greater differentiation
of disliked political figures than liked political figures, and Soucar
[1970] reported higher complexity scores among high school students

for disliked teachers than for liked teachers.

Justification Hypothesis
A justification hypothesis was advanced by Koenig [1971] which
makes the same prediction as does the vigilance hypothesis, i.e., nega-
tive persons will be more highly differentiated than positive persons.
However, the justification hypothesis posits a different underlying
mechanism for this effect. According to Koenig, there is "a trait in

our culture and perhaps others as well, which stresses being friendly,

11

and liking other people" [p. 385]. Dislike for someone requires
justification; one should be able to provide good reasons for dislike.
This implies that negative persons will be more highly differentiated
and discriminated, so that, if required, one can justify one's dis-
like. Since justification for liking is not expected, positive per-
sons would not be as highly differentiated as negative persons.

Soucar and DuCette [1972] demonstrated that differentiation can
be manipulated in an experimental session by asking participants to
defend their choices of liked and disliked persons. When participants
were asked to justify their choice of a liked person, differentiation
(as assessed by the Rep Test) of these persons increased. Results
were less clearcut when participants were asked to justify their
choice of disliked persons, primarily because of problems with the
scale used. The authors concluded that complexity may not be influr
enced by affect per se, but may be influenced by committing oneself
to a statement concerning like or dislike; the threat of postdecisional
justification increases complexity of response.

Finally, Koenig and Seaman [1974b] examined both the vigilance
and justification effects. The authors concluded that they were un-

related, but both contributed to increased complexity of response.

Frequency of Interaction Hypothesis
Contrary to results supporting a vigilance or justification inter-
pretation, Supnick [1964; as reported by Crockett, 1965] found that
participants used more constructs to describe persons they liked than

to describe persons they disliked and to describe peers than to describe

12

older persons. She posited a"frequency of interaction" hypothesis to
explain the results, i.e., participants differentiated more highly
the persons they knew well or liked than less well known or disliked
persons because they came into more contact with the former persons
and avoided the latter persons. Her results<iuinot necessarily dis-
confirm the justification of vigilance hypothesis; the method of
assessing complexity of response in Supnick's study differed markedly
from that of the Rep Test. Participants simply wrote three minute
descriptions of the stimulus persons, and the number of constructs
used in each description was determined. Thus, it seems reasonable
that participants would describe more fully and in greater detail
well known or liked persons. Crockett [1965] cited Supnick's study.
to support his contention that although some generality of complexity
exists in the interpersonal realm, characteristics of the stimulus
person do affect complexity of responding.

Scott [1962] indicated that knowledge of a domain is a necessary,
though not sufficient basis for cognitive differentiation of a domain.
One would expect, on the average, some positive correlation between
complexity and level of information or knowledge regarding a given
domain. Two studies confirmed this. Participants were asked to group
nations and answer factual questions regarding the nations. The infor-
mation tests correlated +.3l and +.37 with H, the measure of complexity.
Scott [1969] also reported studies in which participants were exposed
to instructions about a domain. Dimensionality scores showed a signi-
ficant mean increase from beginning to end of the instructional course.

Seferi [1968; as reported by Scott, 1969] found that intimately

13

known persons were differentiated more highly than casually known
person. Intimacy was assumed to correlate with amount of information
regarding persons, but this assumption was not directly assessed.

Reich [1969] found some support for the frequency of interac-
tion hypothesis. Participants listed adjectives describing different
persons and then grouped the adjectives for each person. The number
of adjectives used, the number of categories and H/Hmax were the three
dependent measures. Participants wrote more adjectives and created
more categories for persons they knew well than for persons they knew
only slightly. The most adjectives were written and the most cate-
gories created for positive persons, followed by negative and neutral
persons.

Lott, Lott, Reed and Crow [1970] examined participants' usage
of the 200 most meaningful words from Anderson's list of 555 person-
ality trait words in describing well known persons who were liked,
disliked or of neutral valence. The greatest number of words were
used to describe liked persons, fewer for disliked persons and fewest
for neutral persons. Liked persons were also described by more favor-
able words than neutral or disliked persons, and disliked persons
were described by less favorable words than neutral persons.

In studies supporting the frequency of interaction hypothesis,
the method of assessing complexity differed markedly from the Rep
Test and involved the rather simple procedure of counting numbers of
adjectives, constructs or categories. Such counts were undoubtedly
affected by verbal fluency. More importantly, the tendency to use

positively evaluated words more fequently and facilely or the Pollyanna

l4

effect [Boucher and Osgood, 1969] probably contributed to higher counts
of constructs and adjectives in descriptions of well known and liked
persons, assuming that such persons were described in favorable terms.
However, when there is a correction for verbal fluency and hence the

positivity bias, another effect is noted.

Neutral Affect Hypothesis

 

Reich [1969] asked participants to respond to six person descrip-
tions: Person You Know Well and Like, Know Well and Dislike, Know
Well and Feel Neutral Toward, Know Slightly and Like, Know Slightly
and Dislike, and Know Slightly and Feel Neutral Toward. The three
dependent measures were number of adjectives used to describe each
person, number of categories and H/Hmax. For the first two measures,
results supported the frequency of interaction hypothesis. On the
last measure, H/Hmax, there was a main effect for Affect. Neutral
persons received the highest complexity scores, followed by positive
persons and negative persons, respectively. Reich hypothesized that
intensity of involvement was the crucial factor; the higher or more
intense the involvement, the lower the discrimination of stimuli.

Glixman's [1965] study also lent support to the neutral affect
hypothesis. He found that participants differentiated more highly
among the attributes of inanimate objects than among the character-
istics of war and self. The three domains were ordered from peri-
pheral to personally relevant. Participants were given statements
regarding objects, war and self and were asked to group the statements.
H/Hmax was then computed. The author found that the largest number

of categories was associated with the least significant domain, objects,

15

and the most unequal distribution of items over the categories used
wmsassociated with the most significant domain, self.

Harvey, Reich and Wyer [1968] also obtained results consistent
with the view that high involvement results in poorer differentiation
than does low involvement. Moreover, they examined the interaction
of involvement with the variable concreteness-abstractness. Two
groups of participants, concrete and abstract (as assessed by the
"This I Believe" test), selected attributes to describe social beliefs
and interpersonal relations, and then rated selected social beliefs
and interpersonal relations (8 positive, 8 negative, and 8 neutral
for each) on the selected attributes. Participants also rated inten-
sity of feeling toward each social belief and interpersonal relation.
Concrete subjects achieved higher differentiation than abstract par-
ticipants when attitudes were of low intensity; abstract participants
differentiated equally well regardless of intensity of attitude.
Overall, neutral domains were differentiated more highly than either
positive or negative domains. Differentiation of neutral domains was
unaffected by intensity of attitudes, while positive and negative
domains were differentiated more highly when attitudes were low in
intensity than when high in intensity. The intensity of ratings of
positive and negative domains did not differ significantly but ratings
for neutral domains differed significantly from both positive and
negative.

Lott, Lott, Reed and Crow [1970] hypothesized that participants
would use the greatest number of words from Anderson's list of person-

ality-trait words to describe liked persons, fewer words for disliked

16

persons and fewest for neutral persons. This hypothesis was based on
the assumed cue properties of liked, disliked and neutral persons.

It was assumed that liked persons would be maximally attention dir-
ecting and salient. A disliked person would be less salient than a
liked person but more salient than a neutral person because of a vigi-
lant reaction to disliked stimuli. The authors found support for
their notions in regard to cue properties. In reconciling their
results with results supporting the vigilance effect the authors
indicated the perhaps liked persons are perceived as very similar

and more trait words can be applied to them, leading to less discrim-
ination among such persons. Disliked persons, as a group, are des-
cribed with fewer characteristics and hence would be more easily
distinguishable from one another. Their findings also suggested

that liked persons as a group and disliked persons as a group have

in common certain personality traits.

Both Reich [1969] and Glixman [1965] indicated that although
participants made a greater number of responses to well known and
liked persons, the responses were highly redundant and didn't contain
much informational complexity. For the neutral person, the amount of
description was reduced, as were the number of independent discriminna-

tions, but actual complexity was high [Reich, 1969].

The Rep Test: Methodological Issues

 

The present study examined two issues regarding the Rep Test:
the affective valences of the role figures and the use of evaluative

trait words as bipolar constructs.

17

In regard to the first issue, Wilkins and Epting [1971] tested
the notion that five of the traditional Bieri role figures were posi-
tive in affective valence and five were negative in affective valence.
They factor analyzed the cognitive complexity scores for the ten role
figures for 82 participants, hoping to find two distinct factors that
would account for a large proportion of the total score variance and
support the assumption underlying the dichotomization of the ten role
figures in terms of social distance and affective valence. Factor I
accounted for only 21.7% of the total score variance and the roles
with high loadings on this factor were the "positive" ones and boss
and person you'd like to help. Factor 2 accounted for only 5.9% of
the total score variance and only two of the "negative" figures loaded
highly on this factor. The authors concluded that the role categories
could not be divided into two distinct sub-groups on the basis of
affective value. This was perhaps the wrong conclusion. The results
more accurately indicated that assumed affective valences for the role
figures were incorrect: perhaps there were subgroups, but they did
not contain the.five "positive" and five "negative" role figures as
outlined by Miller and Bieri [1965].

Two other studies lend support to this latter notion. Turner
and Tripodi [1965] examined the responses of student clinicians to
significant others and clients. For significant others (Bieri's
10 role figures), results supported the vigilance hypothesis, while for
clients, affective valence had no effect on differentiation. The
participants as a group tended to perceive clients more complexly than

the ten role figures. The authors explained these results in terms of

18

a clinician's job requirement to make fine judgments regarding clients
regardless of their affective valence. However, more importantly,
this study indicated that the particular stimuli presented made a
difference in results obtained, depending on the sample of persons
being tested.

Rothman [1973] tested therapists, counselors and counselor
educators and found no differences in differentiation of positive and
negative figures (Bieri's ten role figures). Rothman hypothesized
that for therapists, mother and father might be negative in valence
rather than positive because a difficult or unhappy childhood might
lead one to become a therapist, and boss and person you'd like to
help would be more positive in valence than negative. This was
partially found to be the case. In the case where person you'd like
to help and boss were combined with the other "positive" affect roles,
and mother and father were combined with the "negative" affect roles,
the vigilance hypothesis was supported.

It appears that for different individuals and different samples
of persons, the affective valences of the ten role figures conventionally
presented in the Rep Test are perceived differently. In this study,
participants were required to designate their degree of liking for each
of the role figures in order to test prior assumptions regarding affec-
tive valence.

The second issue in regard to the Rep Test concerns the use of
evaluative trait words. Boucher and Osgood [1969] argued "that there
is a universal human tendency to use positively evaluated words more

frequently, diversely, and facilely than evaluatively negative words"

19

[p. 1]. They called this the Pollyanna effect. The authors provided
cross-cultural evidence for this tendency.

Extending the notion, Shepherd [1972] argued that the existence
of one desirable characteristic implies the existence of other desir-
able characteristics while the existence of one undesirable trait
does not necessarily imply the existence of other undesirable traits.
Therefore, Bieri's procedure for assessing complexity would yield
higher scores for "negative" role figures than for "positive" figures.
Vannoy [1965] noted that "the tendency to use a 'favorable' response
is negatively associated with cognitive complexity as measured by
this test [Bieri's procedure]"[p. 387]. The correlation between the
number of favorable responses and complexity scores was -.46. Thus,
those participants who consistently marked the favorable traits were
those participants with scores indicating a less complex response.

Shepherd had participants complete Bieri's gridform, providing
five evaluative and five nonevaluative bipolar constructs. The evalua-
tive constructs were bipolar adjectives at opposite extremes in favor-
ability according to Andersons' [1968] list of likeableness ratings of
personality-trait words. For the other five pairs of bipolar adjec-
tives the likableness values were similar and close to the neutral
point. The 10 X 10 grid for each participant was then divided into
four matrices, comprised of ratings for positive role figures on
evaluative traits, negative figures on evaluative traits, positive
figures on nonevaluative traits and negative figures on nonevaluative
traits. Participants were more complex in responding to negative

figures than positive figures on evaluative traits, and there was no

20

difference in complexity scores for positive and negative figures on
nonevaluative traits. Thus, the Pollyanna effect may have contributed
to scores for negative role figures; however, complexity of response
to negative figures on evaluative traits was no greater than that
based on nonevaluative traits, indirectly lending some support to a
vigilance effect, over and above the Pollyanna effect.

The above has certain implications for the results one might
expect in using the Rep Test measure. In the studies cited in sup-
port of the vigilance or the justification hypothesis, the measure
of complexity was in all cases the Rep Test. And, as noted above,
Shepherd [1972] found that when nonevaluative trait words were pre-
sented instead of evaluative trait words a vigilance effect was not
found. Thus, it was hypothesized that the vigilance or justification
effect is a measure-specific one. To test this, results for both

evaluative and nonevaluative trait words on the Rep Test were examined.

Aim of the Present Study

 

Past research has indicated that the vigilance or justification
effect occurs when a particular measure of complexity is used, i.e.,
the Rep Test, when certain evaluative trait words are used, and when
certain role figures are assumed to be positive or negative in valence.
When another measure of complexity, H/Hmax, is used, results indicate
that neutral stimulus persons are responded to in a more complex
manner, supporting a neutral affect hypothesis. When quantity of
responding in describing stimulus persons is involved then evaluatively
positive words seem to be used more frequently and facilely to describe

positive persons than evaluatively negative words are used to describe

21

negative persons, supporting the frequency of interaction hypothesis.
In light of the above, the present study addressed the following
issues:

1) Affective valence of role figures

 

This study re-examined the ten role categories or figures con-
ventionally employed when administering the Rep Test, by allowing
participants to freely designate their degree of liking for persons
chosen to fit the role descriptions, and to compare these ratings
with prior assumptions regarding the affective valence of the role
figures. Ratings regarding knowledge of the person, degree of involve-
ment and frequency of interaction provided additional information.

2) Use of evaluative and nonevaluative traits on the Rep Test

 

For the Rep Test measure of complexity, results for both evalua-
tive and nonevaluative traits were examined. (See Appendix A.) It
was expected that for evaluative traits, results would be similar to
prior findings. A vigilance or justification effect would be apparent,
i.e., negative role figures would be responded to in a more complex
manner. For nonevaluative traits it was expected that the affective
valence of the role figures would not be strongly related to complexity
of response [cf. Shepherd, 1972]. Such results would indicate that
the Rep Test yielded results consistent with the vigilance or justi-
fication hypothesis because trait words with evaluative connotations
were used in a particular way. Thus, negative persons would be re-
sponded to in a more complex manner because both positive and negative
traits were assigned to them; such would not be the case for positive
persons. When nonevaluative traits were presented, however, there

was no reason to expect that the vigilance effect would occur.

22

3) Use of an alternative measure of complexity, H/Hmax

Also, a second measure of complexity was examined. It was ex-
pected that results using this measure would not correspond to results
obtained when using the Rep Test measure. The number of adjectives
written to describe stimulus persons and number of categories created
in grouping the adjectives were examined. It was expected that results
for these latter two measures would support the frequency of inter-
action hypothesis, similar to the findings of Crockett [1965] and
Reich [1969]. The prediction for H/Hmax was that results for this
measure would not support a vigilance interpretation.

Given that support for a vigilance effect was not found using
nonevaluative traits on the Rep Test or using number of adjectives,
number of categories and H/Hmax as measures of complexity, the results
for these measures were examined in terms of support for either the
frequency of interaction hypothesis or the neutral affect hypothesis.
It was hoped that results for these measures would converge in support-
ing one or the other of these hypotheses.

4) Generality of responses to two different sets of role figures

Two sets of role figures were examined in conjunction with the
two measures of complexity. It was expected that differing role
figure designations would not affect complexity of response. Most
researchers have used Bieri's ten role figures, but some have not
[Irwin, Tripodi and Bieri, 1967; Turner and Tripodi, 1965; Supnick,
1964; Reich, 1969]. Bieri's ten role figures were chosen so that
results could be compared with prior findings and so that assumptions

regarding the five positive role figures and the five negative role

23

figures could be tested. The second set of role figures, previously
used by Reich in 1969, were such that the two variables of interest,
affective valence and knowledge of the stimulus person, were varied
systematically in the role figure descriptions and allowed examination
of their possible interaction in affecting complexity of response.

Thus, this study examined responses of participants in four
different situations. In the first, participants were exposed to
Bieri's ten role figures and were given the Rep Test. For a second
group, the same ten role figures were presented and participants
responded so that it was possible to calculate H/Hmax as a measure of
complexity. Participants in a third group responded to six person
descriptions: person you know well and like, know well and dislike,‘
know slightly and like, know slightly and dislike, know well and feel
neutral toward, know slightly and feel neutral toward, and completed
the Rep Test. Finally, a fourth group responded to the same six
person descriptions as above in such a way that H/Hmax could be calcu-
lated. Participants in each group also marked questions regarding
liking, frequency of interaction, involvement and knowledge of the
stimulus persons.

In summary, four groups of participants were exposed either to
Bieri's ten role figures or Reich's six role figures and responded in
terms of the Rep Test or the H/Hmax measure of complexity, as follows:

Group 1: Bieri's ten role figures, Rep Test

Group 2: Bieri's ten role figures, H/Hmax measure of complexity

Group 3: Reich's six role figures, Rep Test

Group 4: Reich's six role figures, H/Hmax measure of complexity

24

The present study had implications for the following issues
and predictions:
1) The confirmation of the assumed affective valences of
Bieri's ten role figures. Participants in Groups 1 and 2 were asked
to indicate their degree of liking for each of Bieri's role figures.
2) Results supportive of the vigilance effect for the Rep Test
using evaluative traits but not for the Rep Test using nonevaluative
traits. Participants in Groups 1 and 3 completed the Rep Test, with
ten evaluative bipolar trait words and ten nonevaluative bipolar trait
words. Group 1 was a replication of Shepherd's [1972] procedure.
3) Results not supportive of the vigilance effect but supportive
of either the frequency of interaction hypothesis or the neutral
affect hypothesis using another measure of complexity, H/Hmax. Parti-
cipants in Groups 2 and 4 completed the H/Hmax measure of complexity.
4) No overall difference in supportive evidence for a particular
hypothesis using two different sets of role figures. Participants
in Groups 1 and 2 responded to Bieri's ten role figures; participants
in Groups 3 and-4 responded to Reich's [1969] six person descriptions.
Results for Groups 1 and 3 and Groups 2 and 4 were compared to note

differences in results due to presentation of different role figures.

CHAPTER II

METHOD

Participants

 

One hundred twenty-six introductory psychology students, 60
males and 66 females, at Michigan State University served as partici-
pants. Participants received either class credit or $1.00 in cash.
(Protocols of six female participants were not included in the analysis
because of incomplete questionnaires, N=4, or failure to complete the

grouping procedure, N=2.)

Materials
Groups 1 and 3 (Rep Test)

The Rep Test booklets consisted of a cover page on which each
participant indicated their sex, age, grade point average and year in
school, followed by an instruction page. Subsequent pages were headed
with a role figure name or description and a set of 20 rating scales
anchored by bipolar adjectives. (See Appendix B.) In assembling the
booklets, the order of presentation of the role figures was randomized.
The scales appeared in the same order for all role figures but evalua-
tive and nonevaluative traits were interspersed. For evaluative traits,
the more desirable or favorable trait did not always appear on the same
side of the six point Likert type scale. A scale was provided with
numbers from one to six. Following the scales, questions appeared
regarding liking, frequency of interaction, degree of involvement,

25

II|I [III
.I ll 'II' III.

Ill ['1‘ I'll" .[

26

knowledge of the person and relation to the participant. Participants
responded to such questions by marking a l to 10 scale with 1 indicating
high liking, frequency of interaction, involvement and knowledge, and
10 indicating very little liking, frequency of interaction, involvement
and knowledge. The question regarding interaction was omitted for
Self, Mother, and Father, and the involvement question was not asked
regarding the Self. Booklets either contained the role figures:

Self, Mother, Father, Friend of the Opposite Sex, Friend of the Same
Sex, Boss, Person You Dislike, Person Difficult to Understand, Person
You'd Like to Help and Person With Whom You Feel Most Uncomfortable,

or the role descriptions: Person You Know Well and Like, Know Well

and Dislike, Know Well and Feel Neutral Toward, Know Slightly and Like,
Know Slightly and Dislike, Know Slightly and Feel Neutral Toward.
Finally, except for Self, Mother, Father, Friend of the Opposite Sex,
Friend of the Same Sex and Boss, participants were asked to briefly

describe their relationship to the person.

Groups 2 and 4 (H/Hmax Measure)

Materials for Groups 2 and 4 included cards on which had been
printed the role figure names or descriptions, 3 X 5 blank slips of
paper and paper clips. There was also a sheet on which participants
indicated their sex, age, GPA, year in school; other sheets contained
questions regarding liking, frequency of interaction, degree of involve-
ment and knowledge of the person and a request for a brief description
of the person chosen to fit the role figure, when appropriate. (See

Appendix C.)

27

Procedure

Participants were randomly assigned to one of four groups. All
participants were tested individually in separate cubicles. Partici-
pants assigned to Groups 1 and 3 were asked to fill out a booklet.

The experimenter gave verbal instructions that participants were to
flip through the booklet and familiarize themselves with the role
figures, to be sure they could think of a different person to fit each
description. Participants were encouraged to ask questions, if any,
in working through the booklet. The experimenter left the cubicle
while participants worked. When the participant finished, the experi-
menter debriefed the participant and answered questions.

Participants assigned to Groups 2 and 4 were seated at a table-
in a cubicle with the role figure names or descriptions on cards spread
out in a row before them. The experimenter shuffled the cards before‘
testing each participant. The participant was asked to note the
descriptions and think of a person they knew to fill each description
and indicate that person's name or initials on a slip of paper placed
beneath the role figure designation. Participants were then asked to
think of one word adjectives, as many as could be thought of, to
describe each person and to write each adjective on a separate slip
of paper and pile the slips beneath the appropriate role figure desig-
nation. The experimenter left the cubicle while the participant worked.
When the participant finished, the experimenter returned and asked the
participant to return to each pile of words in turn and group the words
that seemed to go together, to label these groupings and clip the
words in the group and the label together. (See Appendix C for the

instructions given. Some participants found the instructions very

28

difficult to comprehend; the experimenter repeated and varied them so
that the participant reached an understanding of what was required.)
Again, the participant was left alone to complete this task. The
experimenter then gave the participant a sheet for each role figure
with questions regarding liking, interaction, involvement and knowledge
of the person and relation to the participant, when appropriate. Fin-

ally, the experimenter debriefed the participant and answered questions.

Data Preparation

 

Groups 1 and 3 (Rep Test)

For Groups 1 and 3 the experimenter converted participants'
ratings to a gridform for scoring purposes. (See Appendix B.) Scoring
was according to Bieri, et a1. [1966]. Evaluative and nonevaluative
traits were treated separately and a complexity score for each role
figure was computed. A high numerical score indicated highly redundant
ratings, hence a less complex or simple response. The above deviated
from the usual procedure; participants are usually presented with a
gridform to complete and are asked to consider each bipolar construct
fOr all role figures, consider the second bipolar construct, and so on.
However, the obvious advantage of the procedure here was that it was
less confusing for the participant to consider one role figure at a
time, and there was no reason to believe that such a change affected
results. Irwin, Tripodi and Bieri [1967] randomly paired role figures

and constructs, a procedure which apparently did not affect their results.

Groups 2 and 4 (H/Hmax Measure)
For Groups 2 and 4 the number of adjectives and number of cate-

gories for each role figure were counted. These numbers were then used

29

in the calculation of H/Hmax. The formula for H was:
H = log n - 1-2 n. - log n
2 n 1 2 i

where n was the total number of adjectives and ni was the frequency in
the ith category. Hmax was obtained by computing H when all categories

had equal frequencies. The ratio of H to Hmax was then computed.

Data Analysis

 

Groups 1 and 2 (Bieri's Ten Role Figures)

The analysis was correlational in nature. Correlations were
computed based on all ratings and complexity scores for all role fig-
ures and for all participants. In addition, a stepwise multiple regres-
sion analysis was also performed, with the complexity score(s) as the
criterion variable(s) and the other variables (liking, interaction,
involvement, knowledge) as predictor variables. However, in these
analyses, the number of cases was 300, with each of 30 participants con-
tributing ten ratings and/or complexity scores each. This would tend
to inflate the correlations and mask the true strength of the relation-
ships. Thus, although the above analyses were examined, they provided
only tentative information.

Correlations among the ratings of liking, interaction, involve-

ment and knowledge and complexity scores for each role figure were

 

computed and examined. Also examined were the correlations among the
variables for each single participant; for example, the strength of
the relationship between liking and complexity over responses to the
ten role figures for an individual participant was assessed.

For Group 1, ratings were made on a l to 10 scale with 1 indicating

30

high liking, knowledge,etc., and 10 indicating low liking, knowledge,
etc. The higher the numerical value of the complexity score on the
Rep Test, the simpler or less complex the response. Thus, a strong
negative correlation would indicate a relationship between dislike
and greater complexity of response such that disliked or less Well
known persons were being more highly differentiated. A positive
correlation would indicate that disliked persons were not being well
differentiated, that participants tended to be more complex in respond-
ing to positive role figures.

For Group 2, however, a strong negative correlation between any
of the ratings and number of adjectives, number of categories or H/Hmax
would indicate greater quantity of responding or a more complex response

to liked or well known persons.

Groups 3 and 4 (Reich's Six Role Figures)

Data from Groups 3 and 4 were analyzed by a 2 (sex of participants)
X 2 (knowledge of the role figure) X 3 (affective valence of the role
figure) analysis of variance with repeated measures on the last two
factors. For Group 3, a separate ANOVA was computed for complexity
scores on evaluative traits and complexity scores on nonevaluative
traits. For Group 4, three ANOVAs were calculated for three dependent
measures: number of adjectives, number of categories, and H/Hmax.

In addition, overall correlations among all variables and the
correlations among the variables for each role figure were computed

on the data from Groups 3 and 4.

CHAPTER III

RESULTS

Group 1

Participants in Group 1 responded to Bieri's ten role figures.
Complexity scores came from the Rep Test, using evaluative and nonevalu-

ative traits.

Positive and Negative Role Figures

The mean liking rating for all role figures was 3.40, indicating
either a tendency for participants to use the lower end of the 10 point
scale, or that the majority of role figures were rated as more positive
than negative by participants. Boss, Person You Dislike, Person
Difficult to Understand and Person With Whom You Feel Most Uncomfortable
were rated more negatively (mean liking ratings for these role figures
were all greater than 3.40, indicating greater dislike) than the re-
maining six role figures. The overall mean knowledge rating was 3.63,
and the same pattern as above emerged, i.e., Boss, Person You Dislike,
etc. were rated as less well known than the other role figures. Mean
interaction and involvement ratings followed similar patterns.

The mean complexity scores on evaluative traits for these same
four role figures were lower than for the remaining six role figures,
indicating a more complex response to the role figures rated most
negatively and least well known. As indicated previously, the lower
the numerical score, the more complex the response. On nonevaluative

31

32

traits, the mean complexity scores for each role figure did not follow
this pattern; no meaningful pattern emerged. Mean ratings for liking,
interaction, involvement and knowledge and mean complexity scores on
evaluative and nonevaluative traits for each role figure and overall

are presented in Table 1.

Overall Analysis

Overall correlations computed on Group 1 data indicated that
liking and involvement ratings were negatively correlated with complex-
ity scores on evaluative traits (r = -.276 and r = -.109, respectively).
(See Appendix H.) These negative correlations indicated a tendency to
respond in a more complex manner to role figures designated as less,
well liked and less involving. A stepwise regression analysis showed
the liking variable as contributing the most to variability in complex-
ity scores (7%); the other three variables accounted for an additional
3%. (See Table 2.)

Knowledge ratings and complexity scores on nonevaluative traits
showed a low positive relationship (r = .112). This positive correla-
tion implied that well known role figures tended to be responded to in
a more complex manner. The regression analysis in this case showed
the knowledge variable as accounting for only 1% of the variance; the
other variables contributed another 1%. (See Table 2.) Complexity
scores on evaluative traits and on nonevaluative traits were not highly
correlated.

The correlations among the ratings of liking, interaction, involve-
ment and knowledge for each role figure were positive. The overall

correlations were all positive and significant. The two strongest

33

 

 

 

 

 

 

 

 

 

 

 

 

 

 

n~.ca oo.na mo.m Hq.q m~.< oq.m mama: Hamnm>o

ow.HH mq.qH mN.m n~.o mm.q n¢.q euwz .mEoodD

Hmmm cow comumm

mw.a no.8H mo.m oa.m om.m Aa.N swam on mega

p.sow aomumm

mm.oa NH.NH mm.q oo.m mn.q mm.m tcmumumpsa cu

pacofimmwa compom

nn.w CH.SH mm.q up.“ om.c mo.w mxHHmHn

sow cowumm

mm.oa nw.ma oo.m o~.o mw.q oq.¢ mmom

mm.oa mo.om mm.~ mm.N 0H.m nw.a xmm muumommo

ozu mo pcoaum

OH.0H ma.ma mq.N nn.m mo.N om.a wow meow

can mo pcmaum

mm.m mm.wa mm.~ mm.~ ow.H uwnumm

mm.m mq.ma wo.m mm.H mw.H umnuoz

mm.oa m~.N~ oq.~ m~.~ mamm
mufimue .Hm>mcoz wuamua .Hm>m mwpmHBocM acme coauom wdwxaq

1%ufixmaaaoo thuflondEoo oo>ao>nH luwuaH

 

H manna

woudwam oaom H acouu you monocm can mwcfiumm can:

 

34

ummu pmawmunosu .mo. v a «

 

 

 

 

 

 

 

 

 

 

 

,NNo. ooo. oNo. ooo. oHo. Hoo. unmao>ao>uH .o
moo. oHo. omo. moo. ado. moo. cowuumumuaH .m
ooH.- «om. ooh. ooo. oHo. ooo.- .wanxwa .N
HHH. ooH. omo.~ mHo. mHo. «AH. «womasoae .H
mufiwua m>fiumaam>ocoz

thuwxmamsou
Noo.- moo. woo. woo. HoH. ooodu wwwuasoae .4
AAA. ooo. moo.H moo. ooo. ooH.- unmam>ao>aH .m
BAH. Hoo. AHN.¢ oHo. soo. Hoo.u cowuomumuaH .N
mmo.- ooo. oHH.AH oko. oko. «okm.u manage .H

 

 

muamufi o>fiumsam>m

 

 

 

muom mucmoHMAcwwm Housm cu m wwcmno mumcvm m u mamafim
mumsvm m

Imuwxoamaoo

 

H macho you mwmhamc< doommouwmm mmasdmum

N magma

 

35

relationships were between liking and involvement ratings, r = .729,

and interaction and involvement ratings, r = .736. (See Appendix H.)

Complexity on Evaluative Traits

The relationship between liking and complexity was not consistent
for all of the role figures (see Table 3). For four of the role figures,
Friend of the Same Sex, Friend of the Opposite Sex, Person You Dislike,
and Person Difficult to Understand, there was essentially no relation-
ship; for the other role figures there were negative correlations
between liking and complexity. Self (r.= -.454) and Mother (r = -.362)
displayed the strongest relationships. The negative relationship indi-
cated that participants were less complex in responding to liked per-
sons and were more complex in reponding to disliked persons. The
average of these correlations (using an r to z transformation;

[McNemar, 1962]) was -.l66, a significant negative relationship. In
examining the correlations for individual participants, most were
negative; about 16% were positive. (See Appendix D for these correla-
tions.)

The relationship between knowledge ratings and complexity scores
was weak. For five of the role figures the relationship was essentially
zero; for the other five role figures the correlations were nonsigni-
ficant and positive. The average correlation was essentially zero.

For the individual participants' correlations, the majority were non-
significant and negative, but approximately 25% were positive, indica-
ting a fair amount of individual variability in this regard. Tenta-
tively, it appeared that while for individual participants the response

to less well known persons was more complex, for certain individuals

36

the relationship was reversed, and persons known slightly were less

well differentiated than well known persons.

Complexity on Nonevaluative Traits

The relationship between liking and complexity was not consistent
for all of the role figures (see Table 3). For two role figures the
relationship was zero; for the remaining role figures equal numbers
showed positive and negative correlations. The stronger relations,
however, were positive, indicating more complex responses to liked
persons. The average of these correlations was essentially zero. The
correlations for individual participants were both positive and nega-
tive, although the negative correlations outnumbered the positive (see
Appendix D). Again, this suggested a trend opposite to that trend
suggested by the correlations for some of the role figures.

The relationship between complexity and knowledge was only
slightly more consistent. For five of the role figures the correlation
between knowledge ratings and complexity scores was positive, indica-
ting more complexity in responding to better known role figures. The
average correlation, however, was still essentially zero. For individual
participants, both a positive and negative relationship between the two
variables appeared to be Operating, indicating, again, individual
variability in response.

The relationship between the two complexity scores (one on evalu-
ative traits, the other on nonevaluative traits) was nonsignificant
for each of the role figures. For Mother, Friend of the Opposite Sex
and Person You Dislike, the correlation between the two scores was low

and negative; for the other role figures the correlation was zero or

37

positive. The correlations for individual participants were both

positive and negative.

Conclusion

Thus, for complexity on evaluative traits, a vigilant or justi-
fication effect appeared to occur; participants seemed to be respond-
ing to negative role figures in a more complex manner. However, for
individual role figures and individual participants such a relation-
ship often was not apparent. For complexity on nonevaluative traits,
relationships were not at all clear cut. However, the trend appeared
to be opposite to the trend in results for complexity scores on evalua-
tive traits. For nonevaluative traits, the tendency to be more complex
in responding to well liked and well known role figures was very weak
and tentative at best, again hiding a great deal of variability both
for the role figures and for individual participants. Complexity

scores on evaluative and nonevaluative traits were not related.

Group 2

Participants in Group 2 responded to Bieri's ten role figures.
Number of adjectives, number of categories and H/Hmax scores were

examined.

Positive and Negative Role Figures
Mean liking ratings for the role figures showed a pattern similar
to that for Group 1: Boss, Person You Dislike, Person Difficult to
Understand and Person With Whom You Feel Most Uncomfortable were rated
more negatively than the other role figures. In general, although the

pattern was not as consistent as it was for Group 1, participants also

38

on u z
ummu tmafimulosu .mo. v a «

 

«no. moo.

 

 

 

 

 

 

 

 

 

 

 

«we. owe. «oca.l mcoaumamuuoo.
owmuo><
oAH. moo. ooo.- ooo.. HoH.- gun: .oaouao
down now somuom
How. moH. oHN. oko.n ooo.- (mama on mega
v.90w somumm
who. omm. Ham. mom. (mono. oamumumoa: on
“Hoodoofio commas
ooH.- Hem. Nam. Hmo.n Noe. mxAHmHo
cow cowumm
me. mnH.| Hmc. «No. Hom.l mmom
NSN.- ooo.- AAN.- ooH. ooo. woo muamoaao
no oamfium
mew. nos. wees. ooN. ooo.- xmm memo
mo ocmaum
moo. Hmo. HoH.- okH. oo~.- “means
~mH.- ooo. Hmo.n m-.- :mom.- Hugues
mom. soo.- ooH.- ooo.- «mmo.- wamm

 

 

.Hm>mcoZI.Eoo
o .Hm>mu.aoo

.Hm>oCOZI.Eoo .Hm>osozl.aou
o .3oae o wafixaq

.Hm>MI.aou
w .BOaM

.Hm>m1.eoo
o waexfia

 

newsman mHom H ozone How

mouoom hufimeosou can wwcaumm wcoafi msowumamuuoo

m manna

 

39

gave these role figures low ratings in terms of interaction, involve-
ment and knowledge. Boss, Person You Dislike and Person With Whom
You Feel Most Uncomfortable were the role figures with the lowest
means for number of adjectives, number of categories and the H/Hmax
scores. The latter indicated that participants were somewhat more
complex in responding to positive and well-known role figures. (See

Table 4 for the mean ratings.)

Overall Analysis

For this group a negative correlation signified that as complex-
ity increased there was increased liking, or the converse. Number of
adjectives correlated negatively and significantly with ratings of .
liking, interaction, involvement and knowledge. Participants wrote
more adjectives for persons they liked, interacted frequently with,
were involved with and knew well. The same pattern of correlations held
for number of categories and H/Hmax scores. A stepwise multiple regres-
sion analysis showed that for number of adjectives, the interaction
variable explained 7% of the variance while the other three variables
contributed only another 2%. For the number of categories, the inter—
action variable acounted for 4% of the variance and the remaining
variables contributed 2%. In regard to H/Hmax, involvement contributed
5% of the variance; the other variables accounted for an additional 1%.
(See Table 5.)

The correlations among the ratings of liking, interaction, involve-
ment, and knowledge for each of the role figures were positive. The
overall correlations among the ratings were all positive and significant,

with the strongest relationships between liking and involvement (r = .709)

40

 

 

 

 

 

 

 

 

 

 

 

 

 

 

moo. 0H.~ m<.o wH.q on.< mw.¢ mm.m menu: A
Hamum>o
mum. mn.H nm.m oa.c mm.n mn.o oo.m :uHS .msoocp
Hook cow somuom
emu. o~.~ o~.m nc.m mq.q mo.q mc.N can: ou oxfiq
p.3ow somwwm
«no. o~.~ ma.o nm.m o¢.q m~.q no.m ensconced: cu
uacuammwa somuom
Hum. ow.H nn.m mm.m sq.m no.n mm.h mxwamfio
‘ now common
owe. no.H no.m oo.n 5H.n nm.m mm.¢ mmom
me. mm.~ om.n oo.~ mo.m nm.~ mo.H Rom ouamommo
mo panama
Ono. no.~ no.0 nn.m oa.m no.m NH.~ Rom mEMm
mo oceans
nmn. mm.~ ow.o o~.m ow.~ nn.H nonumm
own. mH.N ca.n om.N MH.N mo.H “memo:
mmn. oq.~ om.n no.~ om.~ maom
xmsm\m moauowoumo mm>auoont< omega acme cognac .Awwfiqu
mo Honssz mo Honacz Iaoax I~>Ho>uH lumuaH

 

o maoma

mousmfim oaom N macaw mom mouoom tam mwcfiumm one:

 

ammo toawauloau .no. v a a

 

 

 

 

 

 

 

 

41.

.. oao. so. ooo. ono. «aoa.- amongsoce .o
oko.- ooo. no. ~oo. ooo. «ooo.- coauuauuuoo .n
koo.- man. so. ooo. kno. .oo~.- mousse .~
ooo.- goo. on.HH omo. «no. .o-.- uooau>aopua .~

NEH
ooo. How. oo. ooo. Noo. «no~.- unusu>qo>ua .4
«no.1 own. on. Noo. ~oo. wo~u.: «woodsoau .n
ANA.- oAH. on.~ «no. ooo. «no~.- mowed; .N
nn~.- moo. on.oH ooo. ooo. «o-.- cowuuuuooas .H

 

nuuuowouuu uo nonasz

 

 

 

 

ooo.: are. co. «co. . soc. cama.n mcfixua .Q
mmo. who. mu. ooc. mac. «wm~.| unfilu>do>um .n
omH.I mmo. as.n oao. «mo. «enN.1 cavedaouu .N
cn~.1 coo. an.sn who. Qua. «th.| acquuuuuuau .H

 

oo>wuuofiv< uo amass:

 

 

noon oucuuwuwswum yucca cu m uumuao ouasam m u omwmwm
vacuum a

 

N macaw pom mwmhama< scammmummm mmwsdmum

m manna

 

42

and interaction and involvement (r = .764), as in Group 1. (See

Appendix H.)

Number of Adjectives and Number of Categories
See Appendix E for a discussion and presentation of data for

number of adjectives and number of categories.

H/Hmax

The relationship between H/Hmax and liking ratings was not con-
sistent for all of the role figures; for three role figures, Person
You Dislike, Person Difficult to Understand and Person You Would
Like to Help, the correlation was essentially zero; for two role figures,
Self and Boss, the relationship was positive, and for the remainder a
negative relationship between H/Hmax and liking was indicated. The
latter negative relationship suggested that the greater the liking,
the higher was H/Hmax, and the greater the complexity of response.

The average of these correlations was essentially zero. For individual
participants, the correlation between the two variables was negative
in approximately 80% of the cases.

Knowledge ratings and H/Hmax were negatively correlated for four
of the role figures and positively correlated for three others. The
negative correlations seemed to be slightly stronger than the positive
correlations. For about 60% of the participants, the correlation
between knowledge ratings and H/Hmax scores was negative. Those par-
ticipants who indicated greater knowledge of the stimulus person
responded more complexly than those indicating lesser knowledge.

Thus, data for Group 2 indicated that liked and well known persons

43

Table 6

Correlations Among Ratings and Complexity

for Group 2 Role Figures

Score

 

 

 

 

 

 

 

 

 

 

 

 

 

 

‘1
Liking & Know. & E
H/Hmax H/Hmax ?
3
Self .109 .292 %
Mother -.435* -.364* 2
Father -.304 -.257
Friend of
Same Sex —.276 -.096 4
Friend of '
Qpposite Sex -.213 -.257 .
Boss .397* .048 4
Person You i
Dislike -.034 .080 j
Person Difficult 7
to Understand .054 .183 '
Person You'd
Like to Help -.075 .199
Person You Feel
Uncomf. With —.168 —.134
Average
Correlations -.098 -.032 1

 

 

* p < .05, two-tailed test
N = 30

44

were responded to in a more complex way than disliked and less well

known persons, supporting a frequency of interaction interpretation.

Group 3

Participants in Group 3 responded to six role figures: Person
You Know Well and Like, Know Well and Dislike, Know Well and Feel
Neutral Toward, Know Slightly and Like, Know Slightly and Dislike,
Know Slightly and Feel Neutral Toward. Complexity scores came from

the Rep Test, using evaluative and nonevaluative traits.

Manipulation Check
Mean ratings on liking, interaction, involvement and knowledge
of each of the role figures served as a check. For liking and know;
ledge, mean ratings followed the pattern indicated by the role figure
descriptions to a satisfactory extent. Interaction and involvement

ratings were less clearcut. (See Appendix F.)

Evaluative Traits

An analysis of variance performed on complexity scores on evalua-
tive traits as the dependent measure showed a significant main effect
for Affect (F(2/56) = 17.115, p < .0005). Participants were significantly
more complex in responding to disliked persons (M = 13.02) and neutral
persons (M = 15.38) than in responding to liked persons (M = 19.18),
as shown by a Newman-Keuls test with p < .01. The former two means
did not differ significantly from each other. There was also a signi-
ficant Sex X Affect interaction (F(2/56) = 3.201, p < .05). Females
were significantly more complex than males in responding to the neutral

role figures. For males, the means for liked and neutral persons did

45

not differ significantly from each other; females were significantly
more complex in responding to the neutral figures than to the liked

role figures. (See Table 7 for cell means.)

Tafle7
Group 3

Means and Standard Deviations of Complexity Scores
Evaluative Traits

 

 

 

 

 

 

 

 

 

Liked Disliked Neutral
Participants Role Figures Role Figures Role Figures
Male M 18.97 M 12.93 M 17.40 16.43
SD 6.15 SD 5.56 SD 7.78
Female M 20.77 M 13.10 M 13.37 15.74
SD 6.49 SD 7.13 SD 4.90
19.87 13.02 15.38

Nonevaluative Traits
The analysis of variance for complexity scores on nonevaluative
traits resulted in no significant interactions or main effects. How-
ever, the main effect for Affect approached significance (p < .10).
Inspection of means showed that the trend was toward greater differ-
entiation of disliked persons, followed by liked and neutral persons.

(See Table 8.)

Group 4

Participants in Group 4 responded to six role figures. Number
of adjectives, number of categories and H/Hmax scores were the depen—

dent measures.

46

Tafle8
Group 3

Means and Standard Deviations of Complexity
Scores - Nonevaluative Traits

 

 

 

 

 

 

 

 

 

Role Figures Liked Disliked Neutral
Kn W 11 M . 3 . .
own e 10 0 M 9 67 M 10 53 10.08
SD 3.79 SD 2.87 7 SD 4.44
Known Slightly M 11.17 M 8.73 M 11.50 10 47
SD 6.65 SD 3.37 SD 5.04
10.60 9.20 11.02

Manipulation Check
Mean ratings for liking and knowledge were consistent with the

role figure descriptions. (See Appendix G.)

Number of Adjectives and Number of Categories
A discussion and presentation of the data for number of adjec-
tives and number of categories as dependent measures appear in Appendix

G.

H/Hmax
There was a significant main effect for Knowledge (F(l/28) =
16.633, p < .01). Participants were more complex in responding to
well known persons (M = .76) than to persons known slightly (M = .55).

(See Table 9.)

Groups 3 and 4 Correlations
The pattern of correlations among the ratings and complexity

scores for the role figures in Groups 3 and 4 (shown in Appendices

47

Table 9
Group 4

Means and Standard Deviations of H/Hmax Scores

 

 

 

 

 

 

 

 

 

Role Figures Liked Disliked Neutral
Known Well M .792 M .804 M .685 760
SD .333 SD .295 SD .424
Known Slightly M .513 M .593 M .560 555
SD .467 SD .437 SD .467
.653 .698 .623

F and G) showed much variability, a pattern similar to the pattern
found for the role figures in Groups 1 and 2. This indicated that
the relationship between complexity of response and liking or know-

ledge ratings was not consistent for all of the role figures.

Overall Comparison

 

Comparison of Groups 1 and 3 in terms of their respective over-
all correlations (see Appendix L), showed that none of the correlations
differed significantly from each other. For Groups 2 and 4, one of
the comparisons showed a significant difference. For Group 4 the
correlation between liking and H/Hmax was .068, while for Group 2
the correlation between liking and H/Hmax was -.209. The two corre-

lations differed significantly (p < .05).

CHAPTER IV
DISCUSSION

Interpretation

1) Results from Groups 1 and 2 (in both cases Bieri's ten role
figures were presented to participants) indicated that the role figures
previously assumed by researchers to be positive or negative in affec-
tive valence seemed to have similar valences for the participants.
In ranking the role figures in terms of their mean liking ratings,
Mother, Father, Self, Friend of the Opposite Sex and Friend of the .
Same Sex were the five most positively rated role figures while Boss,
Person you Dislike, Person You would Like to Help, Person Difficult
to Understand and Person With Whom You Feel Most Uncomfortable were
the five least positively rated role figures, thus offering some
support for prior assumptions with a college sample. However, although
the latter five.role figures were rated less positively than the former,
they were by no means rated negatively. Except for Person You Dislike,
which received the highest ratings of dislike (8.03 and 7.93), the
other four "negative" role figures were actually rated more neutrally
than negatively. Such a pattern might indicate a response bias on
the part of the participants to use more often the positive end of the
ten point scales. Or it might indicate that only one of the role
figures was actually negative in valence for the participants. The
only conclusion permissible is that in terms of rankings, the role
figures fell into the positive and negative subsets assumed in prior

48

49

research.

2) The results for Group 1 (Bieri's ten role figures and the
Rep Test) indicated that a vigilance or justification effect did occur
when evaluative traits were used in administering the Rep Test, as
predicted. The prediction was based on Shepherd's observation [1972]
that the tendency to use positively evaluated words more readily than

negatively evaluated words (the Pollyanna effect) operates in complet

ing the Rep Test. Positively evaluated words would be assigned to
positive persons, producing redundancy in response and yielding a low
differentiation score, while both positively and negatively evaluated
words would be assigned to negative role figures, producing higher
differentiation scores for the negative persons.

When nonevaluative traits were used, a vigilance effect did not
appear, and in fact a very weak relationship between knowledge and
complexity and liking and complexity emerged, opposite to that of the
vigilance effect, i.e., participants tended to be more complex in
responding to role figures they rated as liking and knowing well than
to role figures they rated as disliking and knowing only slightly.

Thus, in comparing results for evaluative and nonevaluative traits
on the Rep Test, it was obvious that, as Shepherd [1972] reported,
results were markedly different. Only in using evaluative traits was
the vigilance effect supported. Shepherd's nonevaluative traits and
the additional nonevaluative traits used in this study were from the
middle range in favorability on Anderson's list of personality-trait
words. In rating persons of positive or negative valence on these non-
evaluative traits, the participant would not find words ordinarily

applied to positive or negative persons. This was demonstrated by Lott,

50

Lott, Reed and Crow [1970]. The tendency to use certain words to
describe liked persons and certain words to describe disliked persons
would be thwarted. Participants confronted words that perhaps they
applied to neither liked or disliked persons and hence may have
answered in a random manner or made inferences regarding traits they
ordinarily did not apply to the persons they were rating. If parti-
cipants were answering randomly, one would expect lower numerical
scores or greater complexity of response; this was supported by the
the data. The mean complexity score for nonevaluative traits was much
lower than the mean complexity score for evaluative traits in Group 1.

Results for Group 3 (Reich's six role descriptions and the Rep
Test) offered partial confirmation for Group 1 results. For complexity
on evaluative traits, there was a main effect for Affect; negative
(and neutral) role figures were differentiated to a greater extent than
liked role figures, again supporting the vigilance or justification
hypothesis. The analysis for nonevaluative traits however, indicated
only weak support for such an effect.

3) Results for Group 2 (Bieri's ten role figures and the H/Hmax
measure of complexity) showed a relationship similar to that found in
Group 1 for nonevaluative traits, i.e., as liking and knowledge of the
person increased, complexity of response also increased. This did not
support a vigilance or justification interpretation. The relationships
were weak; the overall regression analysis showed a very low percentage
of variance accounted for by the variables and the direction and strength
of the relationships for individual role figures and individual partici-

pants varied to a great extent. Although relationships were weak, they

51

tended to support a frequency of interaction interpretation.

For H/Hmax, Reich obtained a main effect for Affect; Group 4 data
showed a main effect for the Knowledge variable in this study. AThus,
whereas Reich found support for the neutral affect hypothesis using
H/Hmax (neutral persons were more highly differentiated than either
positive or negative persons), this study found support for the frequency
of interaction hypothesis using H/Hmax. Clearly the Knowledge variable
was important. For number of adjectives and number of categories, the
Knowledge by Affect interaction indicated that Person Known Well and
Liked and Person Known Well and Disliked received significantly greater
numbers of adjectives and categories than the Neutral role figures or
the Known Slightly role figures. The data for H/Hmax showed a similar
pattern (although the interaction was not significant). Thus, this
study found no support for Reich's contention that the greater the
involvement, the less the discrimination of stimuli. The opposite was
fOund; the greater the involvement, the greater was the discrimination
of stimuli. Indeed, in Group 2 (Bieri's ten role figures and the H/Hmax
measure of complexity), Involvement was an important variable (indicated
by the overall regression analysis), and the relationship was such that
as involvement increased, complexity of response also tended to increase.
This study found support for the frequency of interaction hypothesis for
both complexity of response and quantity of responding.

4) In comparing overall correlations for Groups 2 and 4, it was
found that the relation between liking and H/Hmax was essentially zero
for Group 4 while for Group 2 the liking variable correlated significantly

with H/Hmax such that as complexity of response increased, ratings of

52

liking increased, or the converse. In Group 4, the analysis of variance
produced a main effect for the Knowledge variable. However, eta2 was
only .06. For Group 2's overall regression analysis, the Knowledge
variable entered the equation only on the last step (Involvement was
the first variable to enter the equation accounting, however, for only
5% of the variance). The only difference between Groups 2 and 4 were
the role figures presented. For each of the Group 4 role figures,
participants briefly described their relation to the person they had
thought of. An examination of these descriptions revealed that in
nearly all cases peers were selected to fit the role descriptions.
Bieri's ten role figures included older persons (such as Mother, Father,
Boss). Supnick [1964;as reported by Crockett, 1965] found that parti-
cipants used more constructs to describe peers than to describe older
persons. A tentative explanation might be that in responding to peers,
knowledge was the more important variable in influencing complexity of
resonse, while in responding to both older and younger persons, the
involvement variable was more important. Perhaps relationships with
peers are similar in affective experience but differ quantitatively
(i.e., in how often the person is seen and how well the person is
known) while for older persons the role expectations andfibehaviors
required result in stronger and qualitatively different affective re-
lationships. Such an interpretation awaits further testing. However,
the overall conclusion is that the presentation of two different sets
of role figures did not markedly affect results, as predicted.

Finally, the Sex X Affect interaction for complexity scores on

evaluative traits in Group 3 was somewhat similar to the tendency

53

reported by Irwin, Tripodi and Bieri [1967], that "females compared to
males tended to differentiate more among neutral and negative figures,
the latter sex difference being statistically significant" [p. 446].
In this study, females differentiated neutral role figures to a signi-
ficantly greater extent than did males. However, for negative role
figures, males and females did not differ significantly in complexity
of response. The latter differs from the results of Irwin, Tripodi
and Bieri [1967].

Shepherd [1972] also reported a Sex X Valence interaction; females
were less complex in responding to positive role figures but more com-
plex in responding to negative role figures than were males. Irwin,
et al., suggested that females have a greater need to depend upon others,
and hence a greater ability to differentiate among potentially threaten-
ing figures. However, this effect might well be due to the Pollyanna
effect being greater for females than males (consistent with Warr,
[1969]; as reported by Shepherd [1972]). Females would tend to rate
positive figures uniformly favorably and negative figures less uniformly
favorably than males. Lott, Lott, Reed and Crow [1970] found for one
of the samples they tested that females tended to describe their
acquaintances with adjectives from Anderson's list of personality-trait
words which.were slightly higher in likableness value than did males.
These explanations are not appropriate for the results of the present
study; although female participants in Group 3 responded more complexly
to neutral role figures than did the males, they did not respond in a
more complex way to negative role figures.

Other reported sex differences in complexity of response have

been highly inconsistent. Supnick [1964; as reported by Crockett, 1965]

54

fOund a main effect for Sex; females used more constructs to describe
persons than did males. Soucar [1970] and Koenig and Seaman [1974a]
found main effects for Sex; males were more complex in their responding
than were females. Other studies [Soucar and DuCette, 1971; Miller,
1968; Shepherd, 1972] reported no sex differences. Glixman [1965]
found that females exhibited greater complexity of response than men;
females used a greater number of categories than did males, contribu-
ting to higher H scores. However, for H/Hmax, there was no effect for
sex.

The Sex X Affect interaction found in the present study does not
correspond exactly to any reported result in prior research. No main
effect or interaction for Sex was found in Group 4. Thus, there does
not appear to be any ready explanation for the interaction found here.

Thus, the predictions originally made in this study were generally
upheld by the data. However, the data also showed that the overall
relationships or trends masked a great deal of individual participant
variability and variability in responding to the role figures. One
might ask whether the role figures each constituted a separate domain
for participants. In such a case one would not expect generality of
complexity scores across these domains. To assess this, in a subsequent
analysis, the role figures were ranked according to their respective
complexity scores for each of the 30 participants in Groups 1 and 2;
Kendall's coefficient of concordance was then computed on these 30 sets
of rankings. For Group 1, the coefficient for complexity scores on
evaluative traits was .171 (p < .01) and for nonevaluative traits,

.065 (n.s.). For Group 2 H/Hmax scores, the coefficient was .099

55

(p < .01). These low coefficients support the notion of low consistency
of response to the different role figures. The notion of the role
figures constituting separate domains appears to be a plausible one.
This study did not consider the complexity of the individual
participants in the interpersonal realm. Were the more complex parti-
cipants consistently more complex in their responding and were simple
participants consistently simple in their responding? A subsequent
analysis performed on the data for Groups 1 and 2 entailed ranking
the 30 participants according to their complexity scores for each of
the ten role figures and then computing Kendall's coefficient of con-
cordance on these ten sets of rankings. The coefficients were, in.
Group 1, .139 (n.s.) for the complexity scores on evaluative traits,
and .172 (p < .01) on complexity scores for nonevaluative traits. For
Group 2, the coefficient for H/Hmax scores was .272 (p < .01). These
coefficients are quite low. Supnick [1964; as reported by Crockett,
1965] reported a coefficient of .604 for her fourteen participants;
participants were ranked according to the number of interpersonal con-
structs they used in describing eight stimulus persons. One might
argue that generality of complexity scores would be expected to be
stronger for Supnick's data than for this data, since verbal fluency
probably contributed to the scores in Supnick's data, and a very fluent
participant would probably be fluent‘ in all of his/her descriptions.
However, the low generality of complexity for individual participants
and across role figures here, seems to indicate low generality of any
sort and may indicate randomness in the data. It calls into question
the measures of complexity, the motivation of participants, the sample

of persons tested, etc. In comparing mean number of adjectives, number

56

of categories and H/Hmax scores reported by Reich [1969] with the means
from Group 4, one finds that in all cases the means were much lower

for this study. Perhaps this lesser range of response affected overall
results; perhaps it is indicative of the motivation of the persons
tested, given that the H/Hmax task depends to a great extent on how
much time and effort the participant is willing to spend in its comple-

tion.

Implications

This study points to the necessity for examining differences in
the measures used to assess complexity. The basic issue involves the
definition of complexity or differentiation and how it pertains to_
person perception. What does it actually mean to differentiate among
others or respond to them in a complex way? In a paper and pencil test
of complexity the participant responds essentially to a verbal symbol of
a person with whom.he/she interacts. This seems to be a very different
situation than actual face-to-face interaction with others. The defini-
tion of complexity is unclear in this context; studies arising from
different approaches to its measurement have yielded conflicting results.
At this point one cannot adequately examine the variables, such as
affective valence or knowledge of the person, that might affect complexity
of response when the notion of the latter is so unclear and when the
measures available perhaps do not reflect what one might mean by the

notion in actual interactions.

Summary

In summary, this study indicated no strong differences in results

of two different sets of role figures. Participants were asked to

57

indicate their degree of liking for each of the ten role figures con-
ventionally used in administering the Rep Test measure of complexity.
When ranked, the ten role figures accurately reflected the positive

and negative subgroups outlined in prior research.

In addition, this study indicated that results supportive of
either the vigilance or justification effect or the frequency of inter-
action effect could be found depending on the measure of complexity
used. On the Rep Test measure, the presentation of evaluative traits
affected scores in such a way that a vigilance interpretation was pos-
sible. When nonevaluative traits on the Rep Test were presented,
results showed no strong relationship between complexity of response
and liking or knowledge of the stimulus person. This might have occur-
red because the nonevaluative traits were not traits ordinarily used to
describe persons with strong affective valences. For the H/Hmax measure
of complexity, knowledge of the person and degree of involvement were
related to complexity of response. Well known persons and persons with
whom participants indicated greater involvement were more highly
differentiated.‘ Persons wrote a greater number of adjectives to
describe well known and well liked stimulus persons, which also sup-
ported a frequency of interaction interpretation.

The conclusion drawn from the above results was that the vigi-
lance effect appeared to be a very specific one that was found only
when evaluative traits were presented on the Rep Test measure of complex-
ity. When nonevaluative traits were used on the Rep Test and results

for another measure of complexity were examined, no such effect was

58

fOund. Results generally supported the frequency of interaction
hypothesis.

This study also showed that in examining the relationship between
complexity and the other variables for each of the role figures and for
individual participants in two of the groups there was a great deal
of variability in the strength and direction of these relationships,
indicating that the chosen variables were not influencing complexity
to the same degree, in the same way, for different individuals and in
responding to the different role figures. The factors that might
account for this great amount of unexplained variance remain unspecified
at this time and await further investigation. Finally, the generality
of complexity scores for this sample was quite low perhaps indicating
that other (unknown) factors were affecting participants' responses in

the experimental situation.

LIST OF REFERENCES

REFERENCES

Anderson, N.H. Likableness ratings of 555 personality-trait words.
Journal of Personality and Social Psychology, 1968, 9, 272-279.

 

Attneave, F. Applications of information theorygto psychology, New
York: Holt, Rinehart and Winston, 1959.

 

Bieri, J. Cognitive complexity-simplicity and predictive behavior.
Journal of Abnormal and Social Psycholqur 1955, él, 263-268.

 

Bieri, J. Complexity-simplicity as a personality variable in cogni-
tive and preferential behavior. In D. W. Fiske and S. R. Maddi
(Eds.), Functions of varied experience. Homewood, 111.: Dorsey
Press, 1961.

 

Bieri, J. Cognitive complexity and personality development. In 0. J.
Harvey (Ed.), Experience, structure and adaptability. New York:
Springer Pub. Co., 1966, pp. 13-37.

 

Bieri, J. Attitudes and arousal: Affect and cognition in personality
functioning. In C. W. Sherif and M. Sherif (Eds.), Attitude,
ego-involvement and change. New York: John Wiley & Sons, Inc.,
1967.

 

Bieri, J., Atkins, A. J., Briar, S., Leaman, R. L., Miller, H., and
Tripodi, T. Clinical and social judgment: The discrimination
of behavioral information. New York: J. Wiley & Sons, Inc.,
1966.

 

Boucher, J., and Osgood, C. E. The pollyanna hypothesis. Journal of
Verbal Learning and Verbal Behavior, 1969, 8, 1-8.

 

 

Crockett, W. H. Cognitive complexity and impression formation. In
B. A. Maher (Ed.), Progress in experimentalppersonality research.
V01. 2. New York: Academic Press, 1965, pp. 47-90.

Erdelyi, M. H. A new look at the New Look: Perceptual defense and
vigilance. Psyghological Review, 1974, 81, l-25.

 

Glixman, A. F. Categorizing behavior as a function of meaning domain.
Journal of Personality and Social Psychology, 1965, 2, 370-377.

Harvey, O. J., Reich, J. W., and Wyer, R. S. Effects of attitude dir-
ection, attitude intensity, and structure of beliefs upon differ-
entiation. Journal of Personality and Social Psychology, 1968,
19, 472-478.

59

60

Irwin, M., Tripodi, T., and Bieri, J. Affective stimulus value and
cognitive complexity. Journal of Personality and Social
Psychology; 1967, §, 444-448.

 

 

Kelly, G. A. The psychology ofypersonal constructs. New York:
Norton, 1955.

 

Koenig, F. Positive affective stimulus value and accuracy of role
perception. British Journal of Social and Clinical Psychology,
1971!.19' 385-386.

 

Koenig, F., and Seaman, J. Comparison of cognitive complexity of male
and female subjects responding to male and female stimuli.
Perceptual and Motor Skills, 1974, 38, 1345-1346. (a).

 

Koenig, F., and Seaman, J. Vigilance and justification as explanations
of complex cognition. Journal of Social Psychology, 1974, 23,
75-80. (b).

 

Lott, A. J., Lott, B. E., Reed, T., and Crow, T. Personality-trait
descriptions of differentially liked persons. Journal of Per—
sonalityyand Social Psychology; 1970, lg, 284-290.

 

 

Mayo, C., and Crockett, W. H. Cognitive complexity and primacy-recency
effects in impression formation. Journal of Abnormal and Social
Psychology, 1964, pg, 335-338.

 

 

McNemar, Q. Psychological statistics. New York: Wiley & Sons, 1962.

 

Miller, A. G. Psychological stress as a determinant of cognitive com-
plexity. Psycholggical Reports, 1968, 2;, 635-639.

 

Miller, A. G. Amount of information and stimulus valence as determin-
ants of cognitive complexity. Journal of Personality, 1969,
37, 141-157.

 

Miller, H., and Bieri, J. Cognitive complexity as a function of the
significance of the stimulus object being judged. Psychological
Reports, 1965, lg, 1203-1204.

 

Nidorf, L. J., and Crockett, W. H. Cognitive complexity and the inte-
gration of conflicting information on written impressions. Journal
of Social Psychology, 1965, pp, 165-169.

 

Reich, J. W. Affect and frequency of interaction as determiners of the
complexity of interpersonal perception. Perceptual and Motor
Skills, 1969, 28, 107-113.

Rothman, L. K. The "vigilance hypothesis" and psychotherapists.
Journal of CounselingyPsychology, 1973, 20, 169-175.

 

Scott, W. A. Cognitive complexity and cognitive flexibility. Socio-

61

Scott, W. A. Structure of natural cognitions. Journal of Personality
and Social Psychologx, 1969, ll, 261-278.

 

Shepherd, J. W. The effects of variations in evaluativeness of traits
on the relation between stimulus affect and cognitive complexity.
Jourpgl of Social Psychology, 1972, 88, 233-239.

 

Shrauger, S. Cognitive differentiation and the impression—formation
process. Journal of Personality, 1967, 88, 402-414.

 

Shrauger, S., and Altrocchi, J. The personality of the perceiver as a
factor in person perception. Psychological Bulletin, 1964, 88,
289-308. '

 

Soucar, E. Students' perceptions of liked and disliked teachers.
Perceptual and Motor Skills, 1970, 8l, 19-24.

 

Soucar, E., and DuCette, J. Cognitive complexity and political pre-
ference. Psychological Rsports, l971,_29, 373-374.

 

Soucar, E., and DuCette, J. A re-examination of the vigilance hypo-
thesis in person perception. Journal of Social Psychology,
1972, §__8_' 31-36.

 

Tagiuri, R. Person perception. In G. Lindzey and E. Aronson (Eds.),
Handbook of social_psychology, Vol. 3. Reading, Mass.: Addison-
Wesley, 1969.

 

Tripodi, T., and Bieri, J. Cognitive complexity as a function of own
and provided constructs. Psychological Rpports, 1963, l8, 26.

 

Turner, R., and Tripodi, T. Cognitive complexity as a function of
type of stimulus objects judged and affective stimulus value.
Journal of Consulting and Clinical Psychology, 1965, 88, 182-185.

Vannoy, J. S. Generality of complexity-simplicity as a personality
construct. Journal of Personality and Social Psychology, 1965,
8, 385-396.

Wilkins, G., and Epting, F. Cognitive complexity and categorization
of stimulus objects being judged. Psychological Reports, 1971,
22' 965-966 0

Wilkins, G., Epting, F., and VanDeRiet, H. Relationship between repres-
sion-sensitization and interpersonal cognitive complexity.
Journal of Consulting and Clinical Psychology, 1972, 88, 448-450.

APPENDICES

APPENDIX A

PROVIDED BIPOLAR CONSTRUCTS ON THE REP TEST

APPENDIX A
PROVIDED BIPOLAR CONSTRUCTS ON THE REP TEST
(MEAN LIKABLENESS RATINGS ACCORDING TO ANDERSON, 1968)

(SCALE or 0-6)

EVALUATIVE TRAITS (Shepherd, 1972)

Sincere (5.73) Insincere (.66)
Kind-hearted (5.14) Mean (.37)
Interesting (5.11) Dull (1.21)
Friendly (5.19) Hostile (.91)
Trustworthy (5.39) Untrustworthy (.65)

Additional traits:

Open-minded (5.30) Intolerant (.98)
Reasonable (5.00) Unreasonable (.97)
Cheerful (5.07) Ill-tempered (.95)
Thoughtful (5.29) Thoughtless (.77)
Courteous (4.94) Rude (.76)

NON-EVALUATIVE TRAITS (Shepherd, 1972)

Talkative (3.52) Quiet (3.11)
Methodical (3.25) Unmethodical (2.62)
Impulsive (3.07) Cautious (3.34)
Bold (3.36) ' Shy (2.91)
Rebellious (2.58) Conventional (2.60)

Additional traits:

Outspoken (3.13) Reserved (3.48)
Opinionated (2.57) Discriminating (2.83)
Fearless (3.66) Hesitant (2.90)

Suave (3.35) Blunt (2.87)
Nonchalant (3.24) Serious (3.79)

62

APPENDIX B

SAMPLE PAGES AND SCORING GRIDFORM FOR THE

REP TEST--GROUPS 1 AND 3

APPENDIX B

SAMPLE PAGES AND SCORING GRIDFORM FOR THE

REP TEST--GROUPS 1 AND 3

 

Instructions

At the top of each of the fbllowing pages you will find a word or
phrase describing or indicating a person in a certain relation to
you. You are to think of 92§_person whom you know who is appropriate
and place this person's initials in the blank to the right. Then
complete the page with that particular person in mind.

You will first find a set of rating scales with an adjective at each
end. Look at both adjectives first, and then decide how you would rate
that person on a scale of l to 6. For example: ‘

Aggressive l 2 3 4 5 6 Passive
If the person is more aggressive than passive, then circle 1, 2, or 3,
depending on the degree of aggressiveness you perceive this person to
have. Please do not omit any of the ratings. Answer as well as you

can.

Following the rating scales will be a few general questions: they are
self-explanatory.

Please check with the experimenter if you have any questions. Of
course, all your responses will remain confidential. There is no
time limit; work at your own pace.

Thank you for your cooperation and time.

63

64

Person You Know Slightlypand Feel Neutral Toward

Please rate this person:
Sincere l
Thoughtless
Suave
Courteous
Dull
Open-minded
Friendly
Discriminating
Impulsive
Mean
Cheerful
Quiet
Bold
Trustworthy
Conventional
Nonchalant
Unreasonable
Fearless

Methodical

P' P‘ h' h‘ h‘ k‘ F‘ P‘ k‘ P‘ P‘ P‘ h‘ h‘ ha h‘ h‘ k‘ b1

k) k: k) m: n: no a: k: u: n: k: k) k) k) k) n: k) h) k) n)
u: u: u: u: u» u: U1 U1 U2 U1 U1 U1 U1 U1 u: U1 U1 u: u: u:
n. a. b» p. b. a. a, h. n. a. e. p. p. p. a. a. a. n, A. a
U1 U1 U1 U1 U1 U1 U1 U1 U1 U1 U1 U1 U1 U1 U1 U1 U1 U1 U1 U1

Reserved

How much do you like this person?
1 2 3 4 5 6 7 8 9
like
VEIY
much
How frequently do you interact with this person?

1 2 3 4 5 6 7 8 9
frequently

How involved are you with this person?

1 2 3 4 5 6 7 8 9

very
involved

mmmmmmmmmmm-mmmmmmmmm

Insincere
Thoughtful
Blunt

Rude
Interesting
Intolerant
Hostile
Opinionated
Cautious
Kind-hearted
Ill-tempered
Talkative
Shy
Untrustworthy
Rebellious
Serious
Reasonable
Hesitant
Unmethodical
Outspoken

10
dislike
very
much

10

infrequently

10

not involved

at all

65

SAMPLE GRIDFORM FOR SCORING THE REP TEST

Sincere
htless
Courteous

Dul

Friendl

Mean

Cheerf

Trustwor

Unreasonable

Suave

Discriminat

lsi

Nonchalan

r

Methodi

 

Reserved

APPENDIX C

SAMPLE SHEETS AND INSTRUCTIONS--GROUPS 2 AND 4

APPENDIX C
SAMPLE SHEETS-~GROUPS 2 AND 4
Father

How much do you like this person?

1 2 3 4 5 6 7 8 9 10
like dislike
very very
much much

How well do you think you know this person?

1 2 3 4 5 6 7 8 9 10
very not well
well at all

How involved are you with this person?

1 2 3 4 5 6 7 8 9 10
very not involved
involved at all

Person You'd Like to Help

 

 

How much do you like this person?

1 2 3 4 5 6 7 8 9 10
like - dislike
very very
much much

How frequently do you interact with this person?

1 2 3 4 5 6 7 8 9 10
frequently infrequently

How involved are you with this person?

1 2 3 4 5 6 7 8 9 10
very not involved
involved at all

66

67
INSTRUCTIONS FOR PARTICIPANTS IN GROUPS 2 AND 4

Here are descriptions of six (ten) persons. I'd like you to
think of six (ten) persons that you know that fit these descriptions.
Please tell me the initials of the six (ten) persons you think of.

Now, I'd like you to go back to each of the persons, one at a
time and in the order they're arranged here, and list single word
adjectives to describe each person, as many as you can think of.

Write each adjective on a separate slip of paper. In the end you'll
have six (ten) piles of adjectives, one pile for each of the six (ten)
persons.

When you have finished, please leave the piles underneath the
appropriate card. Then call me and I'll introduce a second task for you.
If you need anything, please call me.

Take the stack of words for each of the persons you have described
(in the correct order, of course) and look over the set of words you
have used in describing each person. Then put together into groups
the separate words which seem to go together. You may have as many or
as few groups as you like, and you may have as many or as few words in
a group as you like, so long as the words in each group belong together
for one particular reason. If, after you have throught about the words,
a few do not seem to belong with any of the others, you may put those
into a group by themselves. Of course, your groupings may change from
one person to another.

In brief, then, you are to take the set of words describing a
person, look over that set of words, and sort the words into whatever
groups you wish. Please do this sorting and grouping separately for
each of the six (ten) persons in the same order that you followed in
describing them;

For each person, after you have sorted the words into their separate
groups, you should identify each group; do this by writing a brief de-
scriptive word or phrase or sentence which describes that group of words,
the reason you had for making it a group. Put the label slip with the
reason on tOp of the group of words, then put a paper clip on that group
so that the slips do not become separated. Do this labeling for each of
the groups of slips for each of the persons you've described.

Are there any questions about what you are to do?

APPENDIX D

CORRELATIONS FOR INDIVIDUAL PARTICIPANTS IN GROUP 1

APPENDIX D

CORRELATIONS FOR INDIVIDUAL PARTICIPANTS IN GROUP 1

 

 

 

Liking 6 Knowledge & Liking & Knowledge & Com.-Eva1. &
Complexity- Complexity- Complexity- Complexity- Complexity-
Evaluative Evaluative Noneval. Noneval. Noneval.

-.407 -.277 -.521 -.356 .756*
-.668 -.101 .011 .214 -.066
-.458 -.480 .214 .320 .030
-.l43 .218 -.308 -.520 -.483
.153 .106 -.406 .213 .204
-.563 ~.ll4 -.464 -.415 .133
.650 .155 -.185 .200 -.364
-.359 -.229 .299 .456 .461
-.706 -.556* -.529 -.199 .336
-.502 .421 .155 .151 .131
.456 .011 -.267 -.014 .169
-.405 -.020 -.351 .471 .282
-.396 -.487 .169 .284 .348
-.573 -.371 -.158 .035 .101
-.606 -.302 -.305 .068 .261
-.304 -.267 .245 .208 -.6S9*
-.S76 -.389 .255 -.271 -.101
.468 .408 -.197 -.564 -.236
-.358 -.l74 -.453 -.482 -.386
.052 .186 -.516 -.573 -.475
-.584 -.313 .602 -.042 -.418
.330 .115 -.061 -.759* .064
-.237 -.l70 -.224 .116 .512
-.268 -.267 .205 -.353 -.306
-.816* -.665* -.312 -.141 .471
-.342 -.216 .436 .362 -.509
-.732* -.354 .351 .881* -.436
--040 -.319 .099 .281 -.539
-.554* . -.499 -.220 .337 .255
-.464 -.373 .311 .465 -.238

 

* p < .05, two-tailed test
N - 10

68

 

APPENDIX E
ADDITIONAL DATA FOR GROUP 2

(BIERI'S TEN ROLE FIGURES AND H/Hmax MEASURE)

APPENDIX E

DISCUSSION OF CORRELATIONS AMONG RATINGS

AND COMPLEXITY SCORES FOR GROUP 2 ROLE FIGURES

Number of Adjectives. The correlation between liking ratings and
number of adjectives written in response to the stimulus person for each
of the role figures was negative in all cases except three (Self, Boss
and Person with Whom You Feel Most Uncomfortable); in such cases the
relationship was essentially zero. For the former role figures, the
negative correlation indicated that more adjectives were written in
describing well-liked role figures. The correlation between liking and
number of adjectives for each individual participant was a negative
correlation except for four cases.

The relation between number of the adjectives and knowledge
ratings was a negative one for seven of the role figures, indicating
that more adjectives were written as ratings indicated greater knowledge
of the stimulus person. For all but five participants the correlation

between number of adjectives and knowledge ratings was negative also.

 

Number of Categories. The data for number of categories showed a

pattern very similar to that for number of adjectives.

69

on n z
umou poawwulosu .mo. v a «

 

 

 

 

 

 

 

70

 

 

 

 

 

 

 

«was. nmo.1 moo.1 «mum. «oo~.1 kmoH.1 msofiumamuuool
mwmum><
«awn. me.1 mma. kooq. mmm.1 omo. suwz .maooss
Home sow somumm
«own. mwo. woa.1 «new. omo.1 oom.1 mama ou exam
p.30» nomuom
kmmm. 0mm. «no.1 «mas. omo. mmH.1 pamumuwpa: on
uasofiwmwo domuwm
«awn. oma. «Ho. «mmq. Nom.1 Hoa.1 mxHHmHo
sow somuom
«new. Hue. mmm. «mum. qwa.1 oqo. mmom
«mow. «mwm.1 «Noc.1 «mum. «mqq.1 nqm.1 xmm mafimommo
mo sunfish
«mmm. AHA. mmo.1 «cam. mmo.1 mmH.1 sum 086m
mo sunfish
ammo. wm~.1 num.1 med. Hm~.1 mam.1 umnumm
«cam. «mmq.1 qum.1 «mm. Nom.1 o-.1 umnuoz
«Ham. mmo.1 oHH.1 mew. qu.14 oqo.1 wawm

xmem\z .660 .oz .umo .oz xmam\: .ne< .oz yfie< .oz

8 .umo .02 a .3oax a wcfixaq a .fie< .oz 8 .3oas a masses

 

mHMDUHm MACK N mDQmU mom mmmOOm MBHXMAQZOU 92¢ mUZHB¢m 02024 ZCHfiéflmmmco

 

71

CORRELATIONS FOR INDIVIDUAL PARTICIPANTS IN GROUP 2

 

Liking &

 

 

*

IIA

P
N

.0
10

5, two-tailed test

Liking & Knowledge & Knowledge & No. of Adj. & ,
No. of Adj. No. of Adj. H/Hmax H/Hmax H/Hmax _;
l
.667* -.365 -.453 -.486 -.393
-.116 .114 .090 .232 .128
-.054 -.353 -.118 -.169 -.248
-.096 -.542 -.259 -.547 .709 1
.763* -.082 .000 .000 .000 §
-.460 -.349 -.448 .038 .126 2
-.582 -.375 -.818* -.122 .449 z
-.477 -.703* .511 .523 -.030 i
-.229 -.604 .193 -.021 .538 1
-.215 -.246 .001 .396 .075 g
-.391 -.067 -.811* -.359 .560 g
-.424 -.319 -.744* -.204 .424 ;
-.016 -.123 -.524 .108 —.027 1
-.060 -.164 -.452 -.029 .388
-.292 -.819* .016 -.830* .781* '
-.753* -.689* -.150 -.410 .478 g
-.342 .045 -.567 -.289 .634 !
-.524 -.701* -.330 -.724* .841* g
-.776* -.597* -.198 -.283 .203 s
-.825* -.816* .392 .299 -.070 .
-.445 .205 -.197 .238 -.098 ;
-.S87* -.805* -.156 -.427 .343 §
-.241 -.767* -.117 -.320 .700* 9
-.813* -.733* .049 .253 .194 §
-.336 -.524 .389 .031 .239 1
-.267 -.426 -.270 -.449 .876* g
.161 -.545 -.060 -.133 -.061 ;
-.355 -.389 -.023 .244 .462 z
-.399 -.457 .129 -.058 .378 1
.253 -.204 -.270 -.230 .318

APPENDIX F
ADDITIONAL DATA FOR GROUP 3

(REICH'S SIX ROLE FIGURES AND REP TEST)

APPENDIX F

MEAN RATINGS FOR GROUP 3 ROLE FIGURES

 

Liking Interaction Involvement Knowledge

 

 

 

 

 

 

 

 

 

 

Know Well & Like 1.5 2.17 2.00 1.63
Know Well & Dislike 8.03 5.90 6.53 3.70
Know Well & Feel

Neutral Toward 3.93 4.43 5.33 3.27
Know Slightly &

Like 2.73 3.77 5.17 5.13
Know Slightly &

Dislike 7.93 7.17 8.37 6.37
Know Slightly &

Feel Neutral Toward 4.60 4.90 6.83 6.73
Overall

Means 4.79 4.72 5.71 4.47

 

72

 

73

om u z
ammo ouaamauoSU .mo. v a «

 

 

 

 

 

 

 

 

«ome. wno. HmH.- ka.- «858.- uumsoa Hmuusmz
a saunwwam sous
ems. wee. 465m. Heat: moo. mxsflmga
a sauawfiam some
New. mma. oAN. mma. ooo.: menu
a sHuawuHm 366M
«no. seen. 500. qu. moo. oumsoe Hmuusoz
a saw: some
mNN. mks. wme. meoq. mqfi. mxfiamua
a HH63 some
ems. soc. mm~.1 «80.- “Hm.1 we“;
a HH63 266M
.Hm>maoZI.Eoo .Hm>mco21.sou .Hm>maoZI.Eou .Hmbm1.800 .Hm>mI.Eoo
a .Hm>mu.aoo a .3ocs a manage a .3oae a waaxsg

 

mMMDUHh WHOM m m30m0 mom mmmOUm NBmequOU 92¢ mUZHfifim UZOEd WZOHfidqmmmoo

 

APPENDIX G
ADDITIONAL DATA FOR GROUP 4

(REICH'S SIX ROLE FIGURES AND H/Hmax MEASURE)

APPENDIX G

MEAN RATINGS FOR GROUP 4 ROLE FIGURES

 

 

 

 

 

 

 

 

 

 

 

 

Liking Interaction Involvement Knowledge
Know Well & Like 1.33 1.87 2.13 2.03
Know Well & Dislike 8.23 5.50 6.80 3.10
Know Well & Feel
Neutral Toward 4.50 4.20 5.40 3.93
Know Slightly &
Like 2.53 4.33 5.23 5.53
Know Slightly &
Dislike 7.73 7.27 8.60 7.33
Know Slightly & Feel
Neutral Toward 4.83 6.30 7.27 6.70
Overall ' 3
Means 4.85 4.92 5.91 4.79 ‘

 

74

omuz
ummu omafimulosu .mo. v a «

 

 

75

 

 

 

 

 

 

«kmm. NmH.- «ma. 4Aee. omo.- HOH.I HNH.- moo. uumsoa Hmuunmz
a sauamfiam sons
«Ame. sma.- owm. 4oem. qw~.u 88m. nos.- ham. «engage a
manawgam 366x
48H“. mmo. «No.1 mwm. HNH. Nam. Nqo.- mwN.- mean a
sauewfiam sons
«H85. Hmm.1 omm.- 4amo. 4~m8.- mwm.- 4emm.- «aam.u eumson Hmuusaz
a Hams 366s
«886. moN.- mom. com. mNN.1 ems. on.- mmo. «shaman
a Hams 366M
4Hmo. Noa.- mac. mam. «Ho.- Hoe. 46Hm.- eem.- mean
a Hams 366M

xmam\m .860 .oz .umo .oz xmam\m xmaz\m xmem\m .fiea .oz .naa .oz

8 .umo .oz 8 .366M 8 wanxfia a .ne< .oz 8 .3662 a wages; a .366“ w wafixfla

mmmDOHm mqom v mDomU mom mmmOUm NEHxMAmEOU DZd mUZHfidm UZOEd mZOHEGAmmmOU

 

GROUP 4

DISCUSSION OF NUMBER OF ADJECTIVES AND CATEGORIES

Number of Adjectives. The analysis of variance showed a signi-

 

ficant main effect for Knowledge (F(2/28) = 21.127, p < .01), Affect
(F(2/56) = 9.746, p < .01) and a significant Knowledge X Affect
interaction (F(2/56) = 6.4067, p < .01). Participants wrote signi-
ficantly more adjectives for well known persons (M = 6.38) than for
slightly known persons (M = 4.81). Significantly fewer adjectives
were written for neutral persons (M = 4.67) than for liked (M = 6.38)
or disliked persons (M = 5.733). The latter two means did not differ
significantly. The means for the Know Slightly role figures were not
significantly different from each other, whereas each of the means
for the Know Well role figures differed significantly from each of the
others [Newman-Keuls, p < .05]. The means for the two neutral role
figures did not differ significantly from each other nor did they
differ significantly from the means for the other Know Slightly role
figures.

Number of Categories. There was a significant main effect for

 

Knowledge (F(1/28) = 28.64, p < .01). Participants created signifi-
cantly more categories for well known persons (M = 2.44) than for
slightly known persons (M = 1.89). Also significant was a Knowledge

x Affect interaction (F(2/56) = 4.24, p < .05). The means for the

Know Slightly role figures did not differ significantly from each other,
while for the Know Well role figures, the mean for liked persons

76

77

(M

2.76) differed significantly from the mean for neutral persons

(M 2.10), as shown by a Newman-Keuls test with p < .01. The means
for the two neutral role figures did not differ significantly from each
other and did not differ significantly from the means for the other
Know Slightly role figures. (The interaction was very similar to the

interaction indicated with the number of adjectives.)

Group 4

Means and Standard Deviations of Number of Adjectives

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Role Figures Liked Disliked Neutral
Known Well M 7.70 M 6.37 M 5.07 6 38
SD 3.67 SD 3.10 SD 2.57 '.
Known Slightly M 5.07 M 5.10 M 4.27 4 81
SD 3.70 SD 2.95 SD 2.63 '
6.38 5.73 4.67
Group 4
Means and Standard Deviations of Number of Categories
Role Figures Liked Disliked Neutral
Known Well M 2.77 M 2.47 M 2.10 2 44
SD 1.10 SD .82 SD .88 '
Known Slightly M 1.87 M 1.93 M 1.87 1 89
SD 1.07 SD .87 SD .86 '
2.32 2.20 1.98

APPENDIX H

OVERALL CORRELATIONS FOR EACH GROUP

'78

.omauz .n:o«uadwuuoo qua non
.OOMIZ .n:o«umaouuoo uoauo «do

n

 

 

 

you .Ohulz .ucwao>ao>=H ocq>~o>cu uneducaouuoo non .o-lz .coauouuoucu ocu>uo>cw uneducaouuoo noun
.33 6033195 .3. v a.
.94.. .3... AR... «3.. 21- So. gm
.wmo. onun.1 c¢-.1 cuo~.1 mmo.1 .umu .oz
.821 .39- .68... 311 .34 6:
cmmm. cm~v. aowm. .3ocx
save. «com. .HOSCH
.hmv. .umucu
be moozu
«omm. Hod. -o.1 ovo.1 660.1 .~s>o:021.loo
noo.1 caum.1 cooH.1 coon.1 .Ha>u1.lou
chem. some. aomw. .305
«amp. «mom. .HoacH
«mam. .uquH
An macaw
«Hon. coon. aama.1 am-.1 .non.1 cmo~.1 xusfi\=
«mom. cmhd... choN.I cONN.I «mo~.I .350 .02
.391 .38.- .62.- 63... .6: 6:
anew. anum. «mam. .xosx
«vow. coon. .HO>CH
doom. .uwu:H
UN moozu
Hmo. NHH. Hwo. woo. voo.1 .Au>0:021.lou
ooo.: mo~.1 avo.1 «whN.I .Hu>u1.floo
«one. «awn. came. .3ocu
some. comb. .Ho>:H
chmm. uoucn
0H macaw
.uuU.oz .nv¢.oz .Hu>fl1.ioo .Ibzu .Ao>:H .umucn ocuxun macho

 

 

 

 

 

 

 

 

gomw mug mom mZOHBngU EEO

m Noun—29mg

V

"111111114

“111111101111?

04116