‘— . w— .‘.. . ~. ~ . “-“< ‘q~‘<‘~«“-mws- NONRECIPROCITY AND THE THEORY OF COGNITIVE BALANCE Dissertation for the Degree sf Ph. D. MECHEGAN STATE UNIVERSITY RALPH E. GOOPER 1973 THEM} “ LIBRARY Michigan Stai'e University 1061002 JUN 222002 9530 .05 MAY 0 9 2005 1“ ‘- insight-uh , 21$:3:33:33323:323223333.n-n—Nnnsnfi a mam,.m,u mammmmmmmmmmmamuomaag .............._.fiwm...._.....aammmm...u...nufl.a.mug _h-___p_.hfiFFF._--,_fi__.h_fi~,-F~F~h..___~a.~__n 7 .a.a,an.“wemm,wm3.9mmanu_.uaaaomuuummueua.u.umammu. .mmmnmnnnmfinnmmmn.mn“mammnnmwmmnmflwwmmnmmnanmmm,an. .v......~..........u...............q.......‘....... ,_nnnnnwnnnnnnfinn a,m"mama"nnunnannuwflnn.nnmnnmmnnn_ .NNNNNNNNNNNN~N-~NNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNA _____________,__________________________=_____.___ Ezma:EEBS—omuafiu-5833:3233:3:33:3933:33:2333322322:; _am..=====..=.=== euaaaaega..==._=°=u.=.=qk..=°=°. fi a 1. ‘ .7 a. “cumin mu ABSTRACT NONRECIPROCITY AND THE THEORY OF COGNITIVE BALANCE By Ralph E. Cooper In reviewing Heider's work on cognitive balance theory (1944, l946, 1958, l960), it was noted that Heider was working within the larger framework of Gestalt theory in proposing the balance model. An examina- tion of the balance literature revealed several departures from this approach in the models proposed by Cartwright and Harary (1956) and by Newcomb (l968). The basic tenent of the original Heider theory appeared to be the notion of clusterability of the social-perceptual field. Further examination of the literature indicated that one of the assumptions commonly made in the area of balance research, that of reci- procity in perceptions of interpersonal affect, was not as supported by other research literature as might have been surmised given its broad acceptance and use. Rather, the most cogent available evidence seems to indicate that reciprocity does not hold for perceptions of affect between the perceiver and a disliked other (Price, Harburg, and Newcomb, l966). Based on these two concepts, clusterability and the possible non- existence of reciprocity in the one case, a revised model for cognitive balance was developed. This model involved two submodels for EQ§_situa- tions, one for reciprocal £[Q_relationships and one for nonreciprocity in negative E£Q_bonds. For the EQQ_situations, the existence of two possible loci for nonreciprocity led to the development of four sub- models, one for each combination of reciprocity or nonreciprocity in negative 319 and E£Q_relationships. Ralph E. Cooper Data were obtained from 120 subjects in each of two studies, one involving the £Q§_situations and the other the £99 situations, in an examination of the reciprocity issue and a test of the proposed model in comparison with previously developed models. Subjects were asked to supply the names of same sex peers whom they liked and disliked, and for the EQ§_study, two sides of an issue of their choice. They then inserted the appropriate initials in blanks on cards describing each situation and placed the cards in the appropriate rank order, according to their perceptions of the pleasantness and stability of the situations, with half of the subjects in each study doing the pleasantness ordering before the stability, and the other half using the reverse order. Finally, the subjects rated both their affect toward each of the indi- viduals or sides of the issue they had supplied, and their perceptions of the others' affect toward themselves. The data were analyzed to test three main hypotheses: I. That reciprocity does not hold for perceptions of affect from a disliked other; 11. That the rank orderings of balance situations are function- ally related to affect toward and from the persons and issues in the situations, particularly perceived affect from a disliked other; and III. That the cluster-reciprocity model is a better predictor than either of the two comparison models. Support was obtained for Hypotheses I, in that a substantial number of subjects reported that a disliked other liked them, and affect from disliked others significantly exceeded affect toward disliked others. This result was interpreted as demonstrating the invalidity of Ralph E. Cooper the assumption of reciprocity in the case of affective relations involv- ing disliked individuals. Evidence relevant to Hypotheses II was interpreted as being equivocal, due to the small size and apparent inconsistencies in the correlations obtained, both between the affect measures and individual ranks and between the affect measures and the results of the unfolding procedure. Further research regarding both Hypotheses I and II was recommended. Regarding Hypotheses III, it was observed that none of the three models accounted for more than twenty-five percent of the subjects' orderings on an individual basis, or more than one-fourth of the vari- ability in those orderings when rank correlations were computed. Thus these results were interpreted as indicative of a general failure of models of this type to describe the psychological processes involved in evaluations of the traditional balance stimuli. Additionally, some implications of the support observed for Hypothesis I were discussed, particularly with regard to the use of methods of research requiring an assumption of reciprocity, and the restrictions these results place upon researchers using such methods were outlined. NONRECIPROCITY AND THE THEORY OF COGNITIVE BALANCE By Ralph E. Cooper A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Psychology 1973 ACKNOWLEDGMENTS Most students who successfully pursue higher education owe a great deal to many sources of influence during the course of their lives. In my own, several stand out because they exceeded the normal expectations of their positions. Among these, a high school teacher, John Moore, conveyed the excitement of science and the enjoyment of teaching, providing the impetus to find a career requiring both. Two professors, Larry Messe and Jim Phillips, have been more than members of a student's committee, but have provided touchstones for the devel- opment of both academic and personal life-styles. Most credit must go to Bill Crano, who, by his willingness to work with and for his students, is a model for any advisor; he has been mentor, colleague and friend. Finally, a wife and two sets of parents deserve credit, for their expectations have ever been goals just at the edges of reach, goals without which this effort would not have been undertaken. ii TABLE OF CONTENTS LIST OF TABLES . . . . o . o . o . . . INTRODUCTION . . . . . . . . . . . . . o . . ..... A Problem . . . . . . . . . . . . o . . . Reciprocity . . . . . . . . . ..... . . . . . . A Solution . . . . o . . . . . . . o . o . . Hypotheses . . . . . . . . . . . . . . ..... METHODS . . . ...... . ..... . . ....... . Study I . . . . a . . . . . . . . . . . . o . Study II . . . . . . . o . . . . o o . . . RESULTS . . . . . . . . . . . . . . . . . . . Definitions . . ....... . . . ....... . Affective Relations . . . . . o . . o . . . Affective Measures and Ranks Assigned to Situations .......... . . . . . . . . . . Tests of the Models . . . . . . . . . . . DISCUSSION . . . . . . . . . . . . . . o . . Affective Relations . . . . . . . ..... . ..... . Ranks and Affective Relations . . ....... . . . The Models . . . . . . . . . . . ....... . . On Necessity . . . . . . . . . . . o . . . . 10 16 21 22 22 27 30 30 30 41 54 52 62 64 66 67 LIST OF REFERENCES ........... . . o ........ . . APPENDICES A. INSTRUCTION SHEET FOR STUDY I . . . . .......... B. SAMPLE SITUATION CARD . . o . . . . . . o . . . . . o . . . C. ANSWER SHEET FOR STUDY I . . . . o . o . o . . . o . . . . Do AFFECT QUESTIONNAIRE FOR STUDY I . o . o . . . . . . . . . INSTRUCTION SHEET FOR STUDY II . . o . o . . . . . o . . . ANSWER SHEET FOR STUDY II . . . . . o ......... . . AFFECT QUESTIONNAIRE FOR STUDY II ......... . . . . iv 69 74 75 76 78 79 80 82 10. ll. l2. LIST OF TABLES Stability Scale Values from the Crano and Cooper Study (1973) . . . . . . . . . . . . ..... . . . Application of Semi-Cycles Analysis to All Positive Q[£_Relationships . . . . . . . . . . . . . . . . Reciprocal and All Positive QLP_in Cluster Analysis of EQ§_Situations . . . . . . . . . . . . . . . . Cluster Analysis of PQQ_Situations involving All Positive 913. . . . . ..... . . . . . . . . . . . . Correlations between Affective Relations ..... . . . . Differences in Affect Correlations . . . . . . . . . . . . Means and Variances of Affective Relations . . o . . . . . fl Correlated Variances Test (3) on Affective Relations . . . . . . . . . . . . . . . . . . . . . . . . . Self-To-Other vs. Other-To-Self Affective Relations . . . . . . . . . . ..... . . . . . ..... Number of Disliked Others Reported as Liking the Subject . . . . . . . . . . . . . . . . . . . . . . . . Means, Variances, and Medians of Ranks Assigned to Situations . . . . . . . . . . . . . . . . . . o . . . . Stress Remaining Between Solution and Data After Multidimensional Unfolding . . . . . . . . . . . . . . . . 15 18 20 32 33 34 36 37 39 42 43 13. 14° 15. l6. 17. 18. Maximum Correlations Between Ideal Points and Affective Relations . . . . . . . . . . . Number of Orders Versus Number of Observations per Order . . o . . . . . . . . . . . . . . . . Correlations between Affect Measures and Ranks Assigned to the Situations . . . . . . . ....... Proportions of Observations Fitting Each Model Average Rank Correlations Between Observations and Models . . . . . . . . . . . . . . . . . . . . Average Correlations Between Observations and Cluster Submodels by Reciprocity Conditions in Data 0 O O U 0 O 0 O O 0 O O 0 0 D 0 0 0 vi ....... 49 51 55 59 60 INTRODUCTION While in recent years much social psychological endeavor has focused on cognitive consistency theories (see, e.g., Abelson, Aronson, McGuire, Newcomb, Rosenberg, & Tannenbaum, 1968), relatively little attention has been devoted to Heider's principle of cognitive balance (Heider, 1946), the basic model from which these are, in essence, de- rived. Heider's primary interest as a cognitive psychologist was the perception and cognition of social objects and their interrelations (see Jordan, 1968). In an early paper with Simmel (Heider & Simmel, 1944), Heider reported the results of an experiment in which subjects watched two triangles and a disc move about a simply structured screen in an animated movie. When asked to describe what they had seen, almost all subjects analyzed the cartoon in terms of living beings and of events organized by the perceived motives of those beings. That this should lead to the connection of social perception with Gestalt principles is not surprising, nor is the drawing of that conclusion, expounded more fully by Heider in another paper that same year, unexpected (cf. Zajonc, 1968). In this second paper, Heider (1944) proposed that social percep- tion and cognition were influenced by configural forces in the same manner as the perception of physical stimuli, that social-perception "good-figures" and social-cognition pragnanz are tenable concepts (see 1 also Heider, 1960). In this paper, Heider discussed how this force toward simpler social perceptions arises and examined many examples drawn from diverse areas of human behavior. It is in this paper that Heider introduced the concept of the unit or unitizing relation--a tendency to see effect and cause, when these are social events, as being one social and psychological entity. In his first formalization of the principle of cognitive balance, Heider (1946) set forth the conditions under which the configural or unit-forming forces would be in equilibrium, using only the valences of the relationships between people or between people and objects. In perceptions involving just two people (the self, P, and an other, 9), Heider assumed that a positive relationship would tend to cause a unit to form, bonding the two "objects" into one cognition. Thus if both £19_and 912} were positive, a stable cognition would result (much like two small objects close together on a large screen being perceived as a unit or group). If both £[Q_and Q[E_were negative, then there would be no unit-forming tendency--two separate objects would be perceived and this would be stable. Heider called these equilibrium states balance. However, if one relationship were positive and one negative, then the unit-forming tendency of the positive relation would be counter- acted by a tendency to separate objects related by negatives. The resulting state of disequilibrium was called imbalance. 1Throughout this paper, references to interpersonal affective relationships are to those existing in the perception of P, unless other- wise specified. This is consistent with the cognitive-perceptual basis of Heider's theory. Thus Q[E_indicates Efs perception of gfs affect toward 3, Heider similarly specified conditions for the situations involving three people (3, Q, and Q) or two people and an object (395). In discussing these situations, Heider assumed that each dyadic rela- tionship was balanced. From Heider's Gestalt viewpoint, this assumption 2 seems only logical: To talk of a of balanced dyads, or reciprocity, "good-figure" when parts of it are inconsistent requires some special consideration. Thus Heider discussed the EQ§_and ng_situations as if there were only one relationship between each pair of social objects and considered three different situations as balanced or in equilibrium. If all relations were positive, then the unit-forming forces would be in harmony if the three objects were treated as a group or a unit. (The perceptual equivalent is three small, fairly close objects on a large field.) If two relations were negative and one positive, then the two objects linked by a positive bond could be treated as a single entity in relation to the other. (Two objects close together and another one farther apart is the perceptual analog.) Finally, in contrast to later extensions (e.g., Cartwright & Harary, 1956), if all three relations were negative, the treatment of the three objects as separate entities seemed consistent with the other situations. However, in keeping with his Gestalt orientation, Heider noted, ". . . the case with three nega- tive relations does not seem to constitute good psychological balance, since it is too indetermined (Heider, 1946, p. 110)." Rather than finding forces against balance, Heider noted that three psychologically dispersed elements lacked forces M balance. (The argument on the three negative situations--that it is neither balanced, because it lacks —_z 2The reader is referred to footnote 1. unitizing forces, nor imbalanced, because there is no conflict between unitizing and separating forces—-is of special importance in this paper. That the argument of indeterminism could as well be applied to the negative-negative dyad seems to have escaped Heider and other balance theoriests as well. It should be noted that there is no cognitive or perceptual reduction in considering two objects as two objects or three objects as three objects.) A similar type of analysis can be applied to the remaining triadic situation as well. With one negative and two positive relations, the force to unite objects with a positive bond contradicts the tendency to keep apart those with a negative relationship. If, for example, £19_ and QLQ_are positive, and E£Q_is negative, there would be a tendency to treat £.and 9.as an entity; this entity then has a conflicting relation- ship to 9, being in the one case positive and in the other, negative. There is no means by which the situation can come to perceptual equilib- rium without changing at least one relationship. Heider's model is notable for its essential conceptual simplicity. By knowing the signs of the relationships in a situation, we can classify it as balanced or imbalanced. Heider considered imbalance to lead to a lack of stability in the perceived relations or to tension: "If no balanced state exists, then forces towards this state will arise. Either the dynamic characters (attitudes or affective relations) will change, or the unit relations will be changed through action or cognitive reorgani- zation. If a change is not possible, the state of imbalance will produce tension (1946, pp. 107-108)." Heider's emphasis is again upon the individual's perceptions and cognitions and not with the objective reality of interpersonal relations. In several senses, the prototype for research testing the balance principle has been Jordan's dissertation (reported in Jordan, 1953). Though a student of Heider's, Jordan introduced two modifications which have been rarely challenged in the psychology literature. First, Jordan placed the three-negative triad definitely within the category of 19; balance, even though Heider himself still had reservations about this case (see Heider, 1958, pp. 202-203, 206). This modification, however, permitted Cartwright and Harary (1956) to formalize balance in terms of a product rule--that a dyad or triad is balanced if the product of the signs of the relations is positive and imbalanced if it is negative. Second, Jordan operationalized balance and imbalance as the perceived pleasantness or unpleasantness of the situation. That almost all balance research has accepted this definition is historically understandable-- psychologists, accustomed to thinking in terms of needs and the tensions produced by their unfulfillment, would be more comfortable with measures of tension or unpleasantness than with measures of the changeability or stability of relationships. The strength of this effect can be found in the emphasis that most reviews of the balance literature place on tension (see, e.g., Zajonc, 1968; Phillips, 1971; Taylor, 1970). As noted above, Heider believed that a change in a relationship, either through action or cognitive distortion, would be the primary effect of imbalance, with tension resulting only if change were not possible. The concept of un- pleasantness is thus at least two transformations (change to tension, tension to unpleasantness) removed from the primary result of imbalance POStulated by Heider. This choice of operationalization may have played a keuv role in the conflicting results obtained in much balance research. In addition to these two modifications, Jordan's research raises several other problems for balance theory, some of which are directly attributable to his methodology. Jordan raises subjects with situations like: "I dislike 0; I like X; 0 has no sort of bond or relationship with X (Jordan, 1953, p. 271)." Subjects rated each situation in terms of pleasantness or unpleasantness they experienced by placing a mark on a line with end-points "best" and "worst" and a mid-point of "neutral." Jordan reported that his data did not seem to fit the balance model in two cases--when E_was in a disliking relationship to Q_in otherwise balanced triads, and when there was a negative "unit" (as opposed to affective) relationship in the triad. The first of these is central to this thesis and will be discussed more fully in another context. The second has been a considerable problem for balance theorists in that Jordan found a qualitative difference between affective and other unit relations. This problem may be considered as a functional difference between the compliment and the opposite: Whereas "dislike" is the opposite of "like," “did not cause" is the compliment, not the opposite, of “caused." In this instance the effective opposite of "caused"--and the functional equivalent of "dislikes"--would be "prevented“ (see Cartwright & Harary, 1956, and Harari, 1967, for discussions of this problem). Most balance researchers have concentrated on situations involving only affective relations and positive "unit" relations and effectively have thus avoided this controversy (see, e.g., Rodrigues, 1968). The solution, however, seems to lie in recognition of the logic of opposites and with the suggestion that affective relations are, in effect, only one class of unitizing relations (Jordan, 1968). Thus Jordan's dissertation began a trend away from Heider‘s more systematic theoretical base, toward a conception of balance as simply a model for predicting interpersonal relations and attitudes. The concern over issues raised by this research, together with the mathematical sophistication of the work of Cartwright and Harary (1956) led research and theory away from a deeper consideration of Heider's basic proposi- tions. The research since 1953 has been progressively less directly related to Heider's theoretical orientation and more oriented to issues such as the abstractness of the stimuli, to positivity (e.g., Zajonc & Burnstein, 1965), agreement (Rodrigues, 1968), and quantification (Weist, 1965). Recently, some research has concerned itself with correcting some of Jordan's diversions, most notably the issue of the appropriate measurement variable for balance (Crano & Cooper, 1973; Gutman & Knox, 1972). The purpose of the present paper is to return to an examination of the original premises on which Heider's model is based, to examine certain critical assumptions made by Heider (and since then generally accepted), and to attempt an extension of the model with the goal of a more accurate prediction of the order of the triadic situations. A Problem Several writers, most notably Newcomb (1968), have noticed in balance research certain regularities within the two states in the Cartwright and Harary extension of Heider's theory and several instances of results contradicting this model. On the basis of some of these and the results of the Price, Harburg and Newcomb (1966) study, Newcomb proposed a three state model as a function of the sign of the E£Q_bond. Newcomb reasoned that when E[Q_is negative, P_has difficulty in inter- preting what the Q_/_Q bond means, and it is likely that _P_ will not care what Q_thinks if he is disliked. Thus Newcomb suggested that when PLQ_ is negative, the person is not engaged or involved in the situation and so is not concerned with balance. On the other hand, when ££Q_is positive, the individual, in Newcomb's model, is engaged by the stiua- tion and thus balance forces operate. The result is a three-state model of positive balance, non-balance, and positive imbalance, with the middle level consisting of all situations with a negative Elg_bond. In review- ing the relevant data, Newcomb found evidence to support his model relative to the Cartwright and Harary two-state classification. Crano and Cooper (1973) have tested this extension by means of a methodology which permitted a simultaneous check on the scaling prop- erties of the dependent variables employed. They report that their data for stability evaluations tend to support the Newcomb extension, but that the pleasantness results were not scalable.3 Yet Crano and Cooper note regularities in their data that are not accounted for by the Newcomb extension, some of which parallel results reported by studies using the disputed pleasantness variable. Table 1 shows the results of the Crano and Cooper research on both EQQ_and £9§_situations. It can be seen, for 3This result seems consistent with comments above regarding Jordan's research (1953). Crano and Cooper suggest that the nonscal- ability of pleasantness evaluations, consistent over three independent sets of subjects, may account for some of the inconsistencies in the literature, for example, the relatively different ratings of some situa- tions in the Price, Harburg, and Newcomb (1966) paper and Jordan's original results. This non-scalability seems to arise from intransi- tivity in the use of pleasantness as a judgemental variable, indicating that, even within a single subject, pleasantness is not a consistently defined dimension. vaoo. Amvoo. Amvoo. ARVNM. Akvoe. A fivem. onme. AmVFo. Amvmm. Anew“. “mews. Amv_k. Amvmm. Amvmm. Amvmo. Aevkm. Aevmk. Aevoo. Amvmm. vaom. Amvom. A_V_N._ AFVN_.F Apvmm._ mm ppm mmpmsme dummy m suaum mmFmE Amvmo. Akvm_. Amvoo. Amvoo. vaoo. Amvoo. Amvoo. Amvoo. Akveo. ANVPN. “seem. akvmp. Amvme. fleece. Amvom. Levee. Aevme. onme. Amvmm. Amvmo. Amvme. Amvpm. Aevom. hmvmm. Aevmm. Amvkm. A¢v¢m. Leena. Amvoo. Aevmm. Amvpm. onmm. Amvkm. Amvek. “memo. Amvmm. Amvmm. ANVmN.P Amvow. ANvNO.P Amvmm. ANVF~._ A_V¢F.P A_V_m._ APVwO.F APVmN.P A_V¢_._ “PVP¢.F mm :m mmFmEmL. mm—mE mm :0 mwmemL. mmFmE ammmv N auagm «mmmv P sassm Ammmpv acupm emaoou can cameo mgp seem mos—m> m_mom xprpnmum F mpnmp “Iv Amy Adv Amy on Auv Amy A< umpzmwmz we mmmgm>< mo mmmgw>< eo comm eo :mwm mo cmwm _ mawgm=o_pmpa¢.mum a>wuwmoa ”_< o“ mwmspmc< a_u»u-team to cowpmuP_an< N manh 16 a procedure on the 3-cyc1es with situations not differentiated by this method designated by the same lower case letter. It can easily be seen from Table 2 that a Cartwright and Harary type semi-cycle analysis does not discriminate among the critical pairs of situations differently than do the Newcomb or the Cartwright and Harary models. Thus we are con- fronted with differences which none of the previous approaches to balance seem to be able to explain in a consistent fashion. A Solution The solution to this problem would seem to lie in reconsidering the assumption of dyadic reciprocity together with a return to some possible principle derivable from the Gestalt framework within which Heider was working. Gestalt theory, as expounded by Heider (see especially, 1960), includes as a motivational construct a tendency toward representing any cognitive structure in the simplest possible form, by application of Gestalt principles. Heider's “unit-formation" principle can be expanded into a model which expresses this simplicity concept and makes more detailed predictions regarding the order of triadic situations. Heiders "unit" concept is, in this respect, very similar to the notion of clusters and clustering as put forth by Davis (1967). To Heider, a unit is two objects which are positively related and which are similarly related to (or coact with regard to) other objects in the environment. This is precisely the definition of a cluster given by Davis, except that Davis included more than two objects in a cluster. Heider's concept of imbalance is the inability to form a unit because of a failure of two positively related objects to coact with regard to 17 another object, a principle which defines the conditions under which clustering is not possible. Thus a model which offers to make a distinction between those situations which are clusterable and those which are not, and distin- guishes among clusterable situations by the number of clusters, seems very close to the basic foundation of the Heider theory. It is this model, together with the added consideration of non-reciprocity, that is proposed as an alternative to those heretofore expounded. The left side of Table 3 presents the proposed model for situa- tions in which the Qifl_bond is always the same as the £19_bond (thus only the £19, £15, and Q£§_bonds are considered). It may be seen that this model is highly correlated with Heider's original model, but also more definitive in its predictions. For situations where the 913 bond is positive irrespective of the E£Q_bond (the result of reciprocity except when E£Q_is negative), the model becomes somewhat more complex, and the situations must be considered as the hybridization of two distinct parts in order to apply the model. E_is, in the first of these, the self as actor or emitter and in the second, the self as receiver of affect; his reSponse is a hybrid of these two situations. The right side of Table 3 presents the results of this process with "_£Q_Clusters" referring to E_as emitter and "Q[£_Clusters" to P_as receiver. The resulting rank order is pri— marily by clusterability with more clusterable situations ranked ahead of the less clusterable, and within a level of clusterability, by the number of clusters. 18 m.n m.m m.¢ m.¢ xcmm mgvmanoo mpawmmoa go: we mcwempm=FU mews: meowumzuwm mmpmcmwmmok m.n x c m « m.N . . N N e F . N . m N m m m m.m . N m N m.m . N m N N N N m N F P _ F N mgmumzpu mgmumspo Ncmm .mum o\a xcmm LmNm=_u m>PaPmom.mum _moogawumm.mum mcoNpmzpwm xoa to mwmzpmc< gmumz~u :p.mum m>wwwmo¢ -< new Pmuogawomm m mpamh A-++v A+-+v A++-V A---v A-+-V A+--v A--+v A+++V LIJLLLDI (MUD =ONNN=NNW 19 If the rank orders in Table 3 are compared with Study II in Table 1, it can be seen that the order for reciprocal 9£P_parallels the data for male subjects and that the order for all positive Q££_follows the data for female subjects. A hypothesis could be that these two popula- tions differed to some extent on the reciprocity issue, thus producing the differences in results. The last column in Table 3 represents a ranking of results which might be expected if a population consisted of equal numbers of subjects fitting each of the two conditions of 913_ affect; comparison with the results for all subjects in Study II of Table 1 shows a strong correspondence. A similar type of analysis can be performed for the three person situation (399), but the task is complicated somewhat in that either the Q££_or the Q£E_relationship might not be reciprocal. In the case where both of these relationships are reciprocal, the left side of Table 3 is still appropriate. Similarly, if Q£E_is positive and Q£E_is reciprocal, then the right side of Table 3 applies. Table 4 presents this analysis for the other two cases. On the left side of the table, for Q£P_positive and QLP_reciproca1, only the non-reciprocal relationship is considered and the analysis proceeds in a fashion similar to that presented in Table 3. The right side of the table presents the four combinations relevant for the remaining case in which both Q£E_and Q£E_are positive. Comparing this model to the data in Table l is difficult and, at best, an academic exercise. Given the four orders and the three degrees of freedom available for estimating the proportions of a group that would fit each order, there are many orders which could be predicted for a group of subjects. Thus, new data, including measures of the relevant affect variables, are required to obtain any sense of the possible 20 m.¢ m.¢ NORM xcmm mpawmmoa we: we mcwgmamapu mews: meowumzuwm mmumcmwmpo .1 .mum quosapumg mcw>po>=m mcowumauwm.mmm Low m mpnmh mmm « w k e p k F k F p r r a N N m e r N N F e k N p N k N p F p P memummvw mgmpmz mgmummvw mgmum: Max; mums; “3.: «3.: 9.3.58 Sm EN mum 58 ma 2:3: :< m.N x a w P e m.n « a m N m N N N m.m * N m.m a N F p P msmumapu msmumzpu 5.52. ME NH pmuoagwumm.mum new m>mummoa.mum m=F>Fo>2H meowpmzpwm.mmm No m_mxpmc< Lawmapo e «Pack — TIV (53: T: :1 73 TL :3 T: A1,; LI. mpmg apuuogwu mmocg mew mcowpmpmgeou vocwpgmccs mN.o Nm.o xxxx _—.0 mo.o no.0 Np.o .mmqm oo.o oo.o N~.o .o_ Fo.n no.o mo.o xxxx .m n.m o..- eo.o _o.o me.o xxxx om.o N..- No.- wo.- _..o .qum NN.o .N no.1 pp.o Fo.u mm.o m¢.o xxxx mp.u Fo.u —o.o No.1 mv.o on.o .N nm.o ep.o mp.o mp.» mo.u op.u xxxx mo.u ¢F.o o~.n op.n No.1 .m m..o .mmmm No.0 oo.o No.0 No.o Nm.o xxxx me.o mo.- m..- mo.- .m m mpnm» wo.o mo.o .mmqm mo.o mo.o oo.o mo.o mp.o xxxx Np.o No.1 ¢P.o .q wo.o mo.u m~.o wwwmm om.o mp.o mo.s No.1 op.o xxxx op.o mo.o .m mcopumpmm m>Fuumem< cmmzpmn mcowpmpmggou ‘ a .choma.c.wmw 3°.mn m. .. auspm .mzmmw cm Eogm powwmm pcmmmgamg quoz mmwgpcm mcwmmws ms» ”chommpu ms» m>onm mp Axoav H zcsumm No.r No.1 Np.u NN.o Ammgw mm.o mo.u Fo.o oo.o mm.o xxxx Nm.o 0N No.1 No.0 no.0 NN.o m¢.o mmnm mp.u No.0 no.1 mN.o mm.o xxxx .F U) 0 4.: LL. (I) O 4.: LL! U') D 4.: D V: O 4.: Q U) 0 4.: < LI. 0 4.: W LL! 0 4.: U‘) D O 4.: U) U 0 4.: m op < ON m U) C V) O 4-) m FNMQ’LO‘ONQO‘CDF-N l—l—f— (I) o =0.NN.N¢ 33 Table 6 Differences in Affect Correlations S to +L with +L to S vs. S to -L with -L to S r+L r_L ta p Study I (295) r's .74, .70 .44, .40 2.985.25 Study II t_ 4.92 13.70 8.48 .63 36 (£99) p <.ooos (.0005 (.0005 >.25 aNumbers in parenthesis refer to rows in Table 7, specifying the variance used in the test. bThe degrees of freedom used in each test was 118, with recognition that a larger number could have been used (either 238 or 358 for Study I and Study II, respectively) but would have resulted in a less conservative test. > to S j H O co H'fi co m MO ('1' 0.0 U) U H- O M U: ‘4. O U rn (up 0 U" U3 ('0' o m '11 «+4-0- DO U) vs. VS. VS. V5. V5. V5. aA positive mean difference indicates that self—to—other Table 9 Self-To-Other vs. Other-To-Self Affective Relations Study I (£929 Mean Difference -.425 -.342 1.91 2.57 a (Matched t) E -4.95 -4.02 6.88 8.77 P C .0005 ( .0005 (.0005 (.0005 Study II (399) Mean Difference .t -.225 -3.29 -.183 -2.21 -.183 -1.51 2.20 7.43 2.23 7.22 2.09 7.78 affect was more positive (liking) than other-to-self affect. 37 (.001 (.05 ).10 (.0005 < . 0005 (.0005 38 appears to be supported in that perceived liking from disliked others exceeded affect toward disliked others. Although no prediction was made concerning differences for liked others, the fact that these were, in general, significant and in the opposite direction from differences for disliked others provides an interesting contrast. Finally, Hypothesis Ia predicts that a substantial portion of the subjects would indicate that a disliked other liked the subject, i.e., that at least one 4L to S relationship would indicate liking. Table 10 presents the distribution of the number of -L to S responses on the liking side of the affect scale midpoint. Inspection of Table 10 reveals that about 39 percent of the subjects in Study I and 51 per- cent of those in Study 11 reported that at least one disliked other liked the subject. For Study 1, the lower confidence limit (p = .05) for .39 is approximately .30, indicating that 39 percent significantly exceeds any percentage smaller than 30. For Study II, the similar confidence limit for .51 is .42 or 42 percent. Thus both studies appear to have substantial numbers of subjects indicating that a dis- liked other also likes the subject, thereby supporting Hypothesis Ia. Table 10 also presents a means of examining the independence of reciprocity-nonreciprocity within subjects. The question of whether reciprocity tends to be the consistent perception of a subject, or varies within a subject, can be tested by comparing the observed distribution with one based on an assumption of independence and using the overall percentage of observed liking from disliked others in a binomial expansion. The expected distributions are included in Table 10 and differ significantly from the observed distributions (3? = 9.06, df = l, 2 p (,001 and g_ = 18.86, df = 1, p.4,001 for the two studies respectively). 39 .xucmzamcw umpumaxm zoF op mac N Foo.v. F New.mF oo.oNF oNF o.ooF mm.N m m.N om.FN «N o.o~ mo.Nm mN N.¢N eF.m¢ mm m.m¢ am we .oz aw Fo am we umpuaaxm Lanszz ucmugma memv .. auzpm _Nuahnam as» x so» umaaoem mew: mmFLommpmu ozu ummF mg» .FF Nuaum eon .mqum ucFoa mF co N sauces» F mmmcoammg mqum Foo.v. F oo.m oo.ONF ONF o.ooF mN.N «F N.FF om.m¢ mm m.NN mN.om mm No.0m am we .oz mm Fo am we umuumaxm Lmnsaz pcmugmm ammmv . xezpm m:.... m. 666.686N 6.65.0 66.....o .6 .66232 oF anmF N F _mNOF o .w m:.... 6.6;No 66.....o .6 Langsz 40 The differences between the observed and expected distributions are similar in the two studies in that there were fewer subjects than expected reporting a single positive'eL'to S relationship and more than expected who either reported no positive'éL'to S or gave two or three such relationships. This pattern of differences shows a significant lack of independence in the -L to S responses of the subjects. The same hypothesis can also be tested by examining the correlation coefficients from Table 5 for D to S, E to S and'f;tg_§, For Study 1, the coefficient is .32 and for Study II, .32, .28, and .35; all of these correlations are significantly different from zero (p4(.01, Diem & Lentner, 1970). However, the percentage of variance accounted for by these correlations is sufficiently small (range 7.84 percent to 12.25 percent) so as to contraindicate the conclusion that there is a consistent tendency within individuals regarding the perception of affect from disliked others. ngigw, Support was found for all parts of Hypothesis I in that: a) A substantial number (39 percent and 51 percent) of the subjects reported at least one -L to S as a liking relationship; b) -L to S ratings indicated significantly more liking than S to -L (p (.005 in all cases); c) L_tg_§_and §_tg_L ratings were more highly correlated for liked others than for disliked others (rav = .72 vs. .42, p.4.005 and rav = .68 vs. .51, p 4.05); and d) Variances for -L to S ratings exceeded the variances for both l_.__i_:_g___S_ and S to -L affect (p4.0025 for both comparisons). In addition, it was observed that: a) Although there is a significant lack of independence in the perceptions of subjects re— garding affect from disliked others, the relationship is not so strong as to justify treatment of these perceptions as a consistent tendency within subjects; b) +L to S ratings indicated significantly less liking 41 than did S to +L affect in all but one instance, in which the difference was in the same direction but not significant; and c) There was a greater variance in S’to -L responses than in'S't0'+L ratings (p(.0005 in both studies.) Affective Measures and Ranks Assigned to Situations Summary data for the ranks assigned to the situations are pre- sented in Table 11, including means, variances and medians. The results for the stability rankings in Study I are directly comparable to those reported in Table l for the Crano and Cooper (1973) Study II. Thus, situation A is distinctly the most stable situation, and situation B is clearly the second most stable. Situation E and F are ranked closely together followed by C, D, and lastly the pair G and H. The pleasantness data are quite different from those that would appear to be expected on the basis of the two traditional models discussed above and most past research, particularly the high ranks (unpleasantness) assigned to situations C and D. The stability rankings for Study 11 also appear to be somewhat similar to those reported by Crano and Cooper in their first and third studies (£99), with situations C, D, E, and F all ranked relatively closely together. For pleasantness rankings, however, the results are again somewhat different from the more usual expectations, especially the ranks assigned to situation C. Hypothesis II suggests that there will be a relationship between the rank orders and the affective measures, particularly the ;Q££_bond in the EQ§_triad and the ;Q[£_and'zglf_bonds in the PQQ_triad. There are several approaches to the problem of relating a rank order of objects to 42 Table 11 Means, Variances, and Medians of Ranks Assigned to Situationsa Pleasantness Stability Situation Mean Variance Median Mean Variance Median Study I (39);) A (+++) 1.71 1.27 1 1.97 2.48 1 B (+--) 2.21 1.81 2 2.59 2.94 2 C (-—+) 6.39 3.09 7 5.25 4.26 5 D (-+-) 6.29 2.93 7 5.64 3.66 6 E (---) 4.47 2.58 4 4.43 2.66 4 F (-++) 4.34 2.47 4 4.38 3.52 4 G (+-+) 5.33 3.21 5 5.92 3.25 6 H (++-) 5.25 3.63 5 5.82 3.67 6 Study II ([99) A (+++) 1.26 1.04 1 1.71 2.14 l B (+--) 3.40 2.79 3 3.39 3.16 3 C (--+) 5.59 3.52 6 4.63 4.19 4 D (-+-) 4.81 2.63 5 4.45 3.11 4 E (---) 5.29 3.64 5 4.53 4.27 4 F (-++) 4.37 3.20 4 4.99 3.19 5 G (+-+) 5.02 3.64 5 5.77 3.15 6 H (++-) 6.25 4.61 7 6.53 3.89 7 aLow numbers indicate pleasantness or stability. 4223 other variables, and each of these will be discussed in relation to the present hypothesis. The first approach is to represent the rank order by a single measurement, and then to relate this measurement to the variables of interest. The technique used in the present case is Kruskal's (1968) adaptation of Coombs' multidimensional unfolding method (see Coombs, Dawes, & Tversky, 1970). This program produces output that includes a space of subject ideal points and a set of stimuli values such that the distance between a subject's ideal point and the stimuli points are a monotonic function of his rankings. Because the solution with a large number of subjects is rarely perfect, the program operates by minimizing a measure of the degree of non-monotonicity called stress. Because of the limitations of the capacity of the computer and the program, the data for each of the two studies were divided into four equal groups, and solutions were obtained for each replication for l, 2, and 3 dimensions. Table 12 presents the degree of stress remaining for each group at each of these numbers of dimensions. It should be noted that the degree of stress is smaller as the number of dimensions increases, a factor inherent in multidimensional scaling because of the ability to represent the data more accurately in more dimensions. The numbers reported in Table 12 do not represent large amounts of stress for these numbers of dimensions and any differences in the degree of stress between replica- tions are probably not meaningful. There is some problem in the interpretation of the multidimen- sional scaling data reported here which may reduce the meaningfulness of any associated results. The spaces resulting from the unfolding technique seem to have been strongly influenced by the data of a few Table 12 Stress Remaining Between Solution and Data After Multidimensional Unfolding Replication l 2 3 4 Number of Study I (395) Dimensions A. Pleasantness l .212 .249 .297 .370 2 .189 .174 .228 .234 3 .156 .026 .129 .140 8. Stability l .343 .376 .283 .423 2 .118 .292 .225 .275 3 .105 .210 .129 .179 Study 11 (£99) A. Pleasantness 1 .240 .225 .194 .240 2 .240 .137 .167 .158 3 .162 .096 .126 .128 8. Stability l .364 .379 .391 .362 2 .264 .262 .261 .263 3 .132 .173 _ .220 .195 44 observations in which situation A was given rank 3 or greater or situa- tion B rank 4 or greater. This influence resulted in a very tight clustering of most ideal points at the center of the spaces, thereby discounting, to some extent, the variability within these clusters. One solution to this problem is to arbitrarily remove those subjects from the sample and repeat the unfolding program. However, this solu- tion does not appear to be methodologically defensible, especially since the process of arbitrary removal of subjects could be repeated at the discretion of the experimenter until the only observations remaining were those appropriate to the hypotheses. This factor, together with the phenomenal time and expense of such a process, effectively pre- cludes its use. Relating the ideal points to the affective measures can be accomplished by the calculation of correlation coefficients if the space is unidimensional, since the single dimension is unique up to a change in sign. For a multidimensional space, since the solution is unique only up to a rotation, correlations between the coordinates of the points and the affective measures would depend on the arbitrary choice of the dimensions. To overcome this problem, Kruskal and his associates have programmed a least squares method called Profit, which locates in a space a vector composed of external measurements on some variable (in this case the affect measures) by minimizing the sum of the squared distances between the subject's point in the space and his loca- tion on the vector variable. The goodness of fit of this vector is analogous to the correlation coefficient in the unidimensional case, and in fact, the program output includes the maximum correlation co- efficient possible between the vector and the ideal points, ignoring 45 the sign of the coefficient. Table 13 presents these maximum possible correlation coefficients for the pleasantness and stability solutions for the four replications in each study. Examination of Table 13 can be undertaken in several fashions. First, since the data should be best represented in the solutions having more dimensions, any functional relationship between the rank orders and the affect variables should appear as a stronger correlation as the number of dimensions increases. Second, given that there are four inde- pendent replications each representing 60 observations in Study I and 90 in Study II, any meaningful relationship would be expected to appear in all four replications, or at least three of the four. Finally, the correlation coefficients significantly different from zero at the 5 per- cent and 1 percent levels are underlined by one and two lines respec- tively, for easier inspection. The problem of multiple tests in this instance is somewhat offset by the existence of the four independent replications. Thus, for present purposes, any discussion of the results presented in Table 13 is predicated on the joint requirement that sig- nificance be observed in three of the four replications. Applying these approaches to Study I in Table 13 reveals that there is no relationship consistent across the four replications. For Study II, in three dimensions there are four cases in which at least three of the four correlations are significant. For pleasantness rank- ings these are PLrO, Pj-Q, and -O[P; the last of these also appears for the stability rankings. Thus there appears to be some evidence, though not particularly strong, for a relationship between some of the affect measures, especially for disliked others, and the observed rank orders. Table 13 Maximum Correlations Between Ideal Points and Affective Relationsa Pleasantness Stability Replication 1 2 3 4 l 2 3 Study I (_P_O_)_(_) Relation A. One Dimension Solutions £119_ .03 .02 .10 .19 431. .16 .04 21:9, .16 .16 .22 .04 .17 .01 .14 £115. .10 .OO .08 _.____3_9_ .09 .24 .11 P -X .09 429_ .21 ,2§_ .07 .ll .16 jjyfli .03 .13 .11 .22 .12 .OO .01 :912_ .01 .19 .02 .08 .02 .14 .06 8. Two Dimension Solutions P +0 5.3.: .20 .24 .02 _.___§_9_= .02 .23 31:0 .11 =3; .21 .12 .23 _.__2_8_ .21 £111! .11 .15 £9. .15 .13 ___3__4: ;_2_8_ 3L1}. .12 i0 _-__._3_6_ .04 .07 .03 £1 1% .19 .10 g .05 _28 .04 g -0 P .09 lg§_ .15 .ll .12 ,gg_ .15 C. Three Dimension Solutions 3110 .21 _2_5_ .23 _._:_3_O_ _._=5___4 .14 .23 {2:51 .06 .18 .24- igg; .21 .14 .24 P +X .10 .08 20 _62 .11 .21 .20 P/-X .28 O9 .20 .13 .14 .18 .20 Relation 66666666 :6 6666 16666 .22 .23 Table 13 (cont'd.) C. Three Dimension Solutions .24 .14 .00 .04 .13 .18 .12 .14 .21 .22 .14 Study II (399) A. One Dimension Solutions .02 .10 .11 .11 .01 .14 .04 .10 .00 .02 010 .06 .20. .08 .11 .04 6.2.1. .04 . Two Dimension Solutions .15 .13 .14 _._,_2_§_ .11 .19 23 21 .08 .07 .15 .16 .27 .12 .04 .20 .17 .17 .19 .15 47 .21 48 Table 13 (cont'd.) Relation C. Three Dimension Solutions P +0 .05 .27 .08 .18 .16 .14 'égg: .14 E£;Q_ ,21_ .25 .18 ' .23_ .13 1‘;22| .19 .16 Eflfl .20 .07 .16 ._2_8_ .13 .17 _iL :38 22-.0. .29. .24 .26 .31. ._22 .13 .18 .22 :Q[£_ .18 .28 .16 .17 .12 .05 .14 .17 :91): _._2_1_ .16 .22. ._2_4 .422 _._3; _._2_8 .11 1911 .15 _2_9_ .13 _._2_5_ .06 '._2_€5_ .20 ,2_5_ :OLP_ _._2_1_ .20 .17 =3; 33 .13 .14 .12 aCorrelations underlined with a single line are significantly different from zero at the .05 level; those with two lines at the .01 level, each test. A second possible approach to relating the affective measures and the rank orders involves treating the orders as the independent or classification variable and the affective measures as the dependent variable in the analysis of variance. In the present data, however, there is a problem that is revealed in Table 14, which presents the distribution of the number of observations in each class if the classes are composed of one order each. As can be seen from the table, a large majority of orders are represented only once in the data. The problem of performing an analysis of variance on such data is obvious, given the number of cells and the lack of within cell variance. The possi- bility of grouping the orders together exists, but there is a very large number of possible groupings, each of which could produce different results, and any selection would be, of necessity, arbitrary. 49 Table 14 Number of Orders Versus Number of Observations per Order Observations per Order Study I (391) Study 11 (Egg) Pleasantness Stability Pleasantness Stability 1 122 154 224 255 2 17 13 43 26 3 8 7 11 8 4 5 6 2 4 5 2 O O 6 O O 0 l 7 2 O O l 8 2 O O O 9 O O 1 0 Total 158 183 281 295 50 A third possible approach to determining the existence of any relationship between the affective measures and the orders is the calcu- lation of correlations between the ranks assigned to each situation and the affective measures. While this does not directly involve the orders, any relationship involving the ranks assigned to a situation must be reflected in the orders. Table 15 presents the correlations between the affective measures and the ranks assigned to each of the situations for both stability and pleasantness rankings for both studies. Correlations significantly different from zero at the 1 percent level are underlined for convenience, with the level of significance chosen because of the problem of multiple tests. Inspection of correlations for Study I reveals that six of the 96 correlations are significant at the .01 level, but that the largest of these is .21. Additionally, none of these appear for ;Q[£_bonds, which are of primary interest in Hypothesis II. Given the size and small number of these coefficients, further interpretation of them seems inappropriate and will not be attempted here. For Study 11, 21 of 128 coefficients are significant at the 1 percent level, as a correlation of .13 is sufficiently large with 360 observations to reach significance. Given the dependency between ranks, the correlations between the ranks assigned to two different situations and the same affect variable are not independent and should be of opposite sign. Thus the results in the second half of Table 15 will be discussed according to the type of affective relationship involved. First, the most relevant to Hypothesis II, :Q[£_bonds are significantly related to the pleasantness ranks assigned to four situations, 8, D, F, and G such that the more that -0 likes P, the more pleasant F and G are 51 Table 15 Correlations between Affect Measures and Ranks Assigned to the Situationsa Situation: A B C D E F G H Affect Study 1 (£95) Relation A. Pleasantness £119_ -.02 .14 .03 .02 .ll .07 .Ol .09 P/-O .05 ;,§1_ -.04 .04 .02 .13 .01 -.02 P +x -.Ol .14 -.05 -.1O .01 .02 .04 -.01 P -X -.02 :_._2_O __,1_6_ .03 -.06 -.O9 .14 -.O4 jiyfl: -.08 .08 -.O8 -.01 -.O4 -.02 .06 .07 :912_ .06 -.O7 .03 .02 .02 -.Ol -.04 .00 8. Stability E[:Q_ .09 .12 .04 .09 -.13 ;;29_ -.O4 .04 E[;Q_ —.Ol -.11 -.08 -.02 .ll -.03 .13 .02 £111 .09 _._18 -.O4 -.O3 -.04 -.14 .04 -.03 EL;£_ .04 - 11 .05 -.04 .Ol .03 .08 -.05 :QLE_ .00 .OO .04 .17 -.O8 -.10 -.05 -.Ol :QLE_ -.12 -.08 .04 .05 .01 .01 .04 .02 Study 11 (£99) A. Pleasantness P£:9_ «.02 .01 .OO -.05 .02 .08 .07 —.11 EL;Q .02 ':313_ -.O9 -.10 .02 '.20 .08 .OO P/+Q -.03 -.11 .05 -.ll -.02 .20 .11 -.O9 Table 15 (cont'd.) A. Pleasantness _a 01 d d O _.: 0" 33g -.05 _._2_1_ .06 .01 —.07 .04 .11 +0 P -.05 —.07 .05 .01 -.03 .04 .06 -O/P «.07 _—_,_1_3_ —.02 ;_._1_7_ -.1o _._2; _._13 191]: -.04 -.11 .03 -.06 _.07 .09 __._1_5_ M -.04 ;_._2_2_ .01 .oo -.07 .05 _._16 B. Stability 11:9 _,_1_3_ .01 .04 -.02 .04 .02 -.04 _P_[_;Q -.02 .03 .02 214 .02 _._2; -.01 £130 __._1_g_ -.03 .05 -.07 .oo .11 -.05 32:9 .03 000 .01 -.02 -.07 .02 .11 :02}: .oz .05 .03 .oo -.09 .08 -.07 :0le .02 .00 -.1o -.22 .02 .27 -.02 181.2 .14 -.08 -.08 .01 .02 .12 -.05 _-_Q[_E .10 -.04 -.08 -.02 -.1 .09 .06 aUnderlined correlations are significantly different from zero, p (.01 I O 0‘ -.02 .04 -.05 .06 53 ranked and the less pleasant B and D are ranked. For :919,similar relationships are observed for situations B and G. In addition, ranks assigned to situations are related to bonds such that the more disliked :9 the more pleasant is situation 8 and the less pleasant situation F; the more ;9_is disliked, the more pleasant situation B; the more :9_is liked, the more pleasant situation F; and the more :9_likes 9, the more pleasant situation G. These relationships, particularly those for :919_ and 3919, appear to support Hypothesis II with regard to the pleasant- ness rankings in Study II. For the stability rankings for Study II, ;9[9_is related to situations 0 and F in the same manner as in the pleasantness data, such that the more ;9_likes 9, the more stable the rank of situation F and the less stable that of situation 0. For:9£9_the only significant relation is that the more ;9_1ikes 9_the less stable the ranking of situation E. In addition, the more :9_and :9_are liked, the more stable the ranking of situation A and the less stable situation H; the more disliked :9, the more stable situation 0 and the less stable situation F; and the more :9_likes 9_the more stable situation A is ranked. The results for the stability rankings for Study 11 also appear to support Hypothesis II, even though the relationships are not strong. 999139. Thus, the hypothesis that some relationships exist between the affective measures and the rank orders assigned to situations seems to be supported, although none of the relationships were very strong. The results of the multidimensional scaling analysis could conceivably be either more or less strong than those which would have been observed had the process not been so dominated by the relatively few observations not following the general pattern of low ranks assigned 54 to situations A and B. The correlations using actual ranks assigned to each situation, although not directly addressing the problem of rank orders, demonstrate that the ranks are related to the affective measures, thereby implying some differences between orders as a function of these measures. The failure of the approach based on using the orders as the levels of an independent variable in the analysis of variance, while not providing any information with regard to the issue of Hypothesis II, indicates that there is a great deal of variability in the assigning of ranks to the situations, a variability not even substantially accounted for by the affective measures. Tests of the Models Hypothesis III relates to the relative fit of three models, the Cartwright and Harary, the Newcomb, and the present cluster revisions of Heider's balance theory. One approach to making these comparisons is to count the number of observations where the rank order is non inconsistent with the predictions of the model. For the cluster model, this approach requires that each observation be assigned to one of the appropriate submodels on the basis of the ;9[9_and ;9[9_bonds. For the present analysis, these bonds were defined as reciprocal if ratings between 9 and 15 on the 15 point affect scale were obtained, and as nonreciprocal if the rating was between 1 and 8, this being an approxi- mate median split for both the ;9£9_and the ;9£9_bonds. Table 16 pre- sents the number and proportion of observations fitting the requirements of each model, including the 5 percent confidence limit on the pr0portion observed. In addition, the table includes the proportion of observations which would be expected to fit each model under an assumption of equal Model Cartwright & Harary Newcomb Clustera Cartwright G Harary Newcomb Clustera Table 16 Proportions of Observations Fitting Each Model Observed Proportion (Confidence Range) Pleasantness Stability Study I (995) .0417 (.Ol7-.O78) .1625 (.112-.218) .1458 (.099-.202) .2333 (.152-.268) .0250 (.008-.058) .0458 (.021-.084) Study II'(999) .0750 (.046-.113) .1670 (.126-.214) .1389 (.100-.185) .1611 (.121-.210) .0250 (.009-.052) .0278 (.012-.056) 55 Expected Proportion .0143 .0024 .0005 .0143 .0024 .0003 aCluster model based on a division of -0/P and - /P between responses 8 and 9 on the affect scale, the approximate m both distributions. 1an of 56 probability for all rank orders. For the cluster model, the expected proportion depends on the number of observations in the two or four reciprocity in the two studies respectively. Examination of Table 16 shows that the best fit of any model on an individual observation basis is about 23 percent, obtained by the Newcomb model on the stability data of Study 1, but that, even given the small number of observations fitting each model, significantly more observations fit each model than predicted on the basis of equally likely rank orders. Comparing the degree of fit of the models, that is, determining which demonstrates the better predictability, becomes somewhat difficult with these data. The models differ with respect to the restrictiveness that they impose on an order in determining whether it fits the model. For example, the Cartwright and Harary model allows six degrees of free— dom within which it may be fit, three among the balance situations and three among the imbalanced. Similarly the Newcomb version allows 5 degrees and the cluster model allows either 1, 2, or 4 degrees of free- dom depending on which of the submodels is appropriate. Thus any com- parison of the models must take into account both the differences in the restrictiveness of the models and the fact that the models are not independent of each other. One approach which might permit such comparisons involves the use of Bayesian probability analysis to determine the probability of the truth of the model given the data (see Mendelsohn, 1970). This approach involves determining for each order the probability of the order if the model were true, dividing by the probability (relative frequency) of the order in the data, and taking the product of all such quantities together with some a priori estimate of the model's truth. In the present 57 context, some of the orders observed have a probability of zero for all three models, with the result that the produce becomes zero for all three models, thus not differentiating between the models. An alterna- tive approach is to consider an order occurring below a certain frequency as unusual (and probably representing error), and to calculate the prob- abilities based on the remainder of the orders. In the present instance, as demonstrated by Table 14, this alternative would result in not using most of the data, since most observations are represented by orders which occurred only once. Thus the Bayesian analysis approach does not prove useful in differentiating between the models. Another possible approach to determining the relative fit of the models involves the multidimensional scaling analysis discussed above. The space resulting from the unfolding should theoretically be parti- tionable into areas fitting each of the models. Then the percentage of the total space fitting each model could be used as the expected propor- tion to test against the number of observations in the area in a chi- square test, and the resulting chi-squares could be compared by means of a test such as the F ratio. However, as noted above, the multidimensional scaling analysis placed inordinate weight on the unusual cases as far as the rankings of situations A and B were concerned. This resulted in the location of most cases near the origin to such an extent that points representing differing orders were not distinguishable. Thus it becomes impossible to determine the division between an area fitting a model and those not fitting the model, and the comparison cannot be made. A final approach to comparing the models involves calculating, for each observation, the rank correlation between the observed order and the orders implied by the models. By virtue of having a correlation 58 value for each model on every observation, the models can be compared by use of a t_for correlated means. Table 17 presents the average rank correlation for each of the models in each study. For the cluster model, the ;9[9_and ;9[9_bonds were classified on the same basis as in previous analyses. Inspection of Table 17 reveals that in every case, the average correlation for the Newcomb revision was greater than for either of the other models. Compared by a t_for correlated means, the average correla- tions for both the cluster and the Newcomb models significantly exceeded the average for the Cartwright and Harary version in every case (t_= 5.14 to 17.52, p (.001). In addition, the Newcomb model was more highly correlated with the observations than the cluster model for the stability rankings for both studies (t_= 2.72 and 3.55 for the two studies respec- tively. p (.005) but not for the pleasantness rankings (t_= 1.14 and 0.99, respectively, p;>.20). The differences between the Newcomb model and the proposed cluster model in predicting the observations, although significant, do not appear to be very great. Given the differences in the restrictiveness of the models, it is questionable whether the dif- ferences in correlations are sufficient to warrant a preference for the Newcomb model over the cluster model; thus Hypothesis III is neither strongly contradicted nor is it supported. A question might be raised regarding the utility of not maintain- ing the reciprocity assumption in the cluster model. Table 18 presents the average rank correlation for each of the cluster submodels in each of the reciprocity conditions of each study. For Study I, it must be observed that the nonreciprocity model is more highly correlated with the observed orders than the reciprocity model in both classes. Thus it seems that in this study, the nonreciprocity model is a better predictor even Model Cartwright & Harary Newcomb Cluster Cartwright & Harary Newcomb Cluster Table 17 Average Rank Correlations Between Observations and Models Pleasantness .153 .514 .492 .321 .511 .494 Stability Study I (99);) .278 .554 .495 Study II (999) .278 .555 .508 59 60 Table 18 Average Correlations Between Observations and Cluster Submodels by Reciprocity Conditions in Dataa Study I’(99§) Model: R N Data A. Pleasantness R 9991_ .622 N .292 £999_ B. Stability R .215. .550 N .364 .520 Study II (999) Model: RR RN NR NN A. Pleasantness RR _._59; .375 .505 .532 RN .428 ' ' _._3_7_§ .470 .428 NR .414 .339 ._599 .559 NN .323 .282 .466 .419 B. Stability RR 9§§1_ .433 .438 .496 RN .564 95§§_ .343 .414 NR .498 .357 9595_ .557 RR .497 .342 .519 gggg; aUnderlined entries denote instances where data is appropriate to the specified model. R indicates reciprocity; NR, non reciprocity. 61 for data in which reciprocity is observed. For Study II, the data reveal no such clear trend, although it appears that the model for reciprocity—nonreciprocity is the least strong of the four. It can also be seen, for instance, that the appropriate model is never the most predictive for both the stability and pleasantness data, and that the differences between the three remaining models are neither as consistent or large as that observed in Study 1. Thus the results appear to indicate that the ;9[9_and ;9[9_relationships did not play the significant role in the fit of the models to the data that would be expected given the use of the relationships in the development of the models. DISCUSSION The hypotheses put forth at an earlier point in this paper propose the existence of a set of conditions and of certain relationships between that existence and predictions associated with various models. The results presented here pose a problem in that, even though the pre- conditions were demonstrated to exist, the association between those conditions and the hypothesized effects was not consistently observed. Thus the discussion of this research must focus on these inconsistencies. The occurrence of the hypothesized preconditions, however, merits inte- gration with the previously available psychological lore. Affective Relations It has been argued that the existence of reciprocity in perceived interpersonal relationships is both a safe assumption and a necessary condition for the conduct of balance research. While the necessity issue will be discussed at a later point, the safety of an assumption of reciprocity can, and should be discarded immediately. As noted above, the often cited evidence for this assumption, on closer examination becomes open to interpretation and seems quite limited regarding percep- tions of relationships involving the perceiver. The most direct evidence in regard to this class of relationships in fact contradicts the assump- tion in the case of affect from a disliked other (Price, Harburg, & Newcomb, 1966). The results of the present two studies constitute a further contradiction of the assumption of reciprocity, again in the case of perceived affect from a disliked other. In both studies, a substantial number of subjects reported the belief that at least one or two or three 62 63 disliked others liked them. The remaining results relative to Hypothesis I may permit some insight into the nature of the perception of affect from a disliked other. Although the predominant variance in these perceptions was not between subjects and it cannot thus be said that the perception of being liked by a disliked other is a consistent and strong difference between individuals, the data at hand are most consistent with the notion that some individuals possess more of a tendency to have such perceptions than do others. Further clarification of this lack of independence in such perceptions is a problem for further research to determine whether it represents a consistent behavioral or perceptual trait and to uncover the nature and origins of such a trait. As noted earlier, the modal behavior in our society is to avoid those whom we dislike and to otherwise interact with them in a stereo- typical friendly manner. This relative decrease in interaction and in the meaningfulness of whatever interaction occurs would logically lead to a decrease in the degree of certainty or reliability of perceptions about the affective responses of these disliked others. Such an un- reliability is perhaps reflected in the larger variance associated with such perceptions in the present research. Finally, some psychologists with an investment in the reciprocity assumption might be inclined to view the pattern of results for affective relations as the product of a statistical artifact. That is, the subjects selected those toward whom they had extreme affect and other variables associated with these extremes would be expected to regress toward the mean of the affect scale. Yet the relatively larger difference observed for one end of the affect distribution demands more than this statistical 64 artifact explanation. If one Eg:g_to accept the existence of some regression artifact as a contributory phenomenon and then somehow transformed the data to remove its effects, the differences between —L to S and‘S to —L ratings would be smaller but still of significant magnitude. The artifact hypothesis is not a sufficient explanation for the lack of reciprocity observed in the data of the present research. The most consistant explanation for the present data is that reciprocity, as it relates to perceived affect, is neither the universal nor overwhelmingly predominant phenomenon claimed by balance researchers. Correspondingly, this conclusion dictates certain restrictions on research using interpersonal affective relationships, restrictions counter to assumptions heretofore accepted. Ranks and Affective Relations The mean and median pleasantness ranks assigned to each situation present a pattern somewhat outside the relatively broad range reported in the balance literature. This result is not inconsistent with the Crano and Cooper (1973) contention that noncomparative methods, such as the often used rating scale, may be inappropriate for use with variables whose scaling characteristics are unknown. In their research, Crano and Cooper demonstrated the nonscalibility of pleasantness comparisons, probably due to a lack of transitivity in the judgements of pairs of situations. Although the present method does not allow the researcher to discover whether intransitivity is a normal characteristic of a particular variable, it has the advantage of preventing its occurance in the data by forcing the simultaneous comparison of all stimuli. Thus differences between the present results and previous findings may be 65 solely reflective of the poor scaling characteristics of the pleasant- ness construct. The fact that the stability results are more consistent with previous research supports this notion of an interaction between method and the scaling properties of the variables used in this research. Thus differences between the present results and those of previous studies need not be attributed to other, extraneous variables. The multidimensional scaling approach to finding possible rela- tionships between the affective variables and the rank orders proved somewhat disappointing, although the use of such procedures always involves the risk of uninterpretable output. The one notable result from the procedure appears to be that, where significant relationships were consistently observed, however weak, these involved bonds to or from disliked others as indicated in the hypotheses. The other attempts to observe relationships between rank orders and the affective bonds deserve similar comment. The large variety of orders observed in the data precluded any comparison of means for dif- ferent classes of observations and at the same time seems indicative of a potential difficulty for balance research which will be discussed when the various models are considered. The correlations between the ranks and the affective bonds are again inconclusive, in the one study due to the lack of significance and in the other to the pattern of the small correlations observed. It should be noted that the significant correla- tions in Table 15 for :9[9_and :919 bonds are in directions not incon- sistent with the model proposed in this paper. Thus it appears that some relationships, however weak, do exist between the ranks and the affective bonds. Yet the meaning and theore- tical significance of these relationships cannot be assessed through the 66 present data, due greatly to the relative size and pattern of the rela- tionships and the failure of the multidimensional analysis procedure to produce unequivocal results. The resolution of this question must await research using some other, as yet unrevealed, methodology. The Models The present research set out to compare several models derived from Heider's cognitive balance theory. The results presented above indicate a great diversity in the preferences of individuals among the basic triadic situations so central to this area of research. This constitutes a severe problem for any model attempting to predict these individual preferences, and yet this must be the goal of any psycho- logical theory or model, as opposed to the description of the average behavior, especially in the light of such diversity. A model which would attain any degree of success in predicting the present data would necessarily have to predict differences in the preferences of individ- uals; yet at the same time the criterion of accuracy of the model must be maintained. The cluster model, by the inclusion of the reciprocity variable, was exactly such an attempt, and it must be viewed as a failure given the results presented above. Yet neither is there any support in the present data for either of the other two models considered in this research. The Cartwright and Harary model, which allows but does not predict differences in individ- uals, was shown to be of such small predictive accuracy as to deserve no further consideration here. The Newcomb model, while giving a better account of itself, cannot be said to have been successful in terms of the present data, fitting only about one—fourth of the individual 67 observations, or accounting for an average of only about one-fourth of the variance over all observations, even though the group means for stability were consistent with the model‘s predictions. The failure of the cluster submodels to display greater accuracy in the appropriate reciprocity conditions than in the others seems to be a further problem for this approach. Together with the overall poor performance of the model, this result constitutes a failure of the pre- sent research to demonstrate any practical utility to be gained from the inclusion of the reciprocity variable in this area of research, regard- less of the finding that reciprocity does not hold in the case of affect from a disliked other. Nevertheless, one is confronted with the existence of some relationship, however small and confusing, between the affective variables and the preferences assigned to the triadic situations. And, the task of developing some model capable of predicting those preferences still remains unaccomplished. If the area of cognitive balance is to remain a viable part of the science of social psychology, these two facts are deserving of further psychological effort, both theoretical and research. On Necessity Finally, some commentary on the necessity of an assumption of reciprocity in the conduct of balance research seems in order. Regardless of the area of research, there always exist some methods which seem to be of extreme utility, but which also require assumptions or restrictions which may not be otherwise justified. In these instances it is incumbent upon the researcher to choose between the loss of a valuable method or the possible loss in the generalizability in the results he obtains. If the 68 assumptions or restrictions can be made sufficiently explicit and are included in the expression of the findings of the research, together with notice of the limitations thus placed on the generality of the findings, then the method need not be lost, and the research should proceed. In the present caSe, methods which require the existence of reciprocity in perceptions of interpersonal relationships are apparently only restricted in the case of perceptions involving affect from a dis- liked other toward the subject and situations which might theoretically involve such perceptions. Even in these instances, the researcher has at his disposal the means to observe whether any subject or situation falls within the portion of cases excluded by this restriction. On the whole, the loss in any study may amount to only a small portion of the situations of interest to a particular researcher. Whether the corres— ponding loss of generality is important becomes an additional matter for research effort. LIST OF REFERENCES LIST OF REFERENCES Abelson, R. P.. Aronson, E., McGuire, W. J., Newcomb, T.M., Rosenberg, M. J.. and Tannenbaum, P. H. (Eds.) Theories of Cognitive Con- sistency: A Sourcebook. Chicago: Rand McNally, 1968. Backman, C., and Secord, P. The effect of perceived liking on inter- personal attraction. Human Relations, 19593.12! 379—384. Blumberg, H. On being liked more than you like. Journal of Personality and Social P§ychology, 1969, 11, 121-128. Burnstein, E. Sources of cognitive bias in the representation of simple social structures: Balance, minimal change, positivity, reci- procity, and the respondent's own attitude. Journal of Personality and Social Psyghology, 1967, 1, 36-48. Cartwright, 0., and Harary, F. Structural balance: A generalization of Heider's theory. Psychological Review, 1956, 99, 227-293. Coombs, C. H., Dawes, R. M., and Tversky, A. Mathematical Psychology: An Elementary Introduction. New Jersey: Prentice-Hall, 1970. Crano, W. 0., and Cooper, R. E. Examination of Newcomb's extension of structural balance theory. Journal of Personality and Social Psychology, 1973, 91, 344-353. Davis. J. A. Clustering and structural balance in graphs. flgmgg_ ‘Relations, 1967, 20, 181—187. Davol, S. H. An empirical test of structural balance in sociometric balance in triads.’ Journal of Abnormal and Social Psychology, 69 70 1959, 99, 393-398. DeSoto, C. B., and Kuethe, J. L. Perception of mathematical properties of interpersonal relationships. Perceptual and Motor Skills, 1958, 9, 279-286. DeSoto, C. B., and Kuethe, J. L. Subjective probabilities of inter- personal relationships. gpurnal of Abnormal and Social Ppychology, 1959, 99, 290-294. Diem, K., and Lentner, C. (Eds.) Scientific Tables. Basle, Switzerland: Ciba-Geigy. 1970. Gutman, G. M., and Knox, R. E. Balance, agreement, and attraction in pleasantness, tension, and consistency ratings of hypothetical social situations. gpurnal of Personality and Social Psychology, 1972, 99, 351-357. . Harari, H. An experimental evaluation of Heider's balance theory with respect to situational and predispositional variables. Journal of Social Psyghology, 1967, Z9, 177-189. Heider, F. Social perception and phenomenal causality. Psychological Review, 1944, 51, 358-374. Heider, F. Attitudes and cognitive organization. Journal of Psychology, 1946, 91, 107-112. Heider, F. The Psychology of Interpersonal Relations. New York: Wiley, 1958. Heider, F. The Gestalt theory of motivation. In M. R. Jones (Ed.), I99_ Nebraska Symposium 9n Motivation, 1960, 9, 145-172. Heider, F., and Simmel, M. An experimental study of apparent behavior. American Journal of Psycholpgy, 1944, 57, 243-287. 71 Jordan, N. Behavioral forces that are a function of attitudes and cognitive organization. Human Relations, 1953, 9, 273,287. Jordan, N. Cognitive balance as an aspect of Heider's cognitive psychology. In R. P. Abelson, et. a1. (Eds.), Theories of Cognitive Consistenpy: A Sourcebook. Chicago: Rand McNally, 1968. Kogan, H., and Tagiuri, R. Interpersonal preference and cognitive organization. Journal of Abnormal and Social Psychology, 1958 99, 113—116. Kruskal, J. B. How to use MDSCAL, a program to do multidimensional scaling and multidimensional unfolding. Unpublished report, Bell Telephone Laboratories, 1968. McNemar, A. Psychological Statistics. (3rd ed.) New York: Wiley, 1962. Mendelsohn, G. H. A sequential Bayesian model evaluation technique applied to a model of concept identification. Unpublished master's thesis, Michigan State University, 1970. Morrissette, J. 0. An experimental study of the theory of structural balance. Human Relations, 1958, 11, 189-195. Newcomb, T. M. Interpersonal balance. In R. P. Abelson, et. al. (Eds.), Theories of Cognitive Consistency: A Sourcebook. Chicago: Rand McNally, 1968. Peters, C., and VanVoorhis, W. Statistical Procedures and Their Mathe- matical Bases. New York: McGraw-Hill, 1940. Phillips, J. L. Notes on cognitive social psychology. Mimeographed lecture notes, Michigan State University, 1971. 72 Price, K. 0., Harburg, E., and Newcomb, T. M. Psychological balance in situations of negative interpersonal attitudes. Journal of Personality and Social P§ychology, 1966, 9, 265-270. Rodrigues, A. Effects of balance, positivity and agreement in balanced and imbalanced triads. Journal of Personality and Social ngchology, 1967, _5_, 472-476. Rodrigues, A. The biasing effect of agreement in balanced and imbalanced triads. Journal of Personality, 1968, 99, 138-153. Taguiri, R., Blake, R. R., and Bruner, J. S. Some determinants of the perception of positive and negative feelings in others. Journal of Abnormal and Social Psycholpgy, 1953, 99, 585-592. Taylor, H. F. Balance in Small Group§. New York: Van Nostrand Reinhold, 1970. Taylor, H. F. Balance, tension and tension release in the two-person group. Human Relations, 1968, 99, 59-74. Weist, W. A quantitative extension of Heider's theory of cognitive balance applied to interpersonal perception and self-esteem. Psychological Monographs, 1965, Z9, Whole No. 607, 1-20. Whitney, R. Agreement and positivity in pleasantness ratings of balanced and unbalanced social situations. Journal of Personality and Social Psychology, 1971, 19, 11-14. Zajonc, R. 8. Cognitive theories in social psychology. In G. Lindzey and E. Aronson (Eds.) H9n99gok of Social Psychology, (2nd ed.) Vol. 1. Reading, Mass.: Addison-Wesley, 1968. 73 Zajonc, R. B., and Burnstein, E. Structural balance, reciprocity, and positivity as sources of cognitive bias. Journal of Personality, 1965, 99, 570-583. Zajonc, R. B., and Sherman, S. J. Structural balance and the induction of relations. Journal of Personality, 1967, 99, 635-650. “mfllh'- 2" K's..." 'l u D APPENDICES APPENDIX A INSTRUCTION SHEET FOR STUDY I 74 GROUPS EXPERIMENT The instructions for this experiment are probably relatively complicated. Please read them carefully and ask if you have any questions or doubts. Please take your time; everyone will be working for about 50 minutes on the experiment. This experiment is a study of the stability and pleasantness of various groups of people and their feelings about issues. Your primary task in the experiment will be to judge the pleasantness and stability of situations involving yourself, another person, and an issue. To assist you in doing this, we have defined these as follows: STABLE--A stable situation is one where the feelings of the members of the group are very likely to stay the same. An unstable situation is one where some sort of change seems likely to occur in the feelings of some member of the group. PLEASANT--A pleasant situation is an enjoyable one, one in which you would feel comfortable. An unpleasant situation is one where you would feel uptight and uncomfortable. The study involves relationships between yourself and people whom you know per- sonally, that is, people who also know you. We would like you to think of two persons of the same sex as yourself and whom you like very much. These should be people of about the same age as yourself, for example students at Michigan I State or from your high school class. Place their initials on the following lines (you may keep this sheet after the experiment is over): A. BI Now think of two people whom you know personally, who are of the same sex and about the same age as yourself, but whom you dislike very much. (This may seem harder at first, but most people can think of two people they dislike pretty much, even though they do not wish them harm. For example, think of people you sort of try to avoid as much as possible and really don't like to be around very much.) When you think of two people you dislike very much, write their initials by the letters below. D. E. Now think of some issue about which you feel very strongly. Some examples might be abortion, marijuana reform, financial aid to parochial schools, busing or amnesty. It does not matter yet whether you are for or against the issue, just that you feel very strongly about it. Write the name of the issue in the blank below Now think of a short label for the side of the issue you are in favor of and an- other short label for the side you are against. Write these labels on the appro- priate lines below: (I am in favor of) C. . . (I am against) F. . Throughout the experiment you will find blanks marked with the letters A through F. Whenever you find these blanks you should write in the initials or label from the correspondingly lettered line on this sheet. Be sure to note that the letters do not appear in order on this sheet, but go A, B, D, E, C, P. APPENDIX B SAMPLE SITUATION CARD 75 W25 I like gB) . I dislike (F) . (B) . strongly likes gF) . 1 PLEASANT UNPLEASANT APPENDIX C ANSWER SHEET FOR STUDY I Rank 10 ll 12 13 14 15 16 Code Number ANSWER SHEET (Pleasantness) Rating I did pleasantness ranking: 76 Circle: Female Male second 77 Circle: Male Female ANSWER SHEET (Stability) Rank Code Number Rating 1 2 10 11 12 13 14 15 16 I did stability ranking: first second APPENDIX D AFFECT QUESTIONNAIRE FOR STUDY I 78 AFFECT QUESTIONNAIRE Now we would like you to rate for us just how much you like each of the other peo- ple in these situations and how they feel about you. We would also like for you to rate how strongly you feel toward each side of the issue that you have used in the experiment. The scales below are very similar to those that you have used on the situations. Just fill in the initials in the blanks and place a mark in the space best indicating how you feel about each person and issue or how they feel toward you. My feelings toward (A) : like: : : : : : : : : : : : : : : :dislike very much very much My feelings toward (B) : like: : : : : : : : : : : : : : : :dislike very much very much like: : : : : : : : : : : : : : : :dislike very much very much My feelings toward (D) : like: : : : : : : : : : : : : : : :dislike very much very much My feelings toward SE) : like: : : : : : : : : : : : : : : :dislike very much very much My feelings toward (F2 : like: : : : : : : : : : : : : z : :dislike very much very much (A) feelings toward me: like: : : : : : : : : : : : : : : :dislike very much very much (B) feelings toward me: like: : : : : : : : : : : : : : : :dislike very much very much (D) feelings toward me: like: : : : : : : : : : : : : : : :dislike very much very much (E) feelings toward me: like: : : : : : : : z : : : : : : :dislike very much very much Relative to other experiments that I have been in or have heard about this one was: boring: : : : : : : : : : : : : : : :interesting difficult: : : : : : : : : : : : : : : :easy APPENDIX E INSTRUCTION SHEET FOR STUDY II 79 GROUPS EXPERIMENT The instructions for this experiment are probably relatively complicated. Please read them carefully and ask if you have any questions or doubts. Please take your time; everyone will be working for about Il’minutes on the experiment. This experiment is a study of the stability and pleasantness of various groups of people who you know. Your primary task in the experiment will be to judge the pleasantness and stability of situations involving yourself and two other people. To assist you in doing this, we have defined these as follows: STABLE-~A stable situation is one where the feelings of the members of the group are very likely to stay the same. An unstable situation is one where some sort of change seems likely to occur in the feelings of some member of the group. PLEASANT--A pleasant situation is an enjoyable one, one in which you would feel comfortable. An unpleasant situation is one where you would feel uptight and uncomfortable. The study involves relationships between yourself and peOple whom you know personally, that is, people who also know you. We would like you to think of three persons of the same sex as yourself and who you like very much. These should be people of about the same age as yourself, for example students at Michigan State or from your high school class. Place their initials on the following lines. (you may keep this sheet after the experiment is over, so if writing the names out will help, please do so) A. B. C. Now think of three people who you know personally, who are of the same sex and about the same age as yourself, but whom you dislike very much. (This may seem harder at first, but most people can think of three people they dislike pretty much, even though they do not wish them harm. For example, think of people you sort of try to avoid as much as possible and really don't like to be around very much). When you think of three people you dislike very much, write their initials by the letters below. ‘ Throughout the experiment you will find blanks marked with the letters A through F. Whenever you find these blanks you should write in the initials from the correspondingly lettered line on this sheet. APPENDIX F ANSWER SHEET FOR STUDY II ANSWER SHEET Stability Rank Code Number Rank 1 13 2 11+ 3 15 4 16 5 17 6 18 7 19 8 20 9 21 1O 22 11 23 12 2A I did stability ranking: first second Circle: Female Male Code Number 80 ANSWER SHEET Pleasantness Rank Code Number Rank Code Number .3 .\ \IJ 1A 15 16 17 18 19 \ooox‘soxmrum 21 22 _s O _I _3 23 12 2A I did pleasantness ranking: first second Circle: Male Female APPENDIX G AFFECT QUESTIONNAIRE FOR STUDY II 82 AFFECT QUESTIONNAIRE Now we would like you to rate for us just how much you like each of the other people in these situations and how they feel about you. Just place a check mark in the space best indicating the feeling involved. My feelings toward (A), : like: : : : : : : : : : : : : : : : dislike very much 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 very mucb My feelings toward (B) like: : : : : ‘ : : : : : : : : : : : dislike very much 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 very much My feelings toward (C) like: : : : : : : : : : : : : : dislike very much 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 very mUCh My feelings toward (D) like: : : : : : : : : : - : : : : : : dislike Very much 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 very mUCh My feelings toward (E) like: : : : : : : : : : : : : : : : dislike very much 1 2 3 a 5 6 7 8 9 10 11 12 13 14 15 very much My feelings toward (E), like: : : r___: - ‘-_”3___5_~_5---3 ‘_11F._l} : : dislike very much 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 very much continued 83 (A) 's feelings toward me: like:__~, : : : : : : : : : : : : : :dislike very much 1 2 3 E 5 T 7 8 9 1o 11 12 13 1E 15 very much (B) 's feelings toward me: like: :___} : : : : : : : : : : : : :dislike very much 1 2 3 H 5 6 7 3 9 10 11 12 13 1E 15 very much (C) 's feelings toward me: like:___ : - : : :_ :dislike (D) 's feelings toward me: like :_:—:___: :_: :___: :___: : : :_: :___: dislike verymuch 1 2 3 T 5 T 7 '8’ 9 1o 11 12 13 173—15 verymuch (E) 's feelings toward me: like: : : : : : : : : : : : : : : :dislike verymuch-T-‘TTT?T7TT1O 11 12 WWW-5.verymuch (F) 's feelings toward me: like: : : : : :dislike verymuoh—‘F-T— 3 “T ”“6- MTTfia—W'fi—fi-Tfi verymuch Relative to other experiments that I have been in or have heard about this one was: boring: : : : : : : : : : :interesting 1 2 TT'B’TTTTWTFWTE'B'TTTS difficult: : : : : : : : : : : : : : :easy