PHENOMENOLOGICAL BASES FOR THE ATTRIBUTION
0F BALANCE T0 SOCIAL STRUCTURE

Thesis for the Degree of M. A.
MICHIGAN STATE UNIVERSITY
DAVID WAYNE GERBING
1976

 

 

 

 

 

ABSTRACT

PHENOMENOLOGICAL BASES FOR THE ATTRIBUTION
OF BALANCE TO SOCIAL STRUCTURE
BY

David Wayne Gerbing

The objective of this study was to examine how people
think about social situations. In particular, the takeoff
point of the paper was balance theory which asserts that
some patterns of social interaction are harmonious and
stable over time while other patterns are clashing and
transient. Many studies have attempted to test balance
theory in both its original and its revised forms by
presenting subjects with tasks such as the triadic comple-
tions task in which (1) the subject is told of a group of
people 9, p, and g; (2) the subject is told how 9 and p
feel about each other and how 2 and g feel about each other;
and (3) the subject is asked to predict how 3 and g_will
feel about each other. Astoundingly, none of these studies
has ever asked the subjects to report on the processes
which they use to derive their predictions.

Thirty-one subjects were asked to report their
reasons for making predictions to 33 hypothetical situations

and their answers were categorized. The results showed

David Wayne Gerbing

that subjects did ngt_use either balance theory, revised
balance theory, or transitivity of positive affect to
explain their answers. Instead, most of them explained
their predictions on the basis of inferences about common
interests. Most of the subjects ascribed to some version
of two principles: (1) one person will like another only
if he likes the other person's interests, and (2) two
people who like the same things will like each other.
Thus the fact that balance theory correctly predicted the

response 68% of the time is quite accidental.

Approved:

 

John E. Hunter
Committee Chairman

November, 1976

PHENOMENOLOGICAL BASES FOR THE ATTRIBUTION

OF BALANCE TO SOCIAL STRUCTURE

BY

David Wayne Gerbing

A THESIS

Submitted to

Michigan State University

in partial fulfillment of the requirements
for the degree of

MASTER OF ARTS

Department of Psychology

1976

ACKNOWLEDGMENTS

Scientific work does not proceed in a vacuum and
this paper is certainly no exception. Many people were
necessary and none alone was sufficient for the sequence
of events throughout my life which culminated in this
work. The list of acknowledgments begins with my
parents and grandparents and includes a handful of
teachers through high school, my undergraduate advisors,
and finally my graduate thesis committee consisting of
John Hunter, Ralph Levine, and Jim Phillips. The
chairman, John Hunter, was the largest single contributor
to this work. The largest single contributor to me,
Monica Gerbing, should be thanked for much more than just

the typing.

ii

TABLE OF CONTENTS
Page
LIST OF TABLES O O O O O O O O O O I 0 O O O O O O 0 iv

LIST OF FIGURES . . . . . . . . . . . . . . . . . . . V

INTRODUCTION . . . . . . . . . . . . . . . . . . . . 1
Definitions and Assumptions . . . . . . . . . . . 1
Balance Theory . . . . . . . . . . . . . . . . . . 3
Balance Theory Reconsidered . . . . . . . . . . . 5
The Transitivity of Positive Affect . . . . . . . 8
Evaluation of the Theoretical Perspectives . . . . 9
The Objective of This Study . . . . . . . . . . . 11

THE PHENOMENOLOGY OF THE TRIADIC COMPLETION TASK . . 12
MethOd O O O O O O O O O O I O O O O O O O O O O O 13
SUbjeCtS O O O O O O O O O O O O O O O O O O 0 13
Materials . . . . . . . . . . . . . . . . . . . 13
Procedure . . . . . . . . . . . . . . . . . . . 15
Results and Discussion . . . . . . . . . . . . . . 15
The completion response . . . . . . . . . . . . 16
Balance theory . . . . . . . . . . . . . . . . 18
The certainty of completion responses . . . . . 19
Phenomenological reports: the "++" situation . 21
Phenomenological reports: the "+-" situation . 22
Phenomenological reports: the "--" situation . 24
Conclusion . . . . . . . . . . . . . . . . . . . . 26

APPENDICES

Appendix A: The Implications of the Transitivity
of Positive Affect . . . . . . . . . . . . . . . . 29

Appendix B: Sample Stimulus Page . . . . . . . . 32

Appendix C: The Logic of the Shared Interest
Explanation . . . . . . . . . . . . . . . . . . . 33

LIST OF REFERENCES . . . . . . . . . . . . . . . . . 37

Table 1

Table 2

Table 3

LIST OF TABLES

Page
The Number of Different Completion Responses
as a Function of Structure . . . . . . . . . . 17

The Number of Subjects Who Responded to the
Series of Triads in Accordance with Balance
Theory . . . . . . . . . . . . . . . . . . . . 18

The Mean Uncertainty of the Completion
Responses . . . . . . . . . . . . . . . . . . 20

iv

Figure 1:

Figure

Figure

Figure

Figure

LIST OF FIGURES

The four connected triadic structures
for reciprocated, dichotomous affect.

A "+" and "-" represent positive and
negative affective relations
respectively . . . . . . . . . . . .

The interests of p, g, and q . . . .

The incomplete structure defined by
two positive relations . . . . . . .

The incomplete structure defined by
a positive relation and a negative
relation . . . . . . . . . . . . . .

The incomplete structure defined by
two negative relations . . . . . . .

Page

33

34

35

36

INTRODUCTION

Definitions and Assumptions

 

The processing of socially relevant information does
not ordinarily proceed whimsically or chaotically. Indeed,
it has frequently been assumed that the individual
possesses a systematic "plan"--a set of general principles
called cognitive schemata-—which guides this information
processing. These schemata provide the structure for the
perception, interpretation, and evaluation of social
phenomena. It is this structure which allows the individual
to understand and predict the nature of interpersonal
relations.

This paper considers various schemata that have been
proposed for the perceived patterns of affect in groups of
three people--the social triad. The affective or sentiment
relation is assumed to be dichotomous. Positive and
negative affect correspond to the liking and disliking
relations respectively--the feelings a person has toward
another person or object. This paper will consider only
"symmetric" triads, i.e. triads in which each pair of
people feel the same toward each other. It is also assumed
that each person is either positively or negatively related
to the other two people, i.e. no one is indifferent toward

another. The following three statements provide an example

1

2
of such an interpersonal structure composed of one positive
affective relation and two negative affective relations:
Person 2 and Person 9 like each other; Person 9 and
Person q dislike each other; and Person 2 and Person q
dislike each other.

Since the affective relation is assumed to be dichot-
omous and without indifference, there are four qualita-
tively distinct relational patterns, including the previous
example. All three relations may be positive, all three
relations may be negative, one relation may be positive
and the remaining relations negative, or one relation may
be negative and the remaining relations positive. These

patterns are illustrated in Figure l for the cognitive

P E
0 + q o - q

 

 

_Structure 0— _Structure 1—
B B
2 ' 3 9 ' 3
Structure 2 Structure 3

Figure l: The four connected triadic structures for
reciprocated, dichotomous affect. A "+" and "-" represent
positive and negative affective relations respectively.

3
elements p, g, and q. The labels assigned to the four
structures in Figure 1 (Structure 0, Structure 1, etc.)
correspond to the number of negative affective relations
in the structure. These labels are used to reference the
structures throughout this paper.

Many papers have now been devoted to the following
question: Are all of the structures shown in Figure 1
equally stable? Or are some of the structures likely to
produce forces which will change that structure into one
of the others? Heider (1944, 1946) originally labeled the
structures that he believed to be stable as "balanced".
However since that time, "balanced" has come to be identi-
fied with Heider's particular theory and other terms have
been coined by investigators who have proposed competing
theories.

‘A typical example of the reasoning underlying these
formal theories is the following hypothetical social triad.
In most families there would be considerable conflict if
a man who loves his parents plans to marry someone whom
his parents dislike (Structure 1 with g as "parents").
According to Heider's balance theory this conflict would
result in a shift to some "balanced" triad such as Structure
0 in which the parents have come to adopt a positive

feeling toward their prospective daughter-in-law.

Balance Theory

 

The first formalization of affective consistency is

Fritz Heider's "balance theory" (1944, 1946). The basic

4
explanatory principle for balance theory as expounded by
Heider is the unit relation. Some factors postulated to
lead to the formation of social units are similarity,
proximity, ownership, familiarity, causality, and liking.
For example, if Person p makes or owns Object x, p and x

are a unit. If Person 2 and Person 9 like each other, p
and g tend to be perceived as a social unit by others.
Conversely, negative affect is a force opposed to unit
formation.

According to Heider, a structure is balanced if each
pair of entities which are linked by a unit relation are
positively related to each other and are related in a
similar way to any other entity in the structure.

In applying Heider's principles t0\the triads in

Figure 1, there must first be an assumption concerning
unit relations. The assumption normally made is that
positive sentiment implies a unit relation. Given this
assumption, Structure 2 is easily shown to be balanced:
p and 9 have the only unit relation in the triad, and they
each have a similar view (i.e. negative) toward the other
entities (i.e. q). Structure 0 is also balanced: each
pair has a unit relation and each member of that pair feels
positively toward the third person. Structure 1 is
imbalanced: p and 9 have a unit relation but they disagree
in their feelings toward 3.

Heider (1946) labeled Structure 3 as imbalanced on

the basis of separate considerations (which he later

5

modified in 1958). "The case with three negative relations
does not seem to constitute a good psychological balance,
since it is too indetermined" (p. 110).

Cartwright and Harary (1956) developed a version of
balance theory to accommodate structures with any number
of entities. As a special case of that definition, the
rule is obtained for Heider's balance theory for triadic
structures in which all sentiments are reciprocated and in
which there is no indifference. Thus Cartwright and Harary
label Structures 0 and 2 as balanced and they label

Structures 1 and 3 as imbalanced.

Balance Theory Reconsidered

 

All three central figures in the development and
popularization of balance theory--Heider, Newcomb, and
Davis-—eventually expressed dissatisfaction with the
original formulation of balance theory. Heider (1958)
modified his position of 1946 by saying that the lack of
balance in Structure 3 is not as extreme as the imbalance
of Structure 1. Newcomb (1968) replaced his 1956 formu-
lation by'a new principle called "positive balance theory".
Davis (1967) reversed his 1963 endorsement and postulated
a new principle which he called "clustering".

In the present context (i.e., symmetric triads without
indifference), all three men say the same thing: Structure
3 is not inbalanced, only Structure 1 is inadmissible.

In 1958, Heider noted that Structure 3 has no positive

affect and hence no unit relation. Thus there is nothing

6
in Structure 3 to contradict his balance principle as
stated in terms of unit relations. Having noted this, he
expressed reservations in labeling it as "imbalanced". On
purely logical grounds, that which is not imbalanced must
be balanced. And Heider adopted this relabeling of
Structure 3, though not without a certain nostalgia for
his original position: "If two negative relations are
given, balance can be obtained either when the third
relation is positive or when it is negative, although there
appears to be a preference for the positive alternative"
(1958, p. 206).

Davis (1967) noted that balance theory has one
"unfortunate" implication: If a group is composed of
mutually hostile cliques, then according to balance theory
there can be at most two such cliques. Davis traced this
implication to Structure 3 which can be viewed as a set of
three mutually hostile cliques of size 1. He then proposed
his "clustering theory" which labels Structure 3 as
admissible.

Another affective consistency perspective concerned
with triadic structure is Newcomb's positive balance theory
(Newcomb, 1968; Price, Harburg & Newcomb, 1966). Newcomb
argues that under certain conditions a person may be
indifferent to the attitudes of another person toward a
common attitude object (which may be another person). If
such indifference, or low engagement exists, the situation

is defined by Newcomb to be nonbalanced. A nonbalanced

7
triadic structure represents a state of neutrality inter-
mediate to Heider's balance and imbalance. A nonbalanced
state is characterized by an indifference which is neither
pleasant and hormonious nor unpleasant and unstable.

Newcomb does not provide a substantive derivation of
what constitutes engagement, though he does indicate that
"engagement corresponds, loosely at least to a unit
relationship" (Price et al., 1966, p. 268). It is simply
stated as a hypothesis that a positive affective inter-
personal relation leads to concern about the other person's
attitude toward a common attitude object while negative
affect implies a lack of interest in the other person's
attitude. In order to differentiate his three-category
classification scheme of balance and imbalance with
engagement, and nonbalance with low engagement, Newcomb
(1968) introduces the terms positive balance, nonbalance,
and positive imbalance.

Unlike the affective consistency schemata previously
discussed, Newcomb's schema requires that the consistency
of a triadic structure be evaluated from the perspective
of an individual within the structure. For Person 2,
Structure 0 is positively balanced since two people whom
p cares about like each other. Structure 2 is also
positively balanced since the one person p cares about, 9,
also dislikes 3. Because the two people p cares about
dislike each other, Structure 1 is positively imbalanced.

Finally, Structure 3 is nonbalanced since 2 does not care

8
about 9'3 attitude toward q or q's attitude toward 2. The
classification of the four structures remains the same from
Person 9's and Person q's perspectives, except for
Structure 2 which is nonbalanced for Person q who is
negatively related to both E and o.

How does the revised balance theory compare to the
original? If no sentiment is indifferent, then the differ-
ence between the two approaches can be seen most easily at
the level of Cartwright and Harary's structure theorem.

They proved that a structure is balanced if and only if

the entities may be partitioned into (at most) two

mutually exclusive subsets such that all of the entities
within a subset or "clique" are positively related and all
relations between members of different subsets are negative.
The revised versions of balance theory also imply clustering
with positive relations among group members and negative
relations between members of different cliques as a prere-
quisite for consistency. The difference is that the

revised theories do not restrict the number of mutually

hostile cliques to two.

The Transitivity of Positive Affect

 

Davis ultimately became unhappy even with his revised
version of balance theory and went on to yet a third struc-
ture principle: transitivity of positive affect (Davis &
Leinhardt, 1970; Holland & Leinhardt, 1971; Holland &

Leinhardt, 1976). Positive affect is transitive in a

9
group if whenever A likes B and B likes C then A_likes g.
This principle differs from balance theory in two ways:
First, transitivity does NOT predict that sentiments will
tend to be reciprocal. In fact, the adherents of transi-
tivity believe that unreciprocated sentiment relations are
critical to the study of sociometry since unreciprocated
positive regard forms the basis of social status within
the group. Second, transitivity also permits three or more
negatively related groups.

For purposes of this paper, the critical fact is that
for symmetric triads without indifferent sentiments, the
only intransitive triad is Structure 1. That is, in the
present context, transitivity is equivalent to the other
revised balance models. A more detailed discussion is

presented in Appendix A.

Evaluation of the Theoretical Perspectives

 

It is very difficult to evaluate the conceptuali-
zations of affective consistency by reviewing the many
diverse empirical studies. The absence of regularity in
method across the studies of affective consistency prevents
a meaningful comparison of the obtained results.

At the first level of disparity, there are three
prominent measures used to assess the "consistency" of a
structure. In a triadic completion study (e.g.,
Morrisette, 1958), the subject is given two of the rela-
tions of a triadic structure and is asked to supply the

missing relation. In an affective rating study the entire

10
triad is presented. The subject is asked to rate the
pleasantness of the structure (e.g. Price et al., 1966).
The dependent variable in a learning of structures study
is the number of trials it takes the subject to learn the
relations comprising the structure. A trial consists of
the presentation of each of the relations in the structure
followed by the subject's attempt to recall the relations
(e.g. DeSoto, Henley & London, 1968). Those triads which
are most often formed by completion, rated the most
pleasant, or learned the fastest are expected to be the
"balanced" triads.

The problem is that the construct validity of these
measures has never been systematically determined. That
is, the precise psychological characteristics measured by
each dependent variable and the relation between the three
dependent variables are not adequately understood. For
example, Price et a1. (1966) exclusively employ the
affective rating method in their study which provided
empirical support for positive balance theory. But
recently it has been suggested that the triadic completion
method and affective rating methods measure different
constructs. Fuller (1974) maintains that "a distinction
can readily be drawn between the tension induced by the
perception of inconsistency or imbalance and the tension
induced by the perception of balanced structures which
contain unpleasant elements such as negative interpersonal

relationships" (p. 804).

11

A related problem is that even those studies utilizing
the same general methodology usually differ in many
specific details. For example, the subject may or may not
be asked to identify with one of the members of the triad;
the relations may or may not be reciprocal; the third
element of the triad may be an impersonal entity or another
person; if the third entity is impersonal, it may or may
not be specified. Again, meaningful comparison is diffi-
cult because the theoretical relevance of these variations

is not understood.

The Objective of This Study

 

The lack of comparability in the various empirical
studies stems largely from the fact that the investigators
in these studies did not query their subjects carefully as
to their understanding of the instructions or as to their
method of executing the task set them. Thus if a subject
answers in a fashion not predicted by balance theory, there
is no basis for assessing the reason for the departure.

The present study is intended to fill that gap, i.e.
to assess the subjects' phenomenal reports as to why they
answered the way they did. Since this study was
necessarily exploratory, only one of the types of balance
theoretic studies was considered here: the triadic

completion task.

THE PHENOMENOLOGY OF THE TRIADIC COMPLETION TASK

The main purpose of this study is the investigation
of the phenomenology of triadic structure of interpersonal
affect. Of particular interest is the logic of the indi-
vidual's responses--the language by which he conceptualizes
triadic structure. One concern is the ability of the
individual to present valid, logically consistent justifi-
cations for his responses. If so, then can clusters of
explanations be defined such that the explanations in each
cluster have similar meanings? If this task can be accom-
plished, some of the ways in which triadic structure can
be conceptualized will be better understood.

In the triadic completion task, the subject is given
two of the sentiment relations in a social triad and is
asked to predict the third. According to balance theory,
the subject should react to two positive relations by
completing the triad with a third positive sentiment; he
should respond to a positive and a negative sentiment by
completing the triad with a second negative relation; and
he should respond to two negative sentiments by completing
the triad with a positive sentiment. The revised balance
theories agree on all but the third case in which the two
given sentiments are negative. The revised theories would
accept either a positive or a negative (and hence an
indifferent) sentiment in completing the triad since either
Structure 2 or Structure 3 is admissible.

12

13

From the point of view of the revised balance theories,
the choice of response categories should be very important.
If the subject faced with two negative sentiments is
forced to choose only between "like" and "dislike", then he
would probably choose "like" and would thus be indistin-
guishable from a subject governed by balance theory.
However if the subject is allowed to choose from a full
response range, then according to revised balance theory
many of the subjects should react to the openness of
choosing either "like" or "dislike" by choosing
"indifferent". Therefore this study included triads with
both kinds of response options. When the subject was
restricted to only two categories, he was also asked to
rate his confidence in that response. Subjects also rated
their confidence in their 5-point responses.

For purposes of later analysis, the study included
triads of mixed sex so as to provide data on the phenom—

enology of the attribution of jealousy.

Method

Subjects. Participants in the study were 19 females
and 12 males from introductory psychology classes at
Michigan State University. The students were volunteers
who received extra-credit points for their participation.

Materials. The instrument used by the subjects to

 

record their responses was a booklet of 33 pages, one
stimulus situation per page. (An example appears in

Appendix B.) Each stimulus situation consisted of a

14
written description of a social triad with only two given
symmetric, dichotomous sentiments and a diagrammatic
representation of this triad. Except for three triads,
the possible responses to fill in the missing, symmetric
relation were "dislike very much", "dislike a little",
"indifferent", "like a little", and "like very much". The
certainty of the response was measured on a 5—point scale
anchored by "not at all certain" and "certain". The last
question for each triad was a request for the subject to
explain his reasons for the completion and certainty
responses.

The structures on the first 24 pages represented a
factorial combination of two variables. The primary
variable was defined by the three incomplete structures
defined by the possible pairs of dichotomous, symmetric
relations. That is, each incomplete structure consisted
of two given positive relations, a given positive relation
and a given negative relation, or two given negative
relations. The second variable was defined by the eight
permutations of sex of stimulus person which was indicated
by the use of 18 different unambiguous names of each sex.

Pages 25-27 differed in two ways from the first 24
pages. Although each of these triads represented one of
the three possible incomplete structures, the stimulus
persons of these triads were referred to as Person A,
Person B, etc., i.e. without sex labels. The other differ-

ence was the range of possible completion responses. For

15
these three incomplete triads, the only possible completion
responses were "dislike" and "like".
The remaining six triads were a duplicate of the first
six triads.

Procedure. Subjects participated in the study in four

 

small groups. As each subject entered the appropriate
room, he was handed a written copy of the directions.

When the study was scheduled to begin, the experimenter
read the directions to the entire group and asked for ques-
tions during and after the reading.

Subjects were told that the purpose of the study was
to examine their perceptions of the liking and disliking
relations among groups of three peOple. The nature of
the triadic completion task, including the composition of
the triads in terms of the incomplete structural relations
and the sex of stimuli, was explained. Subjects were also
told that the name of each stimulus person was not
important except to indicate the sex of the person
referred to.

' The task was untimed. When each subject completed the
booklet, he individually turned the booklet in. Each
session was complete when the last subject at that session
finished. The approximate minimum and maximum completion

times across all subjects were 30 and 90 minutes.

Results and Discussion

 

The data for two sets of triads were analyzed: the

three all-male triads with the three—alternative response

16
range (R3) and the three sexless triads using the two-
alternative reSponse range (R2). The data for two subjects
in the two-alternative response range were not analyzed
because of missing data. These two subjects indicated
that they could not meaningfully restrict their responses
to just a positive and a negative alternative.

The completion response. In order to fit the data

 

analysis to the traditional balance theory in terms of
qualitative states, the 5-point response scales were
collapsed to three categories: "like", "indifferent", and
"dislike". The relation each triad was completed with was
scored as follows. A response of "like a little" or "like
very much" was scored as a "+". A "0" represents
"indifferent", and a "dislike a little" or "dislike very
much" was scored as a "-". For the computation of the
mean and standard deviation, a "+", "0", and "-" were
numerically scored as a +1, 0, and -1 respectively. The
completion responses for both sets of triads appear in
Table l.

A repeated measures analysis of variance was performed
for both data sets. For the R3 condition, {(2,60) = 41.74,
p_§ .01. The differences between means were evaluated by
the Dunn test. The critical mean differences for three
comparisons at p_ 5; .05 and p 5; .01 for 60 d_f are 2.47 and
3.06 respectively. The appropriate transformation of the
mean differences for the three possible pairwise comparisons

all exceeded 3.06.

17

Table 1

The Number of Different Completion Responses
as a Function of Structure

 

 

 

 

 

 

Male triads, three-alter- Two alternative-

native response range response range

Given relations Given relations
Response ++ +- -- ++ +- --
+ (1) 30 l 18 29 7 25
0 (0) l 15 7 — - _
- (-l) 0 15 6 0 22 4
Mean .97 -.45 .35 1.00 -.52 .72
S.D. .18 .57 .84 0.00 .85 .70

 

For the R2 condition, F(2,60) = 42.26, p_é;.01.
Multiple comparisons were again conducted using the Dunn
test. In this condition, the transformed mean difference
between the completion responses of the two positive rela-
tions and the two negative relations was l.63--a value
which does not approach either critical value. The
remaining comparisons, however, easily exceeded p 2 .01.

At first glance, the data in Table 1 appear to provide
almost perfect support for transitivity since not a single
subject responded to two given positive sentiments by
completing the triad with a negative relation. However
transitivity is also assessed by the "+-" triads, since to
complete this triad with a positive response is to leave
an intransitive triad. In the forced choice R2 condition,

7 out of 29 subjects did just that. In the free response

18
triads, only 1 subject gave a positive completion, but 15
were neutral.

The "--" triad did not prove as unpredictable as would
have been expected by the revised balance theorists. In
the forced choice condition, only 4 of 29 subjects completed
the triad with a negative sentiment to produce the unbal-
anced triad with all negative sentiments (which is fewer
"inconsistencies" than were observed for the "+-" triads).
In the free choice condition, 18 of 31 gave the predicted
positive prediction, 7 were neutral, and 6 gave the nega—
tive response which produces an unbalanced triad.

Balance theory. Balance theory predicts the subject's

 

response to all three questions, i.e., the entire config-
uration of the subject's responses. How many subjects
answered uniformly in accord with the predictions of balance

theory? The answer to this question is given in Table 2.

Table 2

The Number of Subjects Who Responded to the Series
of Triads in Accordance with Balance Theory

 

Forced choice

 

 

 

Free Not all All

choice balanced balanced Total
All

balanced 3 5 8
Not all

balanced 6 15 21

Total 9 20 29

 

19

In the forced choice condition, 69% of the subjects
conformed uniformly to balance theory; but in the free
response condition only 28% of the subjects followed
balance theory perfectly. In comparing the two response
formats, it is clear that allowing neutral responses lead
to the subjects expressing much more disagreement with
balance theory. This fact is brought out in another way
by considering the subjects' response patterns to both
response conditions simultaneously. The correlation
between the responses which fulfill balance theory in the
two conditions is -.09. Only 5 of the 29 subjects followed
balance theory in all 6 responses.

It is tempting to try to explain the previous results
on the basis of error of measurement in the response.
However, balance theory predicts that there will be an
unequivocal reaction to each of the three situations and
hence no error measurement. Revised balance theory predicts
error of measurement only in the "--" situation and cannot
handle the error in the other conditions.

The certainty of completion responses. Table 3
presents the mean uncertainty for the basic responses in
each of the three completion situations for the 31 subjects
who responded to the free choice situation. A repeated
measures analysis was performed on the overall means for
the three situations, {(2,60) = 2.09, 235 .13, which showed
no differences between the situations. The further break-

down shown in Table 3 for uncertainty by response, or by

20
response-situation interaction, similarly showed no

significant differences.

Table 3

The Mean Uncertainty of the Completion Responses

 

Given relations

 

 

Overall
Response ++ +- -- mean
+ 3.97 2.00 3.78 3.86
0 3.00 3.53 3 71 3.56
- -— 3 67 3.50 3.62
Overall mean 3.94 3.55 3.71
Overall S.D. .73 .85 .97

 

Finally, the certainty of responses across the three
triads for those eight subjects who responded with perfect
balance in R3 was compared with the certainty of responses
for the remaining subjects. The balance subjects were
significantly more certain of their completion responses,
3 = -2.48, d: = 91, p £;.05. The mean certainty of the
completion responses for the eight "balance" subjects was
4.08 versus a mean of 3.58 for the remaining subjects.
Although the magnitude of this difference is not large, it
is supportive of balance theory. Those subjects who
conformed to the balance predictions were more sure of
their responses than the subjects who used other configu-

rations of responses.

21

Phenomenological reports: the "++" situation.

 

Balance theory predicts that subjects would base their
predictions on the harmony or conflict inherent in the
pattern of affect in the triad. Yet this is not what the
data revealed. Instead the subjects say that they made
their decision on the basis of an assessment of the
probability that the hypothetical people in the group would
have common interests. That is, college age people think
of friends in terms of how they spend their leisure time
and they appear to conceive of friendship in terms of
spending time doing the same thing.

Ninety-six percent of the 30 subjects who completed
two positive relations with a positive relation based their
argument on the common interests of the three members of
the triad. The following argument was abstracted from and
is consistent with the explanations of these 28 subjects.
(A more detailed version of the common interest explanation
to the "++" and also the "+—" and "-—" situations appears
in Appendix C.)

If p likes 9, then B has something in common with g.
If 9 likes q, then 9 has something in common with 3. But
if E has things in common with g and g has things in common
with q, then B will have something in common with q. Thus
the "++" situation should be completed with a positive
response.

The crux of this argument is that people interpret the

liking relation as transitive, not because positive affect

22
per se is transitive, but because common interests are
transitive. That is, all three members of the group have
common interests. This interpretation is supported by
the following phrases used by many of the subjects in
their explanations: "same interests", "mutual interests",
"common interests", "similarities", "all value the same
thing", and "birds of a feather flock together".

Phenomenological reports: the "+-" situation.

 

Neither the quantitative nor the verbal responses to the
incomplete structure defined by a positive relation and a
negative relation exhibited the uniformity of the responses
to the two positive relations. Not only are all three
quantitative responses used, but there are a variety of
explanations underlying the same quantitative response.
That is, the cognitions preceding identical behavior
patterns differ.

The first response to be analyzed is the balance
response--the negative relation. There are two kinds of
explanations underlying this response, though they are not
necessarily mutually exclusive. Four subjects thought that
since 9 and p like each other, and since p dislikes q, B
would discuss only q's "bad points" with 9. Person 9
would then tend to form a negative evaluation of q,
regardless of the favorable aspects of q.

The second explanation underlying the balance response
is a version of the same inductive explanation used by

eight subjects to justify the balance prediction to the

23
"++" situation. People who like each other have similar
interests and people who dislike each other have dissimilar
interests. But 9 likes the things 2 likes and q dislikes
the things 2 likes. Thus 9 and q dislike each other.

Two different kinds of explanations provided justifi-
cation for completing a positive relations and a negative
relation with the null relation. Persons 2 and 9 could
like each other for different reasons than the reasons 2
and q dislike each other. Thus 2 and 3 have no reason for
disliking each other.

The second explanation was offered by four subjects
who shared a "positivity bias" which countered the tendency
toward balance. The argument is that there are competing
tendencies: (a) to complete a positive relation and a
negative relation with the balance response, and (b) to
be motivated toward positive interpersonal affect. The net
result of these pressures is a neutral response. A typical
example of this argument is the following response, where
the three members of the triad are Paul, Pete, and Jeff.
"The reasons for Paul and Pete's dislike will influence
Jeff, but not enough to necessarily make Jeff dislike Pete."

One subject completed the "+-" structure with a
positive response, i.e., in direct opposition to the
response predicted by balance theory. This person believed
that two people have a tendency to like each other unless
presented with contrary evidence. Since the reasons for

disliking were considered by this and other subjects to be

24
more diverse than the reasons for liking, this subject
thought that g and q will probably not have a reason to
dislike each other. Thus he completed a positive and a
negative relation with a positive relation.

Phenomenological reports: the "--" situation. The

 

greatest diversity in quantitative and verbal responses
was to the incomplete structure defined by two negative
relations. The subjects in this study confirmed the
ambiguity of this situation noted by Heider (1958).

There is yet another version of the common interest
explanation which forms the premises for the conclusion
predicted by balance theory. But some subjects who completed
two negative relations with a positive relation used a
different reason than the common interest explanation.

Two different psychological mechanisms are responsible for
the same behavioral response.

The following is a summary of the common interest
explanation: Persons 3 and q have attributes in common
which E dislikes and vice versa. That is, g and q like
each other for the same reason they dislike 2 (although
some subjects did admit the possibility that g and q could
dislike p for different reasons).

The second cluster of explanations underlying the
balance response does not invoke common interests per se
as a motivating factor for the development of the liking
bond. This explanation is based more on the affective

processes specified by balance theory instead of the

25
cognitive emphasis on shared attributes. The primary
element uniting 9 and q is their mutual dislike of B. One
subject even suggested that o and q may have dissimilar
interests: "When people dislike other people, they stand
together because of that dislike, despite their differences
toward each other." In this case, the unitizing force
based on their mutual dislike for p is stronger than the
force opposed to unit formation based on the dissimilar
attributes of g and q.

The six subjects who interpreted the disliking relation
as transitive provided a reasonable justification for their
inferences. These subjects maintained that something was
wrong with p since 9 and q did not like p. But there was
also something wrong with g and q since 3 did not like
either of them. That is, g and q did not like each other
for the same reason 2 did not like either of them. All
three people had trouble interacting with other people. For
example, each member of the triad may be perceived as having
"bad" breath by the remaining two members.

Thus, at least in this experimental situation, Heider's
(1958) assertion that two negatives may be completed with
a positive 9£|a negative is correct. If g and q like each
other for the same reason they dislike p, Cartwright and
Harary's version of balance is maintained. If there is
something offensive about all three individuals which the

others cannot reconcile, the grouping schema of DeSoto et

a1. or the clustering principle of Davis provides an

26
accurate conceptualization of the situation.

There were two explanations for completing two nega—
tive relations with a null relation. The most popular
explanation, used by five subjects, may be labeled genuine
indifference. These subjects maintained that the existence
of mutual dislike does not provide sufficient information
for the two people to like or dislike each other.

The remaining subject simultaneously upheld the
possibilities that g and q could have something in common
or that p, g, and 3 could all be offensive people. Since
this subject saw no way to choose either alternative, they
canceled each other out, leaving no general principle as a

guide for predicting the missing relation.

Conclusion

 

In conclusion, the phenomenological analysis of the
quantitative responses provided much information that cannot
be obtained with the more traditional methods of social
psychology. First, reconfirming a point emphasized
repeatedly by Asch (1952), it was demonstrated that there
is not always a one-to-one correspondence between a response
and the psychology underlying this response. That is,
different psychological processes may map into the same
behavioral response. To limit study to an isolated response
ignores the underlying process. This restricted methodology
is particularly harmful if the response may be easily
interpreted to represent a process which is not only

unrepresentative of the actual psychology, but is a

27
distortion of this underlying psychology (e.g. Asch, 1948).
The exclusive study of isolated responses seems justified
only when the theory is sufficiently advanced such that the
experimenter is reasonably assured that he can account for
a single process underlying a given response. It is a rare
area in current social psychology where the knowledge exists
to draw such an inference.

The interim alternative to the study of isolated
responses is the phenomenological method. Indeed this would
seem to be the only method for exploratory studies in areas
such as social interaction.

This study reconfirms another point Asch (1948, 1952)
strongly emphasized; people usually have defendable,
reasoned arguments for their responses--regardless of
whether or not the response confirms a particular theory.
However, under the banner of "scientific exactness",
subjects are usually only allowed to express themselves by
a simple check in a 5-point scale (or by some equally
restricted response). Without the explanation, a response
indicating positive transitivity would probably be treated
as a mistake. If analysis of variance was used to analyze
the set of responses which included this, or any other
"deviant" response, it would have been formally treated
as error.

This final argument subsumes the previous two asser-
tions. The analysis of explanations added new information

to the understanding of the consistency of triadic

28

structures of interpersonal affect. No recent textbooks
in social psychology suggest that balance theory might be
explained in terms of shared attributes. Yet these
explanations were used by virtually all of the subjects to
account for the prediction of the liking relation to
complete the "++" situation, and by many of the subjects
who completed the other structures with balance responses.

This shared attribute explanation is related to at
least one other topic of psychological interest. This
explanation resembles Byrne's (e.g. 1961, 1971) similarity
and attraction hypothesis which states that attraction is
a linear function of similarity. The assertion that two
people who have common interests or attributes will like
each other is equivalent to Byrne's hypothesis. The major
discrepancy between Byrne's hypothesis and Heider's balance
theory is that the latter postulates affect to be the basis
for the underlying dynamics while Byrne is concerned with
more explicit cognitive processes. The pervasiveness of
the shared attribute explanation may imply that cognition
is more important than affect in the prediction of inter-
personal relations. Or, the cognitive explanations the
subjects give which support balance theory may be the
cognitions which they use to describe a process that, when

it actually occurred, was primarily based on affect.

APPENDICES

Appendix A

The Implications of the Transitivity of Positive Affect

This appendix contains the formal proof of the asser-
tions made in the text, i.e. a formal derivation of the
labeling of Structures 0-3 generated by transitivity of
positive affect. The following notation is used for binary
relations: "qu" is defined to be "p is related to q by
R" (Copi, 1967). Symmetry of R is represented by
"qu => qu", and transitivity is represented by
"Rop ° qu % Roq".

The sentiments in a triadic social structure define
two relations: the positive affect relation P and the
negative affect relation N. If no one is allowed to be
indifferent toward another, then these two relations are
completely determined by each other. In symbolic logic,
the lack of indifference would be represented by
'T> é? AIN" or by "N (3’ «'P".

One experimental procedure which has been used to
test balance theory is the triadic completion task in
which the subject is given some of the sentiments between
the people in a triad and is asked to predict others. If
there is no indifference, and if sentiments are recipro-

cated, then only three sentiments must be specified to

29

 

 

I ‘.|.l (In. ‘ ltl', III IIII A.

30
characterize the entire structure. The triadic completion
task is to predict the third sentiment if two of them are
given. If the two sentiments given are both positive,
then transitivity immediately applies to the situation and
no reasoning is required to conclude that the third senti-
ment is also positive (from Pop and qu one would imme-
diately conclude Poq). What is not obvious is that
transitivity also makes a definite prediction if the given
sentiments are positive and negative. The following
theorem shows that for symmetric triadic structures
without indifference, transitivity of positive sentiment
implies that if two of the sentiments in the triad are
poSitive and negative, then the third sentiment must be
negative. In formal terms this means that under the
assumptions ny (s? Pyx and ny (a) ~ny, we have the
equivalence of the expressions (ny - Pyz) =3) sz) é)

(ny ° Nyz % Nyz).

Theorem: Completing two positive relations with a
positive relation is equivalent to asserting that a
positive relation and a negative relation are completed

with a negative relation.

[(ny - Pyz) => sz] 4:) [(Pyx ° Nxz) => Nyz]

Proof: (ny - Pyz) => sz premise

vv(ny . Pyz) V sz material implication
(p9 q) <=? (~qu)

c'ny V e'Pyz V sz De Morgan's theorem
~(p°q) fi~pV~q

31

~ny V sz V ~Pyz commutation and association
~Pyx V sz V ~Pyz symmetry
~~(~Pyx V sz) V ~Pyz double negation ~~p 4:7 p

~(~~Pyx ° ~sz) V ~Pyz De Morgan's theorem
(Pyx - ~sz) =7 ~Pyz material implication

(Pyx - Nxz) => Nyz substitution

Appendix B

Sample Stimulus Page

Jeff and Paul like each other.
Paul and Pete dislike each other.
How do you think Jeff and Pete will feel toward each other?

PAUL
*
* * Please circle your prediction.
* *
like * * dislike dislike dislike indif- like a like
* * very much a little ferent little very much
* *

* * * * * * *
JEFF ? PETE

Please indicate how confident you feel about your prediction by
placing an X in the appropriate box.

not at all
certain / / / / / / very certain

 

 

Please explain the reasons on which you based your decisions.

32

Appendix C

The Logic of the Shared Interest Explanation

The purpose of this section is to provide a logical
formalism for the reasoning which subjects reported using
for the triadic completion task, i.e. a formalism for the
common interests argument. The basic notation is this:
Let there be three people 9, p, and q; and let their
interests be represented by a, b, and g respectively. If
the unit relation is denoted U, then this basic context

can be represented by an, Upb, Uqc, or by Figure 2.
Upb

 

an Uqc

Figure 2: The interests of p, g, and q.

There are two basic laws which form the basis of the
subjects' reasoning. The first is the assertion that one
person will like another only if he likes that person's
interests. In symbolism this becomes a sort of mixed
relation transitivity:

Lop-Upb =7 Lob
The second law is that if two people share common

33

34
interests, then they will like each other:
Lox-pr =7 Lop
These laws can now be used to derive the triadic
completion response. Consider first the "++" situation.
To be concrete, let us assume that Lop and qu, i.e. let

us assume the triad given in Figure 3.

E

 

9. 3.

Figure 3: The incomplete structure defined by two
positive relations.
Verbally the subject's reasoning goes like this:
9 likes p and therefore 9 likes p's interests. q likes
p and therefore q likes p's interests. But that means
that g and 3 have common interests and will like each
other. Logically stated this is:
Lop-Upb =7 Lob
Lqp-Upb =? qu
Therefore
LopoLqchpb => Lob-qu =7 Loq
Thus the prediction of the positive response to the "++"
triadic completion situation follows logically from the
premise of the two laws of common interests.
Next consider the "+-" situation. To be concrete,

consider the situation shown in Figure 4. Let the dislike

35

relation be denoted D. Verbally the subjects' reasoning
goes like this: 3 could like 9 only if q liked 9's
interests. But p likes 9's interests and q dislikes what
2 likes, and hence q cannot like 9's interests. Thus q
must dislike g for the same reason that he dislikes p.
That is,

qu =) ~qu =9 ~an => ~Loq
If the subject is limited to two responses, then «'Loq
would require him to choose Doq. If the response scale
allows indifference, however, then the subject is free to

choose either indifference or dislike willy nilly.

B

 

IO
ll-Q

Figure 4: The incomplete structure defined by a positive
relation and a negative relation.

The "--" situation is shown in Figure 5. It is
different from the other two situations in that the
prediction of a third response does not follow logically
from the two laws of common interests. Rather it is the
case that neither sentiment is inconsistent with those
laws. Thus if a subject were doing his reasoning by going
through cases and checking them out, then (a) if he first

checked out "like", then he would find the "+-" situation

36
of the previous case which is consistent with the rules
of common interests; (b) if he first checked out the
negative sentiment, then he would find no problem there
either. On the other hand, if the subject is reasoning
logically, then there is a fallacious argument to which
he might have fallen prey: qu =) ~Uqb and

Dop => ~Uob and (erroneously) ~Uqb ° ~Uob % Uac =) Log.

2

 

2 q_

Figure 5: The incomplete structure defined by two
negative relations.

LIST OF REFERENCES

LI ST OF REFERENCES

Asch, S. E. The doctrine of suggestion, prestige and
imitation in social psychology. Psychological Review,
1948, 55, 250-276.

 

Asch, S. B. Social psychology. Englewood Cliffs, N.J.:
Prentice-Hall, 1952.

 

Byrne, D. Interpersonal attraction and attitude
similarity. Journal of Abnormal and Social
Psychology, 1961, 62, 713-715.

 

 

Byrne, D. The ubiquitous relationship: attitude
similarity and attraction. Human Relations, 1971,
24, 201-208.

 

Cartwright, D., & Harary, F. Structural balance: A
generalization of Heider's theory. Psychological
Review, 1956, 63, 277-293.

 

Copi, T. M. Symbolic logic. London: The MacMillan
Company, 1967.

 

Davis, J. A. Structural balance, mechanical solidarity
and interpersonal relations. American Journal of
Sociology, 1963, 68, 444-462.

 

 

Davis, J. A. Clustering and structural balance in graphs.
Human Relations, 1967, 20, 181-187.

 

Davis, J. A., & Leinhardt, S. The structure of positive
interpersonal relations in small groups. In J.
Berger (Ed.), Sociological theories in progress.
Boston: Houghton Mifflin, 1970.

 

DeSoto, C. B., Henley, N. M., & London, M. Balance and
the grouping schema. Journal of Personality and Social
Psychology, 1968, 8, 1-7.

 

 

Fuller, C. H. Comparison of two experimental paradigms
as tests of Heider's balance theory. Journal of
Personality and Social Psychology, 1974, 32, 803-810.

 

Heider, F. Social perception and phenomenal causality.
Psychological Review, 1944, 51, 358-374.

 

37

 

 

 

U . l l I III I I'll? III Ila III. 'l U I
I I 1"

38

Heider, F. Attitudes and cognitive organization. Journal
of Psychology, 1946, 22, 107-112.

 

Heider, F. The psychology of interpersonal relations.
New York: John Wiley & Sons, Inc., 1958.

 

Holland, P. W., & Leinhardt, S. Transitivity in
structural models of small groups. Comparative
Group Studies, 1971, 2, 107-124.

 

 

Holland, P. W., & Leinhardt, S. Local structure in social
networks. In D. R. Heise (Ed.), Sociological
methodology. San Francisco: Jossey-Bass, 1976.

 

 

Morrisette, J. 0. An experimental study of the theory of
structural balance. Human Relations, 1958, 22, 239-254.

 

Newcomb, T. M. The prediction of interpersonal attraction.
American Psychologist, 1956, 22, 575-586.

 

Newcomb, T. M. Interpersonal balance. In R. P. Abelson,
E. Aronson, W. J. McGuire, T. M. Newcomb, M. J.
Rosenberg, & P. H. Tannenbaum (Eds.), Theories of
cognitive consistency: a sourcebook. Chicago:

Rand McNally, 1968, 28—51.

 

 

Price, P. O., Harburg, E. & Newcomb, T. M. Psychological
balance in situations of negative interpersonal
attitudes. Journal of Personality and Social
Psychology, 1966, 22, 265-270.

 

 

MICHIGAN STATE UNIV. LIBRARIES