Re-evaluating Newcomb's Position
In this section, I will appraise how Nevvcomb's assertions (please see p. 18) stand in the light of evidence:
a) As I have already reviewed in Chapter 1, there is evidence against Nevvcomb's "disengagement" assertion vvith regard to negative P /O. Yet, I took some precautions against the possible occurrence of disengagement. Disengagement is evidently not
» inevitable even vvith pleasantness ratings.
b) The results of this study shovv that balanced —P /O situa-tions receive ratings different from unbalanced —P /O situasitua-tions, both in terms of pleasantness and consistency. Contrary to Nevvcomb's contention, Heider's balance principles operate also when the P /O relation is negative — except in the opposite direction vvith
pleasantness.
c) The nonbalance phenomenon as described by Nevvcomb, vvas totally absent in the overvvhelming majority of the conditions of this study—except three of the story pairs in the Standard / pleas-antness, and one in the distraction / pleasantness conditions.
d) This study provides additional support to the grovving con-viction that nonbalance is a task-specific or a dependent-measure-specific phenomenon. I observed that, even though both consistency and pleasantness rating tasks did enable the subjects to differen-tiate, vvithin the nonbalance category, the Heiderian balance situ-ations from the Heiderian unbalanced situsitu-ations, pleasantness rat-ings are considerably less povverful and less effective in doing that job.
85
e) We recall that Newcomb had predicted that the two P-0-X configurations in which P /O is negative will resemble each other closely. I interpreted this to mean that they will not differ signifi-cantly from each other. In ali of the conditions, the balanced configurations received very high ratings of consistency, while the unbalanced configurations were rated very low in consistency. Even though the differences were smaller in terms of pleasantness ratings, nevertheless almost ali pleasantness results showed significant differ-ences between balanced and unbalanced configurations, ali within Newcomb's nonbalance category.
f) Newcomb's prediction that the two types of nonbalanced configurations willbe intermediate between positively balanced and positively unbalanced situations was not directly observable due to the design of the present study. However, the implicit ordering assumption within the two nonbalanced configurations could be checked againt the resülts here. Neither Heider's predicted order, nor Newcomb's expectation of a lack of systematic ordering among the nonbalanced configurations was confirmed. In fact, the observed direction was clearly and significantly opposite to what would be expected from a Heiderian point of view in the case of pleasantness ratings. Pleasantness ratings displayed an order here, in nonbalance cases, which is the opposite of that observed in consistency ratings. The second version of Newcomb's predic-tions (that nonbalanced configurapredic-tions tend to receive neutral rat-ings) fared better in this study since the subjects categorized both the balanced and the unbalanced configurations as quite unpleasant, or as not quite so pleasant. Nevertheless, this tendency toward neu-trality as such, was true for the unbalanced configurations only, which were located toward the middle of the unpleasantness scale.
The balanced triads, on the other hand, were rated quite near the unpleasant end of the scale. Thus we received partial confirmâtion for Newcomb's neutrality p rediction.
In sum; Newcomb's predictions are not supported, excepting the one about the intermediateple asantness ratings for nonbalanced configurations, vvhich received partial support. Heider's predic-tions are pretty well supported, if we overlook the inadequate dis-cretion of Heider's followers in choosing dependent measures or tasks to test the theory.
What Does Pleasantness Index?
The reader may be surprised that, in this study and in several • prior studies (Gerard and Fleischer (1967); Whitney (1971); Gutman and Knox (1972); Miller and Norman (1976); and Rodrigues (1968) vvhere P /O is negative, unalanced triads are rated as more pleasant than balanced triads5. Since in unbalanced triads with neg-ative P /O, P agrees vvith O, and since P disagrees vvith O in the balanced triads with negative P /O, it appears reasonable to conclude that subjects find disagreement unpleasant and rate it so even vvhen the triads are balanced. Consistency judgments, hovvever, seem responsive here only to balance, not to agreement.
Does pleasantness index agreement rather than balance? It seems reasonable to suppose that pleasantness indexes both. The evidence for this conclusion is that even the unbalanced (agreement) triads are rated as somevvhat unpleasant in this study.6
One might suppose that subjects find many things pleasant — balance, agreement, positivity of relations, positive attraction of P tovvard O, and probably numerous other factors. Beyond the the evident common sense of this supposition, vve can çite the rela-tively high variability of pleasantness judgments, including differ-ences due to specific stories, as contrasted to the consistency judg-ments, vvhich shovved relatively low variance and no differences among stories vvhich had the same balance configuration. (Appen-dix D, pp_ 115-122).
Accordingly, it seems appropriate that future researchers discontinue the use of pleasantness as a dependent variable in consistency theory research. Its multiple referents make it inappropriate to index the state of tension vvhich Heider assumed is felt given unbalance.
Purely cognitive measures, such as this study's consistency judgments, seem much better suited for investigations of balance
5 Some methodological problems with these studies were discussed in the review of the literatüre. Hovvever, the present study, vvhich hopefully has dealt adeqüately vvith the difficulties cited, shovvs the same result unambiguously.
6 In prior studies vvhich report a similar outcome, the possible addition of a unit rela-tion vvould more conclusively tie the higher pleasantness ratings to agreement alone, since there, ali the triads vvith a negative P /O could be considered unbalanced.
87
theory. Certainly the measure used in this study produced results very much as Heider's theory predicts.
However, a nagging unease is left in this investigator. Heider's theory assumes that unbalanced states lead to tension which in turn makes such states unstable. Asking subjects whether or not the triads are consistent or inconsistent does not, on the face of it, seem to index Heider's tension in a completely face-valid man-iler. Clearly unpleasantness cannot do the job, because it is too sentive to other variables.
Using pleasantness-unpleasantness as a dependent measure leads to confusing, variable results, and to an unfortunate theo-retical conflict between Heider and Newcomb. Yet some alternative is needed to index Heider's tension in an unambiguous fashion, capturing the motivational nature of this hypothesized disturbance without being responsive to irrelevant variables. We must leave this task to future theorists and researchers.7 It is of course also possible that perceived inconsistency is, in itself, not necessarily
"bothersome" or "tension producing" in the emotional sense, yet is adequate to motivate a desire to restore consistency.
Interpretations Concerning Treatments
There is some suggestive evidence in favor of the assumption that the balance principles operate like a Geştalt, since the dis-traction-from-the-whole treatment damaged discrimination of balance from unbalance (a balance effect, or an outcome supporting Heider's position) while the concentration-on-the-whole treatment facilitated it.
There was also some suggestive evidence that "affective"
processes, as indexed especially by the all-sentiment-relations treatment, tend to get in the way of "cognitive" operations of balance principles. All-sentiment-relations, when used together with consistency, produces a relatively weak balance outcome
(although stili considerable, due to high structure). Sentiment, when used together with pleasantness, though stili producing a balance effect (even if low and opposite in direction to that of
con-7 It is interesting to note that although dissonance theory has been frequenty criticized for failure to operationalize satisfactorily the state of dissonance, balance theory has not been subject to this particular çriticism.
88
sistency), loses most of what it can do by vvay of discrimination when used vvith consistency. I assume this is due to the many
"concrete" affective referents (see the section on affective pro-cesses in Chapter 1) supplied by the sentiment treatment vvhich encourage individual differences to come out in pleasantness responses and so svvell the unexplained variation (although it also provides cues to bring about some other type of discrimination, vvhich is "polar opposite" to that of consistency, vvith the help of high structure).
P-identification vvas intended as a precaution to reduce the actor-observer differences (as described in attribution literatüre).
It vvorked vvhen used together vvith pleasantness, but had very little effect vvhen used vvith consistency. Perhaps, actor-observer differ-ences have a tendency to diminish vvhen the task is cognitive in na-ture. The cognitively oriented quality of the standard questionnaires makes it conceivable that, vvhen subjects are dealing vvith a lot of factual information under the appropriate response set (here, cognitive or problem-solving set), their perception of the situation and those of the actors tend to converge.
The primacy treatment came out to be the only one vvhose results vvere not in line vvith our expectations. An ad hoc explana-tion for this failure could be the follovving: In retrospect, it ap-pears that the primacy manipulation vvas unduly vveak. The content of the P /O relation is identical in ali stories, vvhile the remaining bonds ali shovved variation from story to story. A constant P /O description from story to story may have resulted in a habituation phenomenon vvhich vvould vvork against the intended efforts to increase the salience of this particular bond. Moreover, vvhen in interaction vvith high structure, this "identity" characteristic could have resulted in a process that approximates the concentration treatment outcome. Since this identical statement is presented first in each story, and since identical statements do not aid tovvard extracting the distinctive feature betvveen the tvvo paired situations, subjects could have disposed of the identical statements, then and there. This vvay, the burden of comparison could have been reduced by one-third of its original level —comparison being necessary only vvith the tvvo remaining bonds. As a result, primacy not only possibly did not have a chance to vvork as experimental!y intended, but may have made life easier for the subjects by reducing their vvork.
89
I stili think that, with prop6r modification, primacy could be shown to work as originally predicted and would yield results similar to to those of the sentiment treatment.
Even though I did not manipulate the variable of stimulus
"structure",some of the results can be pointed out as suggestive evidence for it. Despite the fact that the standard / consistency treatment was expected to do a fairly good job of producing balance effects (i.e., discriminating balance from unbalance), its level of discrimination was a lot better than I had expected (possibly due to high structure). As a result, when being compared with the other treatments, this condition possibly produced a ceiling effect vvhich underestimated the effects of the other relevant treatments.
It is possible that, if it vvere not for high structure, primacy too could have produced results more in line vvith my expectations.
Similarly, I am arguing that, if it vvere not for a high level of struc-ture, the P-identification /consistency condition could have vvorked better, but the distraction treatment, as well as sentiment treatments could have done a lot vvorse than they did. Therefore, there is reason to believe that, altogether, the results are somevvhat more conserva-tive (due to high structure) in shovving treatment effects.
The degree of structure in the stimulus display has been knovvn for a long time to be inversely related vvith individual's contribution of subjective material to the situation (Sherif, 1935). The variable of stimulus structure, as discussed in Chapters 2 and 3, seem to be responsible for some of the outcomes in this study as vvell. It vvas ex-pected that high individual participation vvill result in idiosyncratic responses (boosting individual differences) — quite likely vvith the use of paper and. pencil procedures. And this vvas expected to result in a lot of "noise" in the data, vvhich vvould be reflected in large amounts of variation not explainable by discrimination of balance and unbalance. With precautionary measures taken tovvard the goal of maximally structuring the stimulus situation:
a) The magnitude of discrimination associated vvith consistency vvas considerably larger in this study as compared to, for example, the results of the Miller and Norman study (1976), (see page 23).
b) By the same token, the magnitude of discrimination associ-ated even vvith pleasantness vvas also larger in this study. In fact, much of the reason vvhy the pleasantness ratings produced a significant dif-90
ference between balance and unbalance here (even though opposite in direction to the pattern of consistency) than in previous studies
— excepting those with future interaction implications—could be attributed to the highly structured content of the questionnaires.
Apart from the fact that the primacy treatment may have lacked sufficient controls in its manipulation here-, the conservative treatment effects due to high structure were almost inevitable in this study, since, being built on a paradox, it aimed at two contradictory goals simultaneously: to produce enough discrimination to show that Heiderian balance principles can operate even within Nevvcomb's
"nonbalance" categories, and also to reconstruct (simulate) some of the conditions (in the form of various treatments) suitable for producing "nonbalance" outcomes.
Conclusions
1) Judgment of consistency is a measure that seems well suited for testing the Heiderian balance expectation that subjects will differ-entiate balanced and unbalanced triads, even with a negative P/O.
2) Judgments of pleasantness are responsive to variables other than balance alone, for instance agreement, and consequently con-stitute an inappropriafe measure for testing consistency theory pre-dictions.
i
When P/O is negative,plea santness ratings do not discriminate strongly the balanced triads from the unbalanced triads as pleasant, but the state of unbalance disposes subjects to rate the triads as unpleasant. Sometimes these two tendencies (desire for balance and desire for agreement) may be roughly equivalent, in which case Nevv-comb's nonbalance effect may be observed. But in some cases, as in the present study, preference for agreement may be sufficiently stronger to produce a pattern of results which is opposite in direction to Heider's predictions, and not in accord with Nevvcomb's expecta-tions, either.
3) The prior studies that supported Heider more than they did Newcomb, or failed to support Nevvcomb, seem to share the property of using a dependent measure which has at least the surface validity of being cognitive in nature — e.g., consistency, comprehensibility, expectedness, probability.
91
4) Nevertheless, using a cognitive rating task alone seems to be a necessary but not a sufficient condition to obtain appreciable balance effects (discrimination of balance from unbalanee) in an experimental setting, or with paper and pencil techniques in particular.
This seems to be the case because the participants do stili have the option to pay differential attention to the component parts of the triadic situation, information about vvhich necessarily comes to subjects in a sequential fashion when presented via paper and pencil techniques.
To the extent that the balance phenomenon is a Geştalt and is achieved as a totality of and above its component parts, participants must take ali three relations into consideration simultaneously. And this seems to point to a phenomenological state neither of information inflow nor of recall, but a state beyond these—more likely that of active processing of the already existing information; that is to say, performing of mental operations on \vhatever information is acquired.
Accordingly, instructions vvhich distract subjects' attention avvay from the totality or which engage affective processes will serve to diminish the arrçount of discrimination betvveen balanced and unbalanced situations, even when utilizing consistency judgments.
5) Prior studies vvhich failed to support Heider seem to share the property of using response indices vvhich could be characterized as affective in nature •— e.g., pleasantness, uneasiness, involvingness, tension.
6) The use of pleasantness ratings in combination with all-sentiment relations results in a drast»cN"eduction in the amount of differentiation betvveen balance and unbalanee. A similar reduetion in discrimination is observed also with the distraction-from-the-whole treatment. There is reason to believe that the primacy of P /O relation vvould have behaved very much the same had it been manipulated better in this study. Further, lack of P-identification and lack of strueture seem also to be responsible for some attenuation in discri-mination. Thus it seems that a combination of certain response measures and some treatments can bring about a considerable reduetion in discrimination.
In sum, the complications in the results revievved in Chapter 1 can be traced to the effect of various treatments vvhich are isolated in this study, or to their joint effect with the types of measures
92
utilized. Since the description of the materials and instruments is not usually stated in detail in journal articles, it is difficult to pin down exactly vvhich of these treatments may be reşponsible for the particular results. But chances are that more than one such factor is simultaneously involved. The studies that came up vvith mixed or contradictory results could, be associated vvith 3 or 4 of these factors. Since these variables vvere not usually controlled in prior research, some of the variation in the results of the previous studies appears due to the variation, among these studies, in the variables delineated by the treatments of this study.
Moreover, the effect of such variables probably extends beyond the rating tasks used in the present study to other kinds of measures as vvell. For example, the eompletion task generally favors a Heiderian position. Yet, some studies utilizing this measure found support for Nevvcomb's position. A systematic bias similar to the pleasantness type of rating is probably not involved in the case of this measure. Yet, it seems likely that a eompletion task vvould
— like the consistency ratings in the present study — be sensitive to variations in such variables as the number of sentiment relations, primacy of P /O (or primacy of P /X), distraction-from-the-vvhole and /or lack of structure vvithin the content of the eompletion task materials. Since the present study included tvvo dependent measures only, the above speculation remains to be further investi-gated in the future.
93
APPENDIX A
I •