III. Experimental Studies
A. PRELIMINARY RESULTS
Here one counsel is valid: Trust the inadequate and act on it; then it will become a fact.
Studies on minority influence have predicted from the very start that the minority's impact on subjects' private responses would be equal to or greater than its impact on their public responses. I know of no experiment where this did not hold and where different results were obtained. It is also evident that the discrepancy between the public and private reactions is more marked where the conflict of influence is sharpest-that is, when a factual judgment is involved -- and less marked where a value judgment or preferences are at stake.
This is apparent in several experiments, three of which I shall present here.
In one experiment (Moscovici, Lage, & Naffrechoux, 1969), groups of four naive subjects and two experimental confederates were asked to make a series of color perception judgments. The stimulus slides were blue. The subjects were asked to report aloud the name of the simple color they saw and to estimate the light intensity in numerical terms. Of course, on every trial, the confederates said that the slides were green. Upon completing this procedure, the subjects were asked to take part in a second, ostensibly unrelated, experiment concerning the effects of training upon vision. The subjects, tested individually, were exposed to a number of disks in the blue-green zone of the Farnsworth Perception Test. For each disk, the subjects were asked to name the simple color they saw.The results of this second experiment indicated that the Perceptual threshold of subjects who had previously been exposed to the consistent minority shifted. They saw as green disks that are usually perceived as closer to blue.
(I shall come back later to this proposition, which requires a stronger foundation). Furthermore, subjects who did not give any "green" responses during the social interaction phase of the experiment were even more likely than those who did join the minority position at least once to call the disks green. The results are satisfactory enough to enable us to conclude that (a) the change in response was stronger in the private than in the public situation; and (b) subjects who felt completely blocked, incapable of adopting the minority judgment in the presence of the others, solved their conflict by adopting the judgment even more, probably without being aware of it, once they were alone. Nothing of the kind was observed when the minority was inconsistent or when the source of influence was a consistent majority (Moscovici & Lage, 1976). This suggests that the consistent majority produced compliance rather than conversion, public without private change.
The experiment by Nemeth and Wachtler (1973) yielded different and more typical results, that is, responses given under social pressure remained the same after this pressure was withdrawn. Subjects, run in small groups, expressed a preference for one of a pair of paintings on each of 19 trials. An experimental confederate consistently expressed a preference for either German or Italian paintings, himself being allegedly German or Italian. Irrespective of his seeming bias or minority position, he exerted a strong influence on the group. A group discussion ensued and, in its wake, the subjects were again asked to state their preferences for each of the 19 pairs of paintings. This time, however, they indicated their choices anonymously in writing instead of giving them aloud. No differences were found between public and private responses, which suggests that subjects had not merely complied but had in fact undergone a real change.
Common to both of these experiments is the fact that they both utilized a direct index, the very same response, to measure the transfer of influence from one situation to another. It was therefore important to see whether this transfer or generalization would occur when more unobtrusive measures were employed, measures associated indirectly with the message received or with the subject matter discussed in the group. An experiment by Wolf (1977) demonstrated that this was in fact the case.
Female subjects, run in groups of four, were led to believe that they were interacting as members of a jury. Their task was to decide upon an amount of compensation to be awarded to the plaintiff in a civil suit. The facts of the case were weighted so as to encourage subjects to award between $20, 000 and $30, 000. The influence agent, ostensibly one of the group members, advocated the minority position of $3000 throughout their interaction. Three variables were manipulated: the cohesiveness of the group (high or low), the behavioral style of the influence agent (high consistency or low consistency), and the opportunity for rejection of the influence agent from the group (rejection possible or not possible).
The primary dependent measure determined the change in private judgment concerning the compensation award from pre- to postdeliberation. Change in the direction of a decreased award reflected the minority's influence. The results showed that subjects in the high-cohesive conditions reduced their awards to a greater extent than subjects in the low-cohesive conditions. Furthermore, within the high-cohesive conditions, the strongest minority influence was found in the high-consistency/no possibility and the low-consistency/possibility conditions.
Four measures of latent influence examined the cognitive-perceptual effects of the minority influence attempt on subjects' judgments. In an ostensibly unrelated task, the subjects responded anonymously to a series of questions concerning the severity of different prison terms and fines and the usefulness to a plaintiff of different compensation awards. Four measures of latent influence were embedded in the items. On 21-point scales, subjects rated the penalty value of a $3000 and a $20, 000 fine and the utility value of a $3000 and a $20, 000 compensation award. The questions concerning $3000 measured subjects' perceptions of the critical stimulus value employed in the social influence situation. The questions concerning $20, 000 measured the generalization of the effect to a different value. The $20, 000 reflected a normative initial judgment in the original influence situation. Subjects whose perceptual codes were influenced by the minority should have viewed the fines as more severe and the awards as more useful than subjects who were not so influenced.
The results of the latent influence measures paralleled those obtained on the judgment-change measure. A major effect for cohesiveness was found on the two utility-value measures. Subjects in high-cohesive conditions perceived both $3000 and $20, 000 as more useful to a plaintiff than subjects in low-cohesive conditions.
Interactions were revealed on the two penalty-value measures. On both measures, the strongest minority influence was found in the high-cohesive/ high consistency/no possibility and the high-cohesive/low-consistency/possibility conditions. The latent influence data, then, replicated both the major effect and the more subtle interaction produced on the judgment-change measure. The results of these unobtrusive measures show that subjects' perceptions of the critical stimulus were altered as a function of the influence situation to which they had been exposed and that this effect generalized to a novel but similar stimulus value. Thus, the minority influence attempt succeeded not only in changing subjects' overt responses but in changing the context or cognitive code underlying those responses as well.
The preliminary outcome of these experiments is clear. A consistent minority produces at least as much, if not more, change in private than in public situations.
The public-private difference reflects both the intensity of the conflict and the nature of the stimulus around which the influence attempt revolves.B. DIRECT AND INDIRECT INFLUENCE
"One cannot break the chains where there are none visible." This statement by Kafka to one of his young admirers is indicative of the kind of incapacity one experiences, unwittingly, when trying to escape from the insistent pressure of a minority. The mere advocacy of a deviant point of view arouses strong feelings in us and binds us by creating an unexpected complexity and ambiguity in a situation that is usually clear and banal. This is enough to stimulate interest, to gain a hearing, and to trigger strongly favorable or adverse reactions.
Any minority that wishes to have an impact must be consistent, but it can be so in two ways. Consistency can be achieved, on the one hand, by speaking and acting in a rigid manner. This unconciliatory behavioral style heightens the threatening and anxiety-producing aspects of the conflict with the majority. A fair behavioral style can be achieved, on the other hand, by consistently manifesting a combination of firmness and flexibility in word and deed, in which the strength of one's own opinions does not preclude taking the opinions and ideas of others into account. This behavioral style provides a way of circumventing the conflict with the minority.
"Rigid" minorities, as a matter of course, block any solution that precludes a wholesale adoption of the deviant position. "Fair" minorities, on the contrary, leave the door open to reciprocal concessions and remove an all-or-nothing character from the adoption of a deviant point of view. The first type of minority creates a more intensive conflict and hence blocks public expression of agreement more completely than the second. It follows that the former will generally produce more conversions, that is, indirect influence, than the latter.
To put it differently, a "fair" minority, exemplified at times by the Socialists, will gain acceptance for the judgments or opinions explicitly contained in its message, whereas a "rigid" minority, such as the extreme left, will on the contrary gain acceptance for judgments or opinions that are implied or derived from the content of its message. A remarkable series of experiments by Mugny (1974) and Papastamou (1979) shed light on the meaning of these suggestions. They were conceived along the same lines as the usual experiments on communication and generally dealt with the topic of pollution. At the same time they presented three special features:
1. The messages were attributed to a minority group or institution.
2. The message style was either "rigid, " that is, expressing dogmatic and extreme points of view, or "fair, " that is, expressing a consistent point of view but with some "conciliatory" overtones. A slogan of the rigid message, for instance, stated: "Let us close down the factories that do not abide by the rules, " whereas the slogan of the fair message expressed its position in these terms: "Let us force the automobile producers to supply their vehicles equipped with antipollution filters."
3. The questionnaire filled out by .the subjects before and after reading the message consisted of an equal number of direct and indirect items. The direct items were nearly identical with those contained in the message, whereas the indirect items related to the topic but were not explicitly contained in the message. A change obtained in the latter items would reflect more than a simple adoption of minority positions; it would imply a certain "generalization" produced by connections with or inferences from the opinions upheld in the latter's statement. Conversion is reflected here in the difference between direct and indirect influence. It is the outcome of each person's own deliberation.
Let us now proceed to review this series of experiments, without going into great detail.
At the outset of the first experiment, the subjects responded in the usual fashion to a questionnaire concerning the causes of and responsibilities for pollution. They were subsequently divided up into four groups, each group receiving a message in either a "rigid" or a "fair" style and with either a "dissonant" or a "consonant" content in terms of the general beliefs of the population. (This constituted four communication conditions: rigid-dissonant, rigid-consonant, fair-dissonant, and fair-consonant.) After reading the appropriate text, each subject responded once more to the same questionnaire.
What was observed? Changes in responses to the indirect items were proportionately greater than changes in responses to the direct items, that is, to those actually contained in the message (F = 5.75, p < .005). However, the details are much more interesting. The minority whose style was fair and whose image was more flexible and less threatening produced an equally great change in responses regardless of the nature of the items. The rigid minority, in contrast, had little influence on responses to the direct items, but its influence was far greater on the responses to the indirect items. The change produced by the rigid minority therefore manifested itself not with respect to the views it upheld but with respect to other, related points. This difference was even more marked when the minority message was dissonant, in other words, when it contrasted sharply with majority beliefs (F = 4.02, p < .05). The results are self-explanatory. Wherever a minority exerts a weak pressure and its opinions do not clash head-on with the convictions of the population, the conflict is less intense and the influence effect is direct. However, as soon as the minority's pressure increases and its opinions begin to clash with the subjects' convictions, the influence observed is largely indirect. There is a lesson to be learned here with respect to our measures and the need to determine in advance, in a theoretical manner, which ones are appropriate. Had Mugny been content to use only direct items, the conclusions drawn from this experiment would have been entirely different and would have confirmed the commonsense notion that a fair minority is more effective than a rigid one.
In a second experiment along the same lines, subjects received a communication whose content was either consonant or dissonant with the attitudes of the general population toward foreigners. Some subjects, however, received a message in a consistent style, that is, a consistently structured text, while others received a message in an inconsistent style. The aim of the experiment was to show that only minorities that appear consistent exert true influence.
The results showed that this is in fact the case. It was also found that when the content of the communication was consonant with the subjects' opinion, it exerted a greater influence than when it was dissonant. To avoid misunderstanding, I repeat that this applied only in the consistent conditions. Moreover, the consonant communication produced response changes primarily for the direct items, while the dissonant communication produced response changes primarily for items related to but not actually contained in the text (F = 3.16, p < .10).
The stronger the ideological "opposition," or the greater the resistance to minority pressure, therefore, the stronger the reluctance to accept its message, and the greater the likelihood that its point of view will be accepted in an indirect manner; the contrary occurs when this opposition is less pronounced. This phenomenon is analogous to one that was observed a number of years ago by Schonbar (1945), to wit, that the longer people resist changing their position under conditions of social pressure, the more persistently will they maintain a modified position in the situation that follows.
In a third experiment, the subjects received either a "rigid" or a "fair" message on the topic of pollution. They were advised at the outset that after reading it, they would respond once more to the questionnaire they had just filled out. The authors of the message, they were told, had asked the experimenters to try to find out whether the subjects had been influenced by the message. This instruction was delivered explicitly to produce resistance, to block the process of coming to terms with the minority. It was assumed, additionally, that this blockage would be intensified by the fact that the subjects were asked to cooperate in one way or another with the source and thus identify publicly with a deviant and extreme point of view. In order to verify this assumption, conditions of unilateral and bilateral influence were created. Subjects in the unilateral influence condition were told: "Your responses are of interest to the authors insofar as they want to take your opinions into account in modifying the message."
The prediction was straightforward: Changes with respect to the indirect items would be greater when the message was rigid and the influence bilateral than when the message was fair and the influence unilateral. For lack of an adequate number of subjects, it was possible to complete only three of the four conditions: fair style-unilateral influence, rigid style-unilateral influence, and rigid style-bilateral influence. The results were as predicted: Greater opinion change was found on the indirect than on the direct items (F = 34.19, p < .001). With respect to the three conditions, it may be noted that the difference between response changes to the direct and the indirect items was practically nil in the first condition (fair style-unilateral influence), that it increased perceptibly in the second condition (rigid style-unilateral influence; F = 8.06, p < .05), and that it became quite large in the third (rigid style-bilateral influence; F = 37.81, p < .001).
The data, which are presented in Table I, clearly confirm our conjectures: There is in all cases an influence effect.
If the minority is perceived as fair, it produces as great a modification in responses, as much agreement, in short, on items contained in its message as on items that are only related to it. If the minority is perceived as rigid, however, it has less or even no impact on the items reflecting its position and relatively more impact on other items that it does not mention but that are related to its message. The more rigid the minority appears and the more subjects are asked to cooperate with it, the greater this indirect impact.
This should not come as a surprise to us. We have observed, over the last several years, how militant minorities and extremists in political and cultural circles have affected our outlook, changed our manner of behaving, dressing, speaking, etc., without at the same time leading most of us to accept their positions or making us act as they would wish.TABLE I MEAN CHANGES OF OPINION WITH RESPECT TO POLLUTION [a]The greater the conflict produced by the minority, by virtue of its rigidity or consistency, therefore, the greater the extent to which the change taking place will be the outcome of the subjects' own inference.
This outcome is far less affected by the specific content of the message than by the global view that this content expresses. Consequently, a seeming lack of influence at any given moment may be misleading. What has been left out of account is the fact that while the persons may have refused to comply publicly and to identify themselves as deviants, they may have been privately converted to the thinking of the minority.C. CONFLICT AND CONVERSION BEHAVIORAll experiments of this kind demonstrate that minority influence is greater on indirect items than on the direct items contained in the message, and that this applies even more to rigid than to flexible messages. It would follow logically that if this difference in impact results from a conflict between the opinions held by the source and those of the audience, it should be more marked if the deviant character of the source is highlighted and its pressure increases. The simplest way to create such a situation is to strengthen the minority numerically: If several deviants express the same point of view, it becomes much more difficult to reject it by attributing it to "subjective" factors; hence one must concede that something "objective" must also be involved. In this case the conflict between two judgments of fact rather than two judgments of value.
In a first experiment based on exactly the same principle as the ones I have just described, Papastamou (1979) has created two conditions: In one condition the subjects read a text that expresses minority positions on pollution, while in another condition they read the two halves of this text, in slightly revised form, but with the first half of the text attributed to minority group X and the second half to minority group Y. The order in which the text segments are presented as well as the attribution to group X and Y is of course neutralized. As was the case in the previous experiments, the message style is either rigid or fair, so that the following communication situations arise: a "fair" source-a "rigid" source-two "fair" sources-two "rigid" sources. On the basis of what we have found out so far, we expect the indirect influence to be greater both in the situation where the source is perceived as rigid and where the two sources are similar. The results support these conjectures. It can be noted, first of all, that while the two messages from the two flexible sources do not exert more influence than one message from a single flexible source, two messages stemming from two rigid sources exert significantly more influence (F 1/68 = 5.709, p < .05) than a message from a single rigid source. However, as was to be expected, this influence is not identical on the direct and the indirect items. It is first of all far lower for the former than for the latter (F 1/68 = 44.186, p < .0005). In addition, on closer inspection it becomes apparent that the difference between the situations with a single source and with two minority sources is not at all significant for the direct items; it becomes significant only for the indirect items (F 1/68 = 4.151, p < .05). In other words, by increasing the intensity of the conflict by means of attributing the message to two sources instead of a single source, a significantly higher conversion rate is obtained than would otherwise be the case.
In a second experiment, which was analogous to the previous one, the subjects were asked to read two texts attributed to different groups expressing minority opinions on the pollution problem. Before reading their texts, however, they were faced with an introductory page in their booklet informing them that in the following section there would be two notorious minority groups expressing their stands with respect to this problem. The relation between these groups was not presented in the same way in all the situations, however. In one situation, as a matter of fact, the groups in question are said to be in a minority position within the general framework of the antipollution campaign, but coordinating their efforts to fight efficiently against pollution. In the other situation, it is stated that these groups are competing for influence, each one trying to outstrip the other in its campaign against existing pollution. Therefore, there are minorities appearing 'to be cooperative, on the one hand, and minorities appearing to be competitive, on the other. In addition, they are either flexible and fair or rigid and extremist. It might have been reasonable to expect that cooperative minorities, generally speaking, would have greater influence than competitive minorities and that their direct influence would be greater. Nevertheless, in keeping with the points I have just presented, this is not the case.
On the whole, the relation between the minorities seems to have no impact, since no difference between the cooperative and the competitive situation can be observed. However, the nature of the message makes a difference, because rigid minorities have greater influence than flexible minorities
(F 1/68 = 4.222, P < .05). The customary effect does not manifest itself here; in fact, it is reversed. It therefore seems that when there are two minority sources, a firm position has a greater payoff, while flexibility is more productive for a single source
. This conclusion must be qualified, however. A breakdown of the effect of the message style on the cooperative and the competitive level reveals the following: Although the two styles have a more or less equal impact for the competitive situation, the same is not true where the two minorities are supposed to cooperate. There a rigid message produces a much stronger impact than a flexible message (F 1/68 = 6.183, p < .01). Let us now compare the responses to the direct and the indirect items (Table II). The nature of the message makes no difference as far as the direct items are concerned. For the indirect items, however, the rigid minorities have a greater positive influence than the flexible minorities. In the light of these results, as a whole, one may say that cooperative relations between minorities, especially rigid minorities, emphasize their deviant character. If they thereby have an impact on the audience, this impact manifests itself more strongly on the indirect than on the direct level, that is, by conversion rather than by compliance.TABLE II. MEAN CHANGES OF OPINION ON DIRECT ITEMS AND ON INDIRECT ITEMS [a] D. MINORITY INFLUENCE, MAJORITY INFLUENCE, AND COMPLIANCE
Minorities are not the only ones to seek and secure conversions. It is probable, though, that they are the ones who most often produce such an effect. To compare their influence with majority influence, I carried out the following experiment in collaboration with Mugny and Papastamou. Although it was similar to the previous experiments in that the communication dealt with pollution and the message read by the subjects were either "rigid" or "fair, " it differed from the others in two important respects:
(a). The same messages were attributed either to a majority source or to a minority source.
(b). The effectiveness of the messages was measured immediately after and after a certain period of time had elapsed, as was done in the experiments on the "sleeper" effect.
The experimental procedure consists of three phases. During a first phase (pretest), the subjects responded to a 20-item questionnaire by circling the number expressing the extent of their agreement or disagreement with each statement. The second phase, the actual experimental phase, took place a week later. The subjects were informed of the positions attributed either to a majority or to a minority group, depending on the conditions, and they read a text on the pollution problem. There are altogether four conditions: rigid-majority, fair-majority, rigid-minority, fair-minority. Right after the perusal of the text, the subjects answered the usual opinion questionnaire. Three weeks later, during the third phase, the subjects once more filled out the questionnaire on pollution problems. They then answered a series of complementary questions intended to measure how accurately they remembered the exact nature of the influence source and the opinions it had expressed. In this case the subjects were Swiss adolescents, about 15 years of age and living in a suburb of Geneva. Like many young people in Europe, they were personally concerned about ecological questions, so one can say that the content of the messages mattered to them.
What were their responses right after reading the text? The results we obtained show that, irrespective of the nature of the source or of the message, there is practically no influence. However. slight as this influence is, a significant difference appears: The subjects let themselves be influenced a little more on the direct than on the indirect items (F 1/80 = 4.876, P < .05). There was, therefore, a slight compliance effect that was reinforced by our instruction. The results obtained 3 weeks later present a different picture. To begin with, the minority appeared somewhat more influential than the majority (F 1/80 = 3.85, p < .10).
As expected, the influence exerted was not the same for the two groups (Table III). On the direct items, the majority and the minority did in fact have a similar effect. On the indirect items, conversely, the majority obtained a negative effect, the subjects returning to the opinion that they held prior to reading the messages, whereas the minority obtained a positive effect, the subjects accepting the point of view upheld in the message to a greater extent at the later time than immediately after having read it (F 1/80 = 7.88, p < .01).
That is not all. The message style does seem to playa certain role. The breakdown of the interaction between the source factor and the style factor of the message provides us with interesting results. With respect to the majority, it made no difference, on the whole, whether the style was rigid or fair, whereas it made a difference with respect to the minority. The fair minority actually lost some of its influence on both the direct and the indirect items, although less, on the whole, than the fair majority. The rigid minority, on the contrary, lost much of its influence on the direct items but gained considerable influence on the indirect items (F = 5.36, p < .05).
This was the minority that won the greatest acceptance for its point of view by the audience 3 weeks later, on opinions that it had not explicitly advocated.TABLE III. MEAN CHANGES OF OPINION BETWEEN THE PRETEST AND THE SECOND POSTTEST [a]
Before I draw any conclusions from this experiment, I would like to mention some other interesting results. When the subjects were asked about the source of the messages after completing the questionnaire 3 weeks later, they recalled the identity of the minority source more accurately than that of the majority source (30 out of 41 as against 13 out of 41). Irrespective of its rigid or fair style, it seems, the minority had a stronger impact than the majority. Similarly, subjects confronted by the minority remembered the content of the message more accurately than those confronted by the majority (F = 17.434, p < .001). The message style seems to have a slight effect in that, on the whole, people remember "fair" messages less clearly than "rigid" messages (F = 3.015, p < .10).The conclusions of this experiment were very clear indications that a majority produces compliance, which disappears over time, whereas a minority produces conversions -- indirect conversions, that is -- especially if it is rigid.
Curiously enough, the "rhetoric, " in other words, the behavioral style, is relevant only for the latter and not the former. The question arises why the identity and the content of a majority message are more easily forgotten than those of a minority. There is one plausible explanation: The majority message is treated more passively than the minority message. Whatever the validity of this explanation, the fact remains that there is no clear-cut relationship between accuracy of recall of the source and the opinion change observed. Thus, subjects confronted with a deviant source remember it equally well whether it is rigid or fair, and yet they do react differently in each case. One finding persistently runs through all these results: These effects obtained by a minority have a specific character and are not, as a rule, obtained by a majority.This experiment has some implications for the body of research on the "sleeper effect." As is well known, this effect consists of a delayed message influence on the opinions of an audience. This delayed influence is explained by dissociating the source from the message: Once the source is forgotten, the content alone continues to act on individuals' opinions. In our experiment, it was rather the opposite that was observed: The delayed influence went hand in hand with a greater recall of the source.
If that is the case, the usual explanation could not be applicable. The' 'sleeper effect" seems to occur, moreover, only when the source is deviant, even extremely deviant. When this source is conformist, it is more likely to cause a resumption of the individuals' original opinion
. It would be more accurate to say: The sleeper effect is associated with a deviant or minority message. Among the minority messages, moreover, those whose content people are more likely to have forgotten are the ones that are the most effective. Actually, subjects confronted with rigid minority remember its exact opinions less accurately than subjects confronted with a fair minority (F = 7.58, p < .05); and yet those subjects show the greatest change
. In short, to the extent that forgetfulness plays a part at all, the opposite of what has been asserted takes place, since it affects the content rather than the source of the message.
How can the results we have obtained be reconciled with those hitherto obtained? There is no clear-cut answer to this question. I would only like to make these observations as a contribution to a possible future discussion.The sleeper effect was demonstrated, not surprisingly, in earlier studies with a typically counterattitudinal message, that is, a message clashing with the norms and beliefs of the audience. The delayed attitude change was explained by saying that once the identity of the source, which had interfered with the acceptance of the content, had been erased from memory, its opinions could be accepted because the content could now be judged on its own merit.
A negative message, however, may have two meanings, depending on the context: On the one hand, it is in opposition to something and, on the other hand, it is deviant in relation to something. It is conceivable that in the earlier experiments it had its impact not so much because of its association with a negative source as because of its expressing a deviant position that upset some widely held beliefs or attitudes, in other words, because it was counternormative and not because it was counterattitudinal. The reason we have obtained such striking effects is that we have, on the one hand, emphasized this counternormative aspect by associating it with a minority and, on the other hand, deemphasized it by associating it with a majority. It is likely that the contradictory results and contradictions of the "sleeper effect" analyzed by Gillig and Greenwald (1974) are attributable to the fact that these two aspects of the message have not until now been distinguished and that experimenters have failed to control this variable, although individuals sometimes react to one aspect and sometimes to the other. If this is the case, it follows that a large part of the attitude change attributed hitherto to the "pro" and "con" character of the message, to the positive or negative content of the communication, could be attributed to its deviant or nondeviant meaning. This calls for a reevaluation of an entire body of generally accepted data and ideas that would benefit from being viewed in a new light. Even dissonance theory might be reevaluated in order to determine whether dissonance results from opposition between two opinions, opposition between a deviant and a nondeviant opinion (Nuttin, 1975), or opposition between a deviant behavior that one is obliged to assume publicly and the norms and beliefs to which one privately subscribes.Another question now arises. In our experiment, the majority failed to obtain conversion, but why did it fail to obtain compliance?
Was this because of the message content or because of some other factor? If it was caused by the content, then the results obtained by us and the conclusions drawn from them are valueless. However, it is much more likely that we failed to obtain compliance because one of the causes of this behavior was missing, to wit, the existence of an external compulsion, that is, social control. Papastamou (1979) confirmed this supposition.
His experiment used the same kind of materials involving the struggle against pollution, but he attributed it to a majority source. This experiment consisted of three phases. In the first phase, the subjects gave individual answers to the pollution questionnaire as well as to some supplementary questions. Ten days later, the second phase took place. At this point a majority communication was read and the subjects were asked to draw up their own arguments justifying their stand on the pollution problem. At the same time the subjects were informed that the group that had drawn up the message asked in exchange to be acquainted with the opinions of part of the persons involved, and to be given their names and addresses so as to be able to contact the various people for a later discussion. This introduced the notion of social control. The experimenters then pretended to draw lots in order to respond to this request. They thereby created four conditions:
(a) In the individual control condition, the subjects were informed that they were among those whose names and replies would be communicated to the authors of the texts they were about to read.
(b) In the collective control condition, the subjects were informed that they were among those whose discussion group and replies of their group would be communicated to the authors of the text they were about to read.
(c) In the individual noncontrol condition, the subjects were informed that they were among those whose replies would be communicated to the authors of the text, but nothing else.
(d) In the collective noncontrol condition, the subjects were informed that they were among those whose names and replies would not be communicated to the authors of the text they were about to read and that only the numbers of the discussion groups would be given to these authors.
In a third phase, which followed immediately, the subjects once more answered individually the pollution questionnaire and some complementary questions. The results show that the majority source has a strong influence and that this influence bears more strongly on the direct than on the indirect items (F = 20.77, p <.001). In contrast to the changes in response produced by a minority, changes in response produced by a majority are limited to the opinions advocated by the majority; there is no impact at all on the items when inferences would have to be drawn from its messages. This connotes a passive acceptance. What are the effects of external compulsion? It can be observed that where social control is "weak, " the degree of influence is the same in the individual and in the collective conditions. In contrast, where it is "strong, " the majority exerts considerable influence in the collective control condition and very little influence in the individual control condition (F = 6.55, p < .05). In other words, an isolated individual tends to offer resistance, while individuals explicitly belonging to a group tend to conform. If one takes into account the fact that the response changes occur only for the direct items and that they are more pronounced for subjects belonging to a group, it follows that compliance takes place and that it is maximized where external compulsion is maximized.
This agrees with Festinger's (1953) and Deutsch and Gerard's (1955) hypotheses. Additional results confirm our conclusion. In the postexperimental questionnaire, the subjects' compliance in a potential discussion with the authors of the message is measured by asking them whether they would try to reach a consensus. As was to be expected, the subjects in the control conditions assert that they would want such a consensus (F = 3.66, p < .10) and this desire is stronger in the collective control than in the individual control condition (F = 4.03, p < .05). Furthermore, as I mentioned at the beginning of this study, such a control on the individual is bound to provoke resistance, as results show. In his experiment, Papastamou proceeded to analyze the arguments written by the subjects to justify their stand, categorizing the arguments as either "majority" or "minority" with respect to their content on the pollution struggle. The score shows that the subjects in the individual control condition produce more minority arguments than the subjects in the collective control condition (F = 2.842, p < .10) and those in the weak social control conditions (F = 3.942, p < .05). In short, they not only refuse to adopt majority opinions but even start adopting minority opinions and thus become deviants. The results obtained seem to me to display great coherence. Majority influence in combination with social control, even weak social control, leads to compliance, which is maximized when there is strong collective control and minimized when there is strong individual control.
The absence of this compliance in the previous experiment therefore makes sense and the interpretations concerning conversion and the "sleeper effect" remain plausible.These results taken as a whole refine our understanding of minority influence. As predicted, fair minorities have their strongest impact on the specific responses of individuals belonging to the majority. Rigid minorities seem to have their greatest impact on the cognitive code underlying these responses. In this sense one may say that the former have a direct effect and the latter an indirect one.