100. DOĞUM YILINDA ATATÜRK'E ARMAĞAN DİZİSİ: 21

(1)

A N K A R A Ü N İ V E R S İ T E S İ SİYASAL BİLGİLER FAKÜLTESİ YAYINLARI NO. 478

100. DOĞUM YILINDA ATATÜRK'E ARMAĞAN DİZİSİ: 21

RE-EXAM!NATION OF THE NONBALANGE PHENOMENON

A Thesis

Presented to the Faculty of the Graduate School of Cornell University for the Degree of

Doctor of Philosophy

by

Fatma Seval GÜREL January 1977

(2)

-

(3)

\ • A N K A R A Ü N İ V E R S İ T E S İ

SİYASAL BİLGİLER FAKÜLTESİ YAYINLARI NO. 478

RE-EXAMINATION OF THE NONBALANCE PHENOMEHON

A Thesis

*

Presented to the Faculty of the Graduate School of Cornell University for the Degree of

Doctor of Philosophy

by

Fatma Seval GÜREL January 1977

(4)

A N K A R A Ü N İ V E R S İ T E S İ B A S I M E V İ - A N K A R A . 1 98 1

(5)

Biographical Sketch

Seval Fatma Gürel was born in Edirne, Turkey on August 17, 1943. She attended Brandeis University from 1964 to 1965, and Robert College (now called Boğaziçi or Bosphorus University) in istanbul, Turkey from 1965 to 1968. Frcm 1969 to 1972, shejoined the faculty of Siyasal Bilgiler Fakültesi (Faculty of Political Sciences), University of Ankara where she is returning after attending graduate school at Cornell University betvveen 1972 and 1976.

iii

(6)

(7)

Acknowledgments

I wish to express my appreciation to various persons involved in this research during the course of its development.

To the Chairman of my Special Committee, Professor Leo Meltzer, I owe my deepest thanks for his profound intellectual stimulation and support of this research through thick and thin. In particular, I wish to thank hini for the genuine enthusiasm he has shown for this' thesis and for the incredible amount of time, effort, and interest he has taken to sapervise and assist in this task.

I also wish to thank the other members of my committee, Professor Herbert Ginsburg, for takingthe time for thoughtful eritirisin of my work and for taking a special trip to appear at my final exam; Professor Robert Kraut for making elarifying sugges- tions on the analysis of treatments; Professor Steven Caldwell for providing the critical guidance in the data analysis just at the right time. Special thanks to Professor Donald P. Hayes, Chairman of the Sociolcgy Department, for providing funds for the pilot study and for computer time. I would also like to express my gratitude to Ms.

Barbara Sandrisser, Program Administrator at the Institute of Inter- national Education, for extendingme financial assistance throughout my graduate work.

I am much indebted to my graduate friends who work constantly under pressure yet did make the time to help a fellow graduate student in grappling with a foreign language. Their generous contributions in the critical judgment phase of the pilot study made the rest of the research possible: Greg Appling, Gene Burns, Robert Chung, Cal Cohen, Jean Dirks, Roland Foulkes, Maurice Haltom, Hank Hankerson, John Hedderson, Bili Hirst, Dennis Hodgson, Lynn Hodgson, Larry

v

(8)

Kraft, Marie Kwan, John Logan, Kate Loveland, John Manuli, Jerone McDaniel, Anne Meltzer, Richard Morrisey, Jane Megaw- Nyce, Dave Phillips (for doing a part of it), Don Poe, Jim Scheirer, Jeff Seaman, Frank Sparhawk, Bob Teicher, Burt Tesler, Kevin Welch and others—I will remember you ali with warmth.

Sincere appreciation is also due to Ceviz for helping me keep sane under the constant pressure of the last days, and especially to Mehmet, the 'soldat inconnu' of this dissertation, as spouses typically are.

(9)

List of Tables

i . i Comparison of Mean Pleasantness Ratings Obtained by Gerard & Fleischer

(1967) and by Whitney(ıg7i) 21 1.2 Comparison of Mean Ratings Obtained by Gutman <fc Knox (1972) and by

Norman & Miller (1976) 23 4.1 Means, Standard Deviations, and Matched (Paired) T-test Results in terms of

Consistency ( 0 = L o w , 10= High) Ratings, över ali Stories for each Treatment 56 4 . 2 Means, Standard Deviations, and Matched (Paired) T-test Results in terms of

Pleasantness ( 0 = L o w , ıo = High) Ratings, över ali Stories for each Treat-

ment 57 4 . 3 Mean Differences of Ratings (MDR's) of Balanced Triads Minus Unbalanced

Triads, with Consistency and Pleasantness, across Treatments 59 4 . 4 Absolute Values of Mean Differences of Ratings ( | M D R | ) of Balanced Triads

Minus Unbalanced Triads, using Consistency and Pleasantness, across Stories,

for each Treatment 61 4 . 5 Comparison of Dependent Measures in terms of Mean Differences in Ratings

(MDR) of Balanced Minus Unbalanced Triads: (p-values shown are for

Consistency MDR vs. Pleasantness MDR; Group T-tests) 62 4 . 6 Number of Subjects Choosing Balanced or Unbalanced Stories as Inconsistent,

and Number Choosing the Same Stories as Unpleasant, över ali Stories for

each Treatment i 64 4 . 7 Discrimination via Dichotomous Choices (combined across Treatments) . . . 65

4 . 8 Discrimination via Dichotomous Choices in terms of Row Percentages . . . . 65 5.1 Analysis of Variance of Composite Consistency Ratings for Treatment and

Configuration (Balanced / Unbalanced) Effects 68 5.2 Analysis of Variance of Composite Pleasantness Ratings for Treatment and

Configuration (Balanced/Unbalanced)Effects 68 5 . 3 Analysis of Variance of Composite Ratings for Type of Rating (Consistency

vs. Pleasantness), Treatment, and Configuration (Balance / Unbalanc3) . . . . 69

(14)

5 . 4 A Priori Contrasts of Groups of Treatments via ANOVA and T-Probabilities, for

\ Consistency, and Pleasantness in terms of Balance or Unbalance Ratings and Dif-

ference Scores 72 5.5 Trend Analysis of Balance and Unbalance Ratings, Difference and Absolute Dif-

ference Scores, p-Values for Consistency and for Pleasantness (N=207) . . . 73 5.6 Trend Analysis (without Primacy) of Balance and Unbalance Ratings, Dif-

ference and Absolute Difference Scores, p-Values for Consistency and for

Pleasantness ( N = I 7 6 ) 74 5 . 7 Trend Analysis of Balance and Unbalance Ratings, Difference and Absolute

Difference Scores, p-Values for Consistency and for Pleasantness ( N = 2 0 5 ) . 75 5 . 8 Trend Analysis (without Primacy) of Balance and Unbalance Ratings, Diffe-

ference and Absolute Difference Scores, p-Values for Consistency and for

Pleasantness ( N = i 7 6 ) 76 5 . 9 Analyses of Variance of Composite Consistency Ratings for Consiguration

(Balanced / Unbalanced) Efîects^{; ;} . . 78 5.10 AnalyseS of Variance of Composite Pleasantness Ratings for Configuration

(Balanced/Unbalanced) Effects 79 5.11 Comparison of Treatments in terms of Proportions of Variation Explained by

Discrimination (Balance vs. Unbalance Configurations) 80 5.12 Comparison of Primacy and Sentiment Treatments 82 A. i Means and ConfidenCe Limits of Pilot Ratings on Eight Dimensions, for the

Four Pairs of Stories of the Questionnaire in the Standard Contidion ( N = 2 o ) . 102 A . 2 Means and Confidence Limits of Pilot Rating on Eight Dimensions for the Four

Pairs of Stories of the Questionnaires in the All-Sentiment Condition ( N =

20) 103 B. i Means, Standard Deviations, and Matched (Paired) T-test Results in terms of

Consistency Ratings ( 0 = L o w , ı o = H i g h ) , for each Story in each Treatment . 107 B.2 Means, Standard Deviations, and Matched (Paired) T-test Results in terms of

Pleasantness Ratings (O = Low, i o = H i g h ) , for each Story in each Treatment . 108 C. i Number of Subjects Choosing Balanced or Unbalanced Stories as being

Inconsistent 112 C. 2 Number of Subjects Choosing Balanced or Unbalanced Stories as being

Unpleasant ' 113 C. 3 Number of Subjects Choosing Balanced or Unbalanced Stories as Inconsistent,

and Number Choosing the same Stories as Unpleasant 114 D. i ANOVA Consistency Ratings of Four Stories in ecah of the Balanced or Un-

balanced Triads 118 D. 2 ANOVA Pleasantness Ratings of Four Stories in each of the Balanced or Un-

balanced Triads 118 D . 3 Means and Matched T-test Results 119

D . 4 Matched T-test Results (between indicidual stories) ^ 120

D . 5 Matched T-test Results (betvveen story Combinations) 121 E. i Distribution of Future Contact Assumptions across Stories in a Sample of 6o

Questionnaires 126 Xİİ \

(15)

G.l ANOVA of Composite Consistency Ratings of Balanced Configuratiotıs

for Treatment Effects 217 G.2 ANOVA of Composite Consistency Ratings of Unbalanced Configurations

for Treatment Effects 217 G.3 ANOVA of Composite Pleasantness Ratings of Balanced Configurations

for Treatment Effects • 217 G.4 ANOVA of Composite Pleasantness Ratings of Unbalanced Configurations

for Treatment Effects 217 H.l ANOVA (Repeated Measures Design: BMDP2V) of Composite Consistency

Ratings for Treatment and Configuratiori (Repeated Factor) Effects 223 H.2 ANOVA (Repeated Measures Design: BMDP2V) of Composite Pleasant-

ness Ratings for Treatment and Configuration (Repeated Factor) Effects . 223 H.3 ANOVA (Repeated Measures Design: BMDP2V) of Composite Ratings for

Type of Rating, Treatment and Configuration (Repeated Factor) Effects 223 H.4 ANOVA (Repeated Measures Design: BMDP2V) of Composite Consistency

Rating for Configuration (Repeated Factor) Effects 224 H.5 ANOVA (Repeated Measures Design: BMDP2V) of Composite Pleasant-

ness Ratings for Configuration (Repeated Factor) Effects 225 H.6 ANOVA (Repeated Measures Design: BMDP2V) of Consistency Ratings of

Four Stories in each of the Balanced and Unbalanced Triads, for Story

(Repeated Factor) Effects 226 H.7 ANOVA (Repeated Measures Design: BMDP2V) of Pleasantness Ratings of

Four Stories in each of the Balanced and Unbalanced Triads, for Story

, (Repeated Factor) Effects 226

' I

(16)

(17)

List of Figııres

i . i Heider's Partition of the Eight Triadic Configurations into Two Categories and

Nevvcomb's Three Category Classification , 8 3. i Detailed View of the Standard Treatment 44

3.2 Research Design 45 3.3 Bird's Eye view of the Research Design with the VVholistic Pattern of

Predictions 46 4.1 Schematic Representation of Location of Grand Means for Balanced and

Unbalanced Triads, Consistency and Pleasantness 60

T

(18)

(19)

\

t

Abstract

The general purpose was to identify conditions which ha ve obscured prior tests between two rival formulations of attitude consistency—Heider (1958) and Newcomb (1968)—and therebyto clarify the applicability of these two theories more definitely than in prior research. More specifically, the purpose was to specify variables which determine whether Newcomb's nonbalance predictions or Heider's balance predictions are more accurate when one person dislikes another.

i

Data were gathered in an experiment using questionnaires.

Stimuli to be judged were presented to the subjects via paired comparisons of stories, one balanced in the Heiderian sense and one unbalanced. The equivalence of the paired stories was established in a pilot study. An attempt was also made to make the stories realistic and slightly humorous. '

Several phenomenological conditions were hypothesized as influencing the extent to which subjects vvould respond to these stories as Newcomb or Heider vvould predict. These include: the number of sentiment relations, order of presentation of the in<er- personal attraction bond, degree of structure in tlıe stimulus situations, subjects' degree of identification with the main character, and subjects' concentration upon or distraction from the totalityof the triadic situation. Most of the relevant factors were translated into treatments and manipulated by instructions; others were held constant. Half the subjects judged the stories for consistency; the other half judged them for pleasantness. The research design con- tained 14 groups (7 treatments x 2 measures) ; 30 persons were randomly assigned to each condition.

xvii

(20)

The following conclusions were ali supported: Even in the the- oretically troublesome condition identified by Newcomb, that in which one person dislikes another, judgments of consistency is an appropriate measure for testing Iieider's expectations concerning discrimination of balance and unbalance. Judgments of pleasantness are less discriminating, more variable, and more responsive to variables other than balance alone, and consequently constitute an inappropriate measure for testing either Heider's or Newcomb's theory. Under some conditions, judgments of pleasantness may be made in the direction opposite to th&t predicted by Heider. Those prior studies which favored Heider's predictions generally utilized cognitive dependent measures while those which favored Nevvcomb's predictions generally utilized affective dependent measures. As expected from the Geştalt nature of Heider's theory, balance outcomes are reduced when conditions lead subjects to pay attention to the component parts of a triadic situation.

In sum, it is concluded that the complications in the literatüre can be traced to the effect of numerous phenomenological conditions, which have varied unsystematically from study to study, as well as to the non-equivalence of the dependent measures typically used in these studies.

xviii

(21)

*

CHAPTER 1

Introduction

THE PROBLEM

Heider's theory of attitudes (1958), 'namely "balance theory", is renowned for its simplıcity and for being in accord with common sense. Yet it has not been fully borne out by research.

Balance theory postulates, in the phenomenal world of a person, the presence of triadic structures that involve a representation of the person himself (P), another person (O), and an issue that concerns them both (X). Heider distinguishes among these structures in terms of whether they are balanced or unbalanced.1

The reader of this thesis is assumed to be familiar vvith Heider's theory, or is expected to consult his 1958 book to acquire that familiarity. Even though it has been frequently described in the social psychological literatüre, the secondary accounts do not alvvays make clear that there are essentially two main propositions to the theory: (1) that subjects partition the set of eight triads into two subsets (balanced or unbalanced); and (2) that the balanced set is stable and the unbalanced set is unstable. It appears that most writers have taken the first propositon for granted, in effect assuming that subjects can and do make this discrimination.

1 Each relation (P /O, P /X, O /X) is viewed as signed ( + or — ). Eight triads are possible in terms of combinations of such signs, and these vvill be listed here in the order P/O, P/X, 0 / X : + + + , + , t-, — H , — + + ,

+ — + , + + —, — — —. Heider distinguishes among these structures such that the first four are labelled balanced and the second four are labelled unbalanced (or imbalanced).

1

(22)

Heider's theory has not received full support in a number of studies, particularly where P's attraction to O is negative (negative P / O relation). Newcomb (1968) attempted to salvage the essentials of the balance framework via a reformulation of the theory.

He proposed that Heider's distinction of balanced and unbalanced configurations applies only in the case of positive P / O relations; but where P's attraction to O is negative,2 a third state, labelled nonbalance, is generated. In effect, Newcomb's theory has two propositions paralleling those given above for Heider: (1) subjects recognize ,a partition of the triadic structures into three groups; and (2) the balanced set (given positive P /O) is the most stable, the nonbalanced set (given negative P /O) is intermediate in stability, and the unbalanced set (given positive

P /O) is the least stable.

Nevertheless the results have continued to be mixed, and Nevvcomb's three category distinction has not been observed consis- tently. In consequence, it has seemed appropriate to some observers that balance theory be abandoned.

On the other hand, my review of the literatüre-to follow later in this chapter - suggests that there are some methodological problems involved with the experiments testing the two rival formulations. Since the experiments have usually employed the same paradigm, it appears to me that their results do not constitute independent tests of Heider's or Nevvcomb's theories, but rather point to some shared artifactual outcomes - outcomes associated with the experimental setting as compared to real life circumstances, or with the paper and pencil assessment techniques in particular.

Procedural artifacts have, as are shown belovv, been brought about by the particular dependent measures used, some variables that are embedded in the stimulus display but have not been explicated so far, and the elicitation of confounding affective or cognitive processes by means of subtle variations in some variables. Thus it appears that a fair test of balance theory — whether of Heider's or Nevvcomb's variety — needs to consider ali of the potential sources that can destroy or prevent the operation of the balance dynamics in the phenomenology of the person

2 The four nonbalanced triads are: 1 , b, b + . - • 2

(23)

especially when participating in a study which relies on paper and pencil assessment tehcniques. The potential effects of the instruments, tasks and their interactions with the dependent measures remain to be investigated. Fairness demands that this alternative be exhausted before the balance theory is abandoned, or even before substantive modifications of the theory such as that of Nevvcomb are introduced.

To repeat, both Heider's and Nevvcomb's theories have two majör propositions: (1) a recognition proposition: Subjects divide the set of triads into subsets having shared phenomenal properties;

(2) a stability proposition: some of these subsets are more stable than others. Most of the literatüre has concentrated on the second proposition. It is my contention that the first proposition is logi- cally prior and should be investigated first. Accordingly, this thesis will concentrate on a test of the first proposition and the factors affecting subjects' partition of the triadic structures into two subsets (as predicted by Heider) or three (as predicted by Newcomb). However, since there is no disagreement between Nevv- comb and Heider in the case of the four triads vvhere P / O is positive, nor does the literatüre fail to find the discriminations between these two subsets as predicted by either theory, this thesis will confine its investigations to the four triads with negative P /O.

The revievv of literatüre to follow in this chapter identifies the phenomenological conditions hypothesized by me as determining vvhether Heider's balance-unbalance effects or Nevvcomb's nonbalance effects are to be expected in experimental tests, using paper and pencil procedures. It is hypothesized that the factors directly or indirectly responsible for the failure of balance predictions include: the number of sentiment relations, the order of presentation of the P / O relation, the degree of structure of the stimulus situations, the subject's lack of concentration on the to- tality of the triadic situation and vvhether or not the instructions generate identification of the subject with P.

Another reason for the confusions and contradictions in the literatüre, it seems to me, is the use of a large number of different dependent variables in a more or less intersubstitutable manner.

For example, the triads have been investigated in terms of such diverse measures as pleasantness, consistency ratings, ease of learning, attitude change, comprehensibility, involvingness, memory,

3

(24)

ete.; moreover, sometimes incomplete triads are presented and subjects are asked to supply the missing relation. However, to test the recognition postulate, ratings or judgments of consistency seem most appropriate and will be used in this thesis. I predict that when appropriate methodological controls are maintained, and if consistency ratings or judgments are made, subjects will divide the set of triads (with negative P / O) into the two sets delineated by Heider rather than fail to differentiate these, as suggested by Nevvcomb's nonbalance category.

I will also inelude in this thesis a second dependent measure, i.e., pleasantness ratings. It is my contention that this is an inap- propriate measure of the recognition postulate, even though it

has been a widely used response measure in prior studies. My predietion is that, using this dependent measure, there is no compel- ling a priori reason to believe that Heider's or even Newcomb's predietions vvill be supported.

In short, I hope to demonstrate in this thesis that Heider's theory is basically correct and preferable to Newcomb's reformulation of it. A secondary purpose is to show that, by specifying certain conditions, the balance theory can also serve as a tool for investigating the funetioning of the mind.

LİTERATÜRE REVIEW Basic Findings and General Concepts

Theories which incorporate the idea of cognitive consistency have occupied a position of considerable importance in social psychology. It has even been argued that they constitute the field's

"first truly general... and compelling theoretical system", (Brown, 1962). The consistency formulations as a whole, although each is direeted to a slightly different aspect of behavior, rest on the assumption that the individual strives for consistency among his cognitions, attitudes, or aetions.

Heider's Balance Theory: According to the earliest of these theories, namely by Heider (1958), the phenomenology of the person P is represented as consisting of perceived relations betvveen the elements: P himself, another person O, and an imper- sonal entity X. Heider recognizes two interehangeable types of relationships between the elements. Sentiment relations refer to 4

(25)

evaluative responses while unit relations refer mainly to Geştalt principles. The two types of relations .are further divided into positive and negative so that we may have: a) positive unit relations ( + U ) based upon unit forming principles, b) negative unit relations (—U) indicatirig a segregation between two elements, c) positive sentiment relations ( + L ) implying liking or approval, d) negative sentiment relations (—L) implying dislike or disap- proval.

Balance theory grew out of a viewpoint that is Gestaltist, field theoretic, and organismic. A balanced configuration is analo- gous to agood Geştalt: balance is defined as "aharmonious state, one in which the entities comprising the situations and the feelings about them fit together without stress" (Heider, 1958, p. 180).

Elements (P, O, X) and relations (P / O, P / X, O / X) are organized into cognitive systems like dyads or triads. These systems tend toward balance. A dyad is balanced when its two relations bear the same sign. A triad is balanced when the product of its three relations is positive, and unbalanced when this product is negative -i.e., in accordance with the multiplicative rule. According to Hei- der, a balanced state is preferred to an unbalanced one. When an unbalanced state occurs, this will give riSe to forces in the direction of restoring balancd; a balanced state is achieved by modifying the relationships that constitute the triad.

f.

In accordance with Geştalt theory, cognitive consistency can be conceived as a perceptual or cognitive field process. This viewpoint was shared by Festinger (1957), and a similar explanation for balance dynamics was offered by Jordan (1968). Heider himself interpreted "consistency" to mean "economy" or "redundancy"

(1958, p.51), and Pepitone described it as "simplicity of cognitive structure" (1966). Cognitive contents tend, not only to not to contradict one another, but also to imply each other to some extent (Peeters, 1971). Thus the mind tends to simplify the informa- tion it deals with by shaping it so as to make it "harmoniously"

related with a pre-existing cognitive structure. Such structures have been referred to as "meanings as integrating factors" (Heider, 1958, p.47), "implicational principles, or molecules" (Abelson, 1968), and "conceptual good figures" (De Soto and Albrecht, 1968).

Heiderian balanced configurations represent such structures, and Heiderian dynamics seem to belong in the general category of 5

(26)

cognitive functioning (later in the thesis to be distinguished from affective processes).

The Jordan Paradigm: The first systematic study of Heider's balance formulation was done by Jordan (1953). Since there are four kinds of relations (like, dislike, positive unit relation, negative unit relation) and since each triad of P-0-X has three relations, 4³ or 64 possible triads can be constructed by using them. Jordan formulated each triad in a simple, abstract fashion, rather devoid of content. The positive and negative sentiment relations were designated as "like" and "dislike". Positive and negative unit relations were translated respectively into "has some sort of bond or relationship with" and "has no sort of bond or relationship with". P was replaced by "I" while the symbols "O" and "X"

were maintained for the remaining entities. The triad "P + L O;

P - L X; O + U X " was thus translated into "I like O, I dislike X, O has some sort of relationship with X".

Jordan asked subjects to rate the pleasantness-unpleasantness of each triad. Balanced triads were expected to be pleasant and unbalanced (or imbalanced) ones unpleasant. These expectations were in general confirmed so that the four balanced triads ( + + + ,

•I , — H , - + ) taken as a whole were ratcd more pleasant than the four unbalanced triads ( + + —, + — + ,

— + + , — )• Jordan observed, however, that this overall result was primarily due to the pleasant ratings given to + + 4- and triads. These are the triads in which the P to O bond is positive, and in which P and O agree in their orientation toward X. The remaining six triads, including the two balanced triads with negative P / O, had means indicating approximately the same degree of unpleasantness.

Jordan revised the nature of the determinants of "pleasantness"

and "unpleasantness" as follows: Negative relations per se, between humans are unpleasant; therefore the operational definition of balance in terms of pleasantness is incorrect. "Balance implies propriety and not necessarily pleasantness... Balanced situations can therefore be considered to be experienced as more proper than imbalanced situations, independent of their degree of experienced pleasantess... But there are also other (other than attraction) determinants of unpleasantness. (Some situations have positive

(27)

attraction) yet are rated as being quite unpleasant. Here then the rating of unpleasantness must be attributed to the state of imba- lance" (Jordan, 1953, p. 282). The tendency for the triads vvith the largest number of negative relations to be rated as unpleasant even if they are balanced is generally referred to as the positivity bias. In Jordan's study, the positivity effect was more typical of sentiment than of unit relations; it depended more upon the P / O relation and less upon the P / X and O / X relations.

Newcomb^,s Reformulation of the Balance Theory: Largely on the basis of such phenomena as mentioned above, Nevvcomb proposed a functional distinction between a positive and a negative P / O relation, even in situations of Heiderian balance. Nevvcomb (1968) grouped ali structures containing a negative P / O relation into a single middle category of nonbalance, which he thought are characterized by lovv engagement and therefore implying "a state of little or no preference for balance or unbalance vvithin the to- tal set of relations", (Nevvcomb, 1968, p.33). Thus, as can be seen in the follovving illustration, tvvo rival classifications of the triadic situations came into being (Figüre 1.1).

Nevvcomb tested his hypotheses about the three categories against Heider's tvvo categories in an experiment by Price, Harburg and Nevvcomb (1966) vvhere the subjects vvere asked to name their bestfriends on campus, and also tvvo students vvhom they disliked.

The hypothetical P-0-X triads vvere presented vvith O and Q (Q standing for a third person in the place of X) as either a friend, or as one of the disliked persons. The subject indicated on a 90 mm şcale hovv pleasant or uneasy he felt about each of the eight structures. Since Nevvcomb considers the triads vvith a negative P / O relation as "neither clearly balanced nor clearly imbalanced"

(1968, p.36), he expected and found that these nonbalanced triads differ from either of the other tvvo categories and "shovv not only far less unanimity, but also more frequent "neutral" responses and larger differences betvveen the tvvo situations in each category — exactly as to be expected if non-balanced situations differ from both balanced and imbalanced ones," (1968, p.36).

The difference in the pattern of results in the Price et al.

study compared to the one in Jordan's study could be due to the follovving: Jordan averaged the scores över sentiment and unit 7

(28)

I

Figüre 1.1

HEİDER'S PARTITION OF THE EIGHT TRIADIC CONFIGURATIONS INTO TWO CATEGORIES

B a l a n c e d TJnbalanced

X X

' 3 \ -

X

'5\~

X

' 7 \ +

X

- / 8 - 0

I

NEWCOMB'S THREE CATEGORY CLASSIFICATION Configuration

number 1 and 2 i 5 and 6 )

i

7 and 8 ) 3 and 4

Label

Positively balanced Negatively balanced Negatively unbalatıced Positively unbalanced

Characteristics

Most balanced, pleasant, stable Intermediate

Intermediate

Least balanced, pleasant, stable

Relation

P/O

P / X

o / x

THE TWO RIVAL CLASSIFICATIONS Configuration number:

A l A2 A5 A6 A7 A8 A3 A4

+ + +

Newcomb's Nonbalance

+ —

— + + — + —

+ + + —

— +

Heider's Balance Heider's Unbalance

(29)

relations together; so the unit relations could have helped "di- lute" the more acute scores associated vvith sentiment relations.

Further, Price et al. utilized stimulus situations with some, though little, content to them relative to the skeletal structures which Jordan had presented to the subjects. Moreover, X was presented as a third person by Price et al.

Later, Rodrigues (1967) employed the Jordan paradigm except for its unit relations and replicated those of Jordan's results vvhich fail to support either Newcomb or Heider. Even though their differential effect on the outcomes is not clear, we note that Jordan, and also Rodrigues, used a scale anchored best-neutral- worst while Price et al. used a scale anchored pleasant-uneasy.

Rodrigues gave his version of the unspecific, unstructured instructions as: "The two persons are you an^i another person, O. The X is not specified and may be anything at ali toward which you and O have a positive or a negative feeling."

We note that the majority of test of balance theory involved the use of paper and pencil procedures. The earlier studies on the theory employed ratings of pleasantness. They reported the follovving pattern of results: liking and agreement situations were rated as highly pleasant, but the remaining configurations, which are the three-fourths of ali the structures, were rated as rather unpleasant thus being "lumped" together in one big category of low preference. Since there were no clear distinctions in degrees of unpleasantness between these categories, neither Heider's expectation of a distinction between balance and unbalance, nor Newcomb's expectation of nonbalance were supported.

Early Tests of Balance Formulations via Pleasantness Ratings:

Nevvcomb's 3 category partition of the triadic structures has not been observed regularly. Nevvcomb (1968) himself reported a study by Steiner and Spaulding (1966) which had the order of the two triads vvithin the nonbalance category the reverse of what Price et al. had observed. Steiner and Spaulding had presented four selected triadic situations and, unlike the Price et al. study, P and O were represented by common first names and X was represented by a specified object or activity. Contrary to Nevvcomb, Steiner and Spaulding observed a clear ordering among the triads; and in this order approximated that of Heider's categories.

9

(30)

Gerard and Fleischer (1967) presented subjects with descriptions of real-life situations. In one experiment they asked the subjects to rate the descriptions for pleasantness or uneasiness; and in another experiment, they asked them to do a recall task. Gerard and Fleischer observed that the triads with —P / O bonds behave differently from those with + P / O bonds. Hovvever, the nature of this difference, they thought, was ambiguous. When P / O was positive, the recall was better for unbalanced than for balanced triads; when P / O was negative, there appeared a statistically insignificant trend in the opposite direction. In line with the Price et al. findings, balanced triads were rated as more pleasant only when the P / O relation was positive. While Price et al. had observed no difference in pleasantness for triads having a negative P / O relation, Gerard and Fleischer found an unexpected rever- sal (contrary to conventional assumptions) such that balanced negative P / O triads were experienced as less pleasant than unbalanced, negative P / O ones — a result consistent with the recall data. We note that, compared to the Price et al. study, Gerard and Fleischer presented the subjects with quite elaborate descriptions of stimulus situations. However, the descriptions vvere loosely and casually worded. Further, Gerard and Fleischer did not control for the confound of implicit anticipation of future contact; morever, pleasantness or uneasiness or tension maynotbe interchangeable. Similar results were observed by Whitney (1971) who, employing the rating task of Gerard and Fleischer, replicated their methods and are therefore subject to similar criticisms.

Oqe possible clue to the unstable nature of findings with respect to Newcomb's nonbalanced triads can be seen in the vary- ing nature of the procedures employed in different studies. Often the researchers made implicit assumptions or oversimplifications regarding the form and content of the stimulus situations, and failed to systematically investigate them.

I mentioned that Whitney (1971) replicated Gerard and Flei- scher's findings. Hovvever, vvhile Gerard and Fleischer reported a highly significant interaction between balance and attraction and no main effects for either balance or attraction, Whitney reported significant main effects for agreement and for attraction but no 10

(31)

interaction. Whitney concluded that there is no indication that balance had any effect on the pleasantness ratings.³

The observation that pleasantness is determined mainy by agreement and attraction but not by balance received support from other studies vvhich, however, differed also among themselves vvith respect to the details of their procedures and their results.

Fuller (1974) used the procedures of Rodrigues (1967) to ob- tain pleasantness ratings with the modification that the words

"highly pleasant" and "highly unpleasant" appeared at the anchor points of the rating scale. Fuller obtained a significant main effect for agreement, and for attraction; balance tendencies vvere also reflected in the significant interaction between attraction and agreement. With the completion task also employed in the same study, she observed that subjects completed the configurations in the manner predicted by Heider's theory. In line with the finding that the configurations with a negative P /O bond tended to receive neutral pleasantness ratings, Fuller expected that a "lack of engagement"

on the part of the subjects could lead to such neutral responses which would indicate an avoidance of negative responses. She found no relationship between the frequency of neutral responses and vvhich relation was negative. Contrary to the disengagement expectations, Fuller concluded that subjects have not been choosing the neutral response in order to avoid choosing a negative balancing response.

Tesis of Balance via Measures Other than Pleasantness Ratings:

The above mentioned pattern of results reflect the complex nature of the observations associated vvith pleasantness ratings.

Even though they do not clearly support either Nevvcomb or Heider, the common thread running across ali such observations is that the results obtained vvith pleasantness ratings have something to do vvith the component parts of the triad to a greater extent than vvith the vvhole of it.

The test of balance formulations have not been restricted to the ratings of pleasantness. Other measures such as the number of

3 The terminology has the follovving usages: "Attraction" refers to the P/O relation, usually the preference for the positive P /O bond. "Agreement" refers to the similarity of signs betvveen the P/X and the 0 / X bonds. "Balance" takes into account ali three relations in accordance vvith the multiplicative rule.

11

(32)

mistakes made in learning given structures (De Soto, 1960; Zajonc and Burnstein, 1965) or probability (expectancy) statements made by subjects (De Soto and Kuethe, 1959), or subjects' reactions to the stimulus situations, the assumption being that their responses will reflect the existing "cognitive biases" (Morisette, 1958; Kuethe, 1962; Ohashi, 1964; Feather, 1965). Even though the results are mixed also in the case of these measures, they do not come near the extent of complications associated with pleasantness ratings. For instance, partial structures containing either a negative or a positive P /O relationship have been completed in such a manner as to produce balanced configurations (Aronson and Cope, 1968; Deutch and Solomon, 1959; Sampson and Insko, 1964) except- ing Wyer and Lyon (1970), and Truzzi (1970) whose results do not support Heider's predictions.

Crano and Cooper (1973) investigated the relative success of tvvo dependent measures — i.e., pleasantness, and stability —in sup- porting either of the tvvo theories.The stimulus situations were of the three person, P-0-Q type. As in the Price et al. experiment, subjects vvere asked to supply the names of four persons currently attending the same university, tvvo of whom strongly liked, and tvvo they strongly disliked. Ali bf the eight configurations vvere presented to the subjects in pairs such that ali possible pairs of these situations vvere represented. In a second study, the P-0-X type situations vvere used.

\ In both studies, perceived pleasantness proved to be non- scalable, and therefore Crano and Cooper concluded that it is not a unidimensional, rather a multidimensional variable. A perceived stability measure (defined such that "some sort of change of feeling betvveen individuals vvould not be expected") was scalable, and this scale clearly supported Nevvcomb's revision of balance theory^ as did the pleasantness scale in previous studies. The nonbalanced situations received intermediate degrees of stability ratings; they vvere significantly different from positively balanced and unbalanced situations, but they did not differ significantly among themselves. We note here, hovvever, that the presentation of ali possible pairs of configurations introduces a complication into the results. The tvvo triads in a pair could simultaneously differ both in attraction and in agreement (or disagreement) factor in some cases, and in others they could differ both in attraction and in balancedness (or unbalancedness). Unless there is only one 12

(33)

distinctive feature between the paired items, that systematically relates to the hypothesis in question, the results fail short of providing a conclusive answer.

In a third study, Crano and Cooper (1973) studied the reasons why the classification which Newcomb predicted occured in their results. Subjects were asked to rate the pairs of P-0-Q situations in terms of comprehensibility and of involvingnees. The two dimensions were defined as:

"COMPREHENSIBLE— A comprehensible situation is one in which you have a rather clear idea of the feelings that each member of the group has for each of the others. In other words, you can inter- pret unambiguously the feelings of the other people in the situation.

INVOLVING— An involving situation is an engaging situation, one in vvhich the relationships betvveen the various members of the group prove to be of real interest to you," (p. 350).

Crano and Cooper expected that, if Newcomb's motivational explanation in terms of disengagement is correct, subjects would distinguish the triads on the basis of involvingness, with groups having positive P /O bonds rated as more involving than the remaining ones. They did not predict, however, a similar differen- tiation on the basis of comprehensibility since the subject vvho disengages himself from the situation because of a negative P /O bond would not be expected to explore the remaining relationships closely enough to make such a distinction. The results confirmed the predictions of Crano and Cooper vvho concluded that the negative P /O bond is botb a necessary and a sufficient reason for disengagement.

As an explanatory mechanism for disengagement phenomena, Crano and Cooper proposed a biphasic decision or attention process determining subjects' reactions to each configuration.

In the first pbase, the critical P /O bond is supposed to be merely inspected; and if negative, indifference to further concern for the situation would emerge. If positive, this would activate the second phase, the investigation Of the remaining bonds.

Involvement would be greater in the second phase, and this would explain why the judgments elicited by positive P / O situations would be more extreme than the nonbalanced ones.

' Crockett (1974) manipulated the characteristics of the triadic situation such that X was either a third person, or a controversial 13

(34)

political issue; O was either an acquaintance oranational political figüre. He employed the procedures of Price et al. and obtained ratings in terms of the pleasantness, the expectedness of the situation, the truthfulness of O, and whether or not P would modify his feelingS for O and X. Crockett argued that, in the case of a disliked O, the expected disengagement of P from the situation may not occur, because this is not ahvays possible. Thus P may not be able to remain indifferent to O's opinions about X vvhether he likes O or not.

Crockett obtained a complex pattern of results. Only for the ratings of pleasantness, the prediction that P becomes disengaged from situations that involve a disliked O and a third person Q, was confirmed such that the results duplicated those ofPriceetal.

In a P-0-X situation, pleasantness ratings showed consistent a- greement preference whether attraction was positive or negative.

Ratings of expectedness, and O's truthfulness showed clear balance effects. The results from the latter measure were not easily inter- pretable, but were generally in the direction predicted by balance theory. Crockett concluded that "a person's phenomenal exper- ience of a situation cannot ordinarily be captured on a single dimension of jııdgment; it consists in a mix of a number of different dimensions... The second conclusion is that the effects of different organizing principles wary with the the dimension of judgment, and with the interpersonal context within which the judgment takes place, " (1974, p. 110).

Like Crockett, Gutman and Knox (1972) suspected that differ- rent measures may be tapping into a variety of aspects a triadic situation may involve. They thought that balance principles may be associated witn one of those multiplicity of aspects; the perceived feelings of pleasantness or uneasiness were not at the center of balance concerns, but were something quite irrelevant to what balance principles were intented to get at. Gutman and Knox (1972) suspected that agreement and attraction may come out as more important than balance when the response measure is based primarily on affect (pleasantness or tension). The re verse would be true however if the measure related more directly to the cogni.tive component of social perception.

Gutman and Knox investigated the relationship betv/een the ratings, on the same triadic situation, of pleasantness, tension,

(35)

and consistency to determine the relative contribution of balance, agreement and attraction to these different types of social judgment tasks. Effects due to agreement, attraction, and balance were significant for ali three types of ratings. Attraction, and to a lesser extent agreement, were observed to be the primary determinants of the pleasantness and tension ratings which tended to covary; only a small variation was due to balance. However, balance was the most important determinants of consistency ratings, with attraction being second in importance.

Gutman and Knox interpreted their results to mean that dependent measures based on affect are inappropriate to test Heider's theory. They also observed that, when an appropriate cognitive task—here, a rating task in terms of consistency — is employed, balance effects emerge for both positive and negative P /O relations. Contrary to Nevvcomb's assertions, Gutman and Knox found balance effects on consistency ratings also for the cases of negative attraction. Hovvever, they reported balance effects on ratings of pleasantness and tension only when P /O was positive —in line vvith Nevvcomb's predictions. They concluded that Nevvcomb's reformulations are restricted to cases employing measures of affect.

There are some irregularities connected vvith the Gutman and Knox study. Zajonc (1968) had revievved the social perception studies, and reported that approximately half of them favor agreement över balance, and attraction did not do very vvell. Since the bulk of the evidence in favor of agreement came from studies employing pleasantness ratings as their dependent measure, the fact that Gutman and Knox found attraction as the most important determinant of pleasantness ratings is not easily reconcilable.

Even though the strength of the attraction effect contrasts vvith the findings from most previous studies, Gutman and Knox state that Zajonc-indices (vvhich Knox (1963) computed from adults and vvhich Gutman (1970) and Ohashi (1964) obtained from cnildren) also shovv attraction to have contributed more to the ratings than agreement or balance.

Problems of Implicit Future Contact Assumptions:

A more serious problem vvith the Gutmaı and Knox study hovvever, is that, like some other studies, it did not control for the

15

(36)

potential lack of balance at the level of the dyad stemming from the implicit subject assumptions, or from the tacit cues in the wording of the stimulus descriptions, or both.

The possible effects of some implicit assumptions on the part of the subjedts became the topic of an experiment by Aderman (1969) who manipulated them in the form of instructions. Aderman was interested in the implicit assumption, by the subjects, of anticipation of future contact with O. He instructed the subjects to assume future contact vvith O concerning X, or future contact but not concerning X, or no future ccntact vvith O. Balance favoring ratings vvere obtained only under the explicit assumption of no future contact. Since Aderman's future contact anticipation can be interpreted to mean an addition of a unit relation to the existing triadic situation (Insko, 1974), it seems to me that it can produce an effect equivalent, in its end result, to increasing the salience of the P /O bond. This vvould be so because novv the balance prin- ciples can operate solely at the level of the dyad; particularly in cases of nonbalanoe, the future interaction ( + U ) vvould conflict vvith dislike of O (—L) to produce unbalance simultaneously at the dyadic level as vvell. Checking some of the previous studies against this idea, vve can see other examples of this confound in some of their results.

The Price, Harburg and Nevvcomb study, the Rodrigues study, the Crockett study and others that asked the subjects to supply the name of somebody they knovv in the place of O, ali could sug- gest future interaction. Gerard and Fleischer (1967) stated that they utilized descriptions of real life situations and gave the example of a consultation betvveen tvvo medical doctors who were also colleagues.

In this case, the possibility for further interaction betvveen the tvvo persons is a rea,l one. It seems reasonable to expect that vvhether or not future interaction is implied in the stories depends on the specific content of the stories. The existence of a positive unit relation is clearly implied in the follovving exeerpt from Whitney:

"Wong and Lee are good friends. They both happen to be at the same dance one night vvhen a new singing group is performing,"

(1971, p 12). Steinef and Spaulding used commpn names of people, campus stories about objects like science courses, ete., as their issues.

Since campus relations betvveen students are quite voluntary and casual, referring to a hypothetical fellovv student, unless so specified, 16

(37)

would probably not imply the presence of a unit relation.

However, the Jordan paradıgm with its skeletal form of presenting the situations does not have this protection against the intrusion of future contact assumptions. Aderman had observed that the ratings made by the subjects who were given only standard instructions and by those explicitly instructed to assume future contact with O shovved close correspondence. He concluded that a sufficient number of subjects did implicitly assume such contact even in the standard condition where no mention of it was made.

Therefore, I suspect that the Jordan paradigm which offers no word to the contrary, or some other paradigm without sufficient structure in its wording, can also be amenable to such future-contact interpretations by the subjects.

Corning backto the study by Gutman and Knox,we see that the vvording of their stimulus descriptions practically correspond to, in Aderman's study, the explicit instruction of assuming further interaction with O. The statement they included in their instructions read as follows: "It is important that you view the association between the people in the situation as continuing över a fairly long time period. The idea is that these people must remain in contact with one another, see each other regularly, and contin- ue to interact whether they like each other or not. Assume that the circumstances are such that they cannot, in the near future, terminate their association by finding other accomodation, quitting their jobs, or resigning from the committee" (1972, p. 353).

In their replication of the Gutman and Knox study, Miller and Norman (1976) also observed that preference for agreement and for attraction vvere stronger for pleasantness and tension ratings than for consistency ratings, while the preference for balance was stronger for consistency ratings than for pleasantness and tension ratings. Nevertheless, in a separate condition where they eliminated the + P / O unit relation from the instructions, the balance preference increased while the attraction preference decreased. Furthermore, Miller and Norman observed that the effect of attraction decreased, from the unit condition to the no unit condition, more for pleasantness ratings than for tension ratings, and more for tension ratings than for consistency ratings. Thus we see that, unless controlled by the experimenter, unit relations creep into the stimulus situations 17

(38)

and add to the extent of the complications already existing in the area of investigation on the balance theory.

Specification oi Nevvcombian Assertions against Heider

Having reviewed the majör findings in the area of balance theory, let us now see how Nevvcomb's reformulations of the theory stand against the accumulated evidence:

a) Does negative P / O create disengagement ? Nevvcomb's predictions had their focus on the negative P / O relationship.

With regard to this particular bond of a triadic configuration, Nevvcomb claimed that it has negative motivational consequences for the subject who, upon seeing this bond, "disengages" himself from the situation or becomes indifferent to it.

We recall that Crano and Cooper (1973) provided evidence in favor of disengagement phenomenon via "involvingness" ratings, and proposed a bi-phasic attention model based on these findings.

On the other hand, we recall that Crockett (1974) found evidence (in the form of attraction effect) for disengagement in pleasantness ratings only in the case of P-0-Q (three persons) situations on vvhich Crano and Cooper had obtained their involvingness ratings. Since Crockett had asked the subjects to identify their ovvn approved and disapproved issues, he expected that P-0-X situations vvould maximize their involvement. In this case, the pleasantness ratings shovved clear agreement effects vvith either positive or negative attraction.

Further, Rodrigues (1968) reported persistent agreement effects, even vvith the measure of completing the missing leg of the triad, vvhich one vvould expect to be especially sensitive to disengagement effects because, when using this measure, the subjects could actively avoid agreement with a disliked O. As also stated by Fuller, if agreement is judged to be more pleasant than disagreement even vvith a disliked O, P cannot be considered to have - disengaged from the situation.

b) Are balance principles inoperative given negative P / O ? Nevvcomb's contention that balance principles are inoperative in situations involving a negative JP / O relation has been borne out only partially by research. Even the studies vvhich employed some sort of an affective measure came up vvith results that do not en- tirely favor Nevvcomb.

18

(39)

For example, while Whitney reported main effects for attraction and agreement but no interaction Gerard and Fleischer found no main effects but an interaction between balance and attraction.

On the other hand, Fuller obtained main effects for attraction and for agreement, and also an interaction between attraction and agreement. We recall that Crockett's pleasantness results supported Newconıb in the case of of P - O Q situations, but supported Heider in the case of P-0-X situations.

İt is clear that balance effects are not always non-existent when P / O bond is negative and the ratings are made for pleasantness. Moreover, the Aronson and Cope (1968) experiment in which a subject develops a more favorable attitude toward his enemy's enemy than toward his enemy's friend, directly contradicts the statement that balance principles are inoperative in situations ınvolving a negative P / O relation:

An experiment was performed to explore the generality of the proposition that people like those who punish their enemies and reward their friends. Results indicated that the attractiveness of a person who punishes one's enemy or re- vvards one's friend is not limited to situations which provide indications of attitude similarity, gratitude, or social support. Specifically, the experiment was designed to show that this phenomenon occurs in spite of the fact that that the situation was arranged so that: (a) The stimulus person's behavior in no way suggested that his attitudes vvere similar to those of S ;(b) the stimulus person was clearly unaware of S's relationship to the latter's enemy or friend—thus he was not trying to help S; and (c) it was clear that S and the stimulus person vvould have no opportunity to meet and gain any social benefit from sharing a mutual friend or enemy. Methodologically, the possibility of bias was reduced by using separate Es, who vvere partially blinded as to treatment, and a third person (similarly blinded) to collect the dependent variable data, (Aronson and Cope,

1968, p. 8).

c) Do nonbalance effects occur regularly? The above studies can also be presented as evidence for the rather unreliable occurance of nonbalance effects, occasionally even when using ratings of pleasantness.

d) Do nonbalance effects occur regularly for certain types of tasks or measures ? Newcomb's assertion that balance principles will be inoperative in situations with negative P / O implies that the nonbalance phenomenon will be relevant across tasks, or across dependent measures. This point gains in importance when we remember that Newcomb's reformulation of balance theory is built on results obtained from studies that utilize pleasantness or tension ratings 19

(40)

as their dependent measure —i.e., Jordan, 1953; Rodrigues, 1967;

Price, Harburg, and Newcomb, 1966. On the other hand, Newcomb's position fared less well when subjects were required to predict missing relations in incomplete triads (Morisette, 1958; Rodrigues,

1968; Wellens and Thistlevvaite, 1971) even though some support came for Newcomb with this measure (Wyer and Lyon, 1970;

Truzzi, 1970). Nevertheless, the generalizability of Newcomb's reformulation across dependent measures is further reduced by studies whichreport balance effects for such ratings as expectedness, or truthfulness (Crockett, 1974), or consistency (Gutman and Knox, 1972). Since the results in these studies as well are mixed, even though balance is the predominant effect (e.g., Crockett 1974, p. 109), and since some did not employ adequate controls, the generalizability of nonbalance outcomes across tasks or measures has been an open issue.

e) Do the structures in which P / O is negative resemble each other more closely than they resemble the other structures?

We know that Newcomb has characterized the psychological state aroused in the subject by P-0-X triads cöntaining negative P / O bonds as one of relative indifference or disengagement (nonbalance) in vvhich principles of balance exert little or no influence. From this notion Nevvcomb derived two related predictions one of which is as follovvs: "The four possible P-0-X situations in which P / O is negative will resemble each other more closely than they resemble other situations" (1968, p. 33). İf we vvere to interpret the vvord "resemble" as "vvill not significantly differ", the evidence for this assertion is summarized by Nevvcomb as follovvs:

Jordan's results shovved a statistically significant tendencv for harmonious relations to be rated (less unpleasantly) than unbalanced ones. But...the data also reveal higher ratings of unpleasantness for Heider-balanced triads in which P /O is negative than for similarly balanced ones in which P /O is positive; and the former though balanced from Heider's point of view, are no more pleasant than those considered unbalanced by Heider. The only triads that are consis- tently rated as pleasant are the first two, characterized by positive P /O together vvith agreement, and the only two that, according to my formulation, are posi- tively balanced. The remaining six situations are undistinguishable in terms of mean ratings, as shown by inspection... And (though perhaps only by chance) the single sign pattern of these six that is rated most unpleasant ( 1- ) is considered balanced by Heider (Nevvcomb 1968, p. 35).

The Price, Harburg and Nevvcomb (1966) study itself came out vvith the findings vvhich Nevvcomb describes as follovvs:

20