22
Tocqueville on conformism
The mechanisms of belief formation I considered in Chapter 7 operate for the main part at the level of the individual, in the sense that the beliefs held by one person owe little to those held or expressed by others. In this chapter I discuss some mechanisms of collective or interactive belief formation. To illustrate the distinction, consider Tocqueville's analyses of American conformism. One explanation why Americans tend to have the same ideas is simply that they live under similar conditions: Since “men equal in condition … see things from the same angle, their minds are naturally inclined towards analogous ideas, and while each of them may diverge from his contemporaries and form beliefs of his own, all end up unwittingly and unintentionally sharing a certain number of opinions in common.” Another explanation relies on the pressure to conform: “In America the majority erects a formidable barrier around thought. Within the limits thus laid down, the writer is free, but woe unto him who dares to venture beyond those limits. Not that he need fear an auto-da-fé, but he must face all sorts of unpleasantness and daily persecution.”
This last passage suggests that people conform outwardly, because of social pressure, but not necessarily inwardly. As he also writes, if you hold a deviant view, “your fellow creatures will shun you as one who is impure. And even those who believe in your innocence will abandon you, lest they, too, be shunned in return.” Other passages suggest that conformism reaches all the way to the soul, so that people eventually develop a sincere belief in the majority view. Two mechanisms are suggested, one “cold” or cognitive and another “hot” or motivational. On the one hand, “it seems unlikely … that everyone being equally enlightened, truth should not lie with the greater number.” On the other hand, the fact that “American political laws are such that the majority is sovereign … greatly increases its inherent influence over the intellect, for there is no more inveterate habit of man than to recognize superior wisdom in his oppressor.”
Experimental findings
I have cited Tocqueville at some length (and shall cite him again in this chapter), because of his acute insights into these matters. The questions he identified – outward versus inward conformism, and cognitive versus motivational mechanisms – are very much with us today. To address them I shall first cite some classic experiments on conformity.
In the most famous experiment, subjects were asked to indicate which of three lines A, B, and C was closer in length to a given line D. There were three conditions: private, doubly public, and singly public. In the private condition, subjects stated their answer when no one else was present, besides the experimenter. In this case 99 percent indicated that D was closest to B, suggesting the unambiguous correctness of this answer. Yet in the two public conditions a substantial minority of subjects gave different replies. In both conditions, the subject answered after several others (confederates of the experimenter) had unanimously said that A was closer in length. In the doubly public condition, in which the subject gave his answer in the presence of the confederates, about one-third agreed that A was closer.1 In the singly public condition, in which subjects stated their opinion privately after they had heard what the others said, conformism was reduced but not eliminated.
The excess conformism in the doubly public condition was arguably due to fear of disapproval. The residual conformism in the singly public condition could be due to learning (“so many others are not likely to be wrong”) or to dissonance reduction. The latter explanation seems the more plausible. Those who conformed privately with the majority are unlikely to have done so on the basis of rational learning only, given the poor cognitive status of the majority view. Some motivational factor must have been at work.
Another experiment strengthens that interpretation. Here the subjects had a more ambiguous task, detecting the distance a light source in a dark room had traveled. Although the source was in fact immobile, isolated subjects judged it to have traveled about four inches (the “autokinetic effect”). Having heard one confederate say that the light had moved between fifteen and sixteen inches, the subjects estimated the distance to be about eight inches. With two confederates making estimates in the sixteen-inch range, the estimate of the subjects was about fourteen inches. The presence of one confederate, that is, led to a four-inch increase in the estimate, and that of a second to a further six-inch increase.
In a process of Bayesian learning (Chapter 13) I can rely on other observers to correct my perception or memory. Their estimates of some fact, such as the distance traveled by the light, can serve to modify my initial assessment. How much they will affect it depends on my beliefs about the reliability of their perception and on the number of these other observers. In this experiment, the subject would presumably attach the same reliability to each confederate. Whatever that reliability might be, the change in his estimate caused by one confederate's stating the distance as sixteen inches should be greater than the additional change caused by the second confederate.2 This contradicts the findings, however, since the second confederate caused a greater adjustment than did the first. There seems to be a dissonance-reduction effect, caused by the discomfort of finding oneself disagreeing with the majority, which cannot be reduced to rational learning.
The second experiment had a further, interesting feature. It ran for several “generations,” over which the confederates were gradually replaced by naive subjects. Thus in the second generation of a two-confederate experiment, one confederate was replaced by a naive subject from a first-generation experiment, while in the third generation the other confederate was also replaced by a naive subject from an earlier generation. In subsequent generations all participants were naive subjects who had been previously exposed either to confederates or to other subjects who had been exposed to confederates, and so on. The experiment was designed so that the newly inducted subject in each generation spoke after the two others. The designers of the experiment had anticipated that the artificially high estimates would be maintained indefinitely, but were proved wrong. After about six generations in three-person groups and eight generations in four-person groups the estimates converged to four inches, that is, the distance estimate given by isolated subjects. The belief in the emperor's new clothes did not perpetuate itself indefinitely. If some cultural beliefs with poor support in reality do maintain themselves over time, it could be because the discrepancy is hard to observe or because they are supported on other grounds. The use of lotteries to identify good hunting or fishing sites, as is the practice in some societies, may have survived because of their religious significance.
Pluralistic ignorance
At the outset of this chapter I distinguished between two reasons why people at a given time might hold or profess similar beliefs: because they are influenced by similar conditions (correlation) or because they influence each other (causation). A special case of the first is provided by the many examples of simultaneous discoveries, such as the invention of the calculus by Newton and Leibniz more or less at the same time. Although no one knows exactly what the “similar conditions” were in that case, the idea may have been “in the air.” For another case of the simultaneous appearance of similar ideas consider the idea of the emperor's new clothes. Hans Christian Andersen's tale was published in 1837. In the second volume of Democracy in America, published in 1840, Tocqueville came up with a similar idea to explain the apparent stability of majority opinion:
Time, events, or individual efforts by solitary minds can in some cases ultimately undermine or gradually destroy a belief without giving any external sign that this is happening. No one combats the doomed belief openly. No forces gather to make war on it. Its proponents quietly abandon it one by one, until only a minority still clings to it. In this situation, its reign persists. Since its enemies continue to hold their peace or to communicate their thoughts only in secret, it is a long time before they can be sure that a great revolution has taken place, and, being in doubt, they make no move. They watch and keep silent. The majority no longer believes, but it still appears to believe, and this hollow ghost of public opinion is enough to chill the blood of would-be innovators and reduce them to respectful silence.
A passage from Tocqueville's Old Regime (1856) makes a similar point about religion. In the course of the French Revolution “those who retained their old faith became afraid of being alone in their allegiance, and, dreading isolation more than heresy, joined the crowd without sharing its beliefs. So what was still only the opinion of a part of the nation came to be regarded as the opinion of all, and from then on seemed irresistible even to those who had given it this false appearance.” In Democracy in America, he made the same argument with regard to the false appearance of faith: “The unbeliever ceases to believe in true religion but continues to deem it useful. Looking at religious beliefs from a human angle, he recognizes their power over mores, their influence on laws … Having lost the faith … he is afraid to take it from anyone who possesses it still. Meanwhile, the man who continues to believe does not hesitate to expose his faith to the view of all … With those who do not believe hiding their incredulity and those who do believe showing their faith, public opinion develops in favor of religion.”
In these passages, Tocqueville refers to beliefs that people profess to hold (or abstain from disavowing), not to beliefs they actually and sincerely hold. In this respect his analysis differs from behavior in the moving-light experiment and in the singly public condition of the line-matching experiment.3 This is not, however, a hard and fast distinction. As I have argued in several places, it is not always clear what it means to “believe” that something is the case. Even in the singly public condition, the “belief” of the subjects who said that A was the matching line may have been somewhat faint. They might not, for instance, have been willing to bet money on the proposition. Also, stating a belief may, under some circumstances, induce a tendency to endorse it (Chapter 7).
Modern psychology rediscovered Tocqueville's insight under the heading of “pluralistic ignorance.” In extreme cases, nobody believes in the truth of a certain proposition but everybody believes that everybody else believes it. One could imagine, for example, a society in which everybody holds Assurance Game preferences, but believes that everybody else holds Prisoner's Dilemma preferences (Chapter 18).4 Nobody would cooperate, and each would take the non-cooperation of others as confirmation of their beliefs. Society would be stuck in a bad equilibrium. In more realistic cases, most people do not believe the proposition but believe that most people do. Thus in Israel almost 60 percent agree to Palestinian autonomy in the territories, but only 30 percent realize that this is the majority position. This state of affairs would obviously tend to discourage advocates for Palestinian autonomy, and to maintain a status quo that the majority would like to change. According to joint Israeli–Palestinian opinion polls, a majority of both populations believes that a return to the 1967 borders is the preferred solution. I conjecture that the majorities falsely believe that only a minority on the other side holds this view. If this is so, that discrepancy would also have a chilling effect on the prospects for a durable peace.
These cases differ from the pathological situations in which everybody publicly professes a certain belief while knowing that nobody actually holds it in private. Communism displayed this culture of hypocrisy to an extreme degree, at least in its final gerontocratic stage. Pluralistic ignorance and cultures of hypocrisy can be sustained by the same mechanism, namely, fear of disapproval or punishment for stating deviant views. The difference is that in pluralistic ignorance, the disapproval is horizontal – meted out by fellow citizens who falsely believe they have to ostracize deviants lest they themselves be ostracized. By contrast, the culture of hypocrisy works by vertically imposed punishment: those who do not express enthusiasm for fulfilling the plan or hatred of the class enemy are likely to lose their jobs or worse. The vertical punishment may then induce horizontal measures, if people avoid or punish deviants lest they be punished as deviants themselves. Thus in a description of the Great Soviet Purge of 1937, we read that “Boris Pasternak was in trouble for not signing a collective request from well-known writers for the execution of Kamenev and Zinoviev; the prose writer Iurii Olesha was in trouble for defending Pasternak.”
Pluralistic ignorance also differs from the mechanism underlying the passive-bystander syndrome observed in the “Kitty Genovese” killing. In that case, each individual believed that the passivity of others justified his or her own. The cause cannot have been social pressure or a desire to conform to group norms, since the thirty-eight bystanders were too isolated from each other to form a community. Rather, the passivity seemed justified by an inference: since nobody else seemed to be doing anything, the situation could not be very serious. The “raw data” (her cries) were overwhelmed by this inference. We shall look more closely at this mechanism shortly. Here I only want to note that the situation did not involve pluralistic ignorance, since there was no discrepancy between what each person privately believed and the beliefs he or she imputed to others.
The culture of drinking has been shown to illustrate pluralistic ignorance. On many American campuses, there is a culture of heavy drinking among undergraduates, especially male. Most students do not feel comfortable with the heavy levels of drinking but go along because they believe, wrongly, that most others do.5 Their drinking behavior conforms to what they wrongly believe to be the typical attitude on campus rather than to their private attitudes. Another example can be taken from an experiment in which students were told to read an article written in a deliberately obtuse style that made it virtually incomprehensible, and then asked how well they had understood it and how well they thought others had understood it. In one condition, the students had the option of seeking out the experimenter and asking for assistance; in another they were expressly told they could not do so. Even in the former condition, no students went to see the experimenter because the procedure for doing so required that they risk embarrassing themselves. Each student seems to have believed, however, that whereas he or she stayed put out of fear of embarrassment, others did so because they understood the article and needed no help. Hence students in that condition tended to believe that others had understood the article better than they had themselves. The difference disappeared in the other condition. Conjecturally, this effect might be due to an “older sibling syndrome.” As noted in Chapter 17, we are all aware of our own inner anguishes and fears, but since we do not have direct access to the inner life of others, we tend to see them as more mature and self-possessed.
In the study of drinking on campus, it was also found that over time private attitudes, beliefs about the attitudes of others, and behavior moved into line with one another, raising the question of the stability of pluralistic ignorance. There are in fact two ways in which it might disappear: by the false beliefs about others becoming true or by people ceasing to hold them. If each person adopts the belief he or she (falsely) imputed to others, that imputation would in fact become true. This would most likely happen by dissonance reduction, caused either by the discomfort of disagreeing with the majority or the discomfort of saying one thing and believing another. This seems to be what happened with drinking on campus.
On the other hand, the situation might unravel.6 Suppose that 20 percent of group members show in their behavior that they do not hold the belief in question, and that the remaining 80 percent pay lip service to it because they require more than 20 percent of non-conformists in the group to become non-conformists themselves. Specifically, suppose that in a group of a hundred, there are twenty non-conformists, ten who would be willing to “come out” if at least twenty-five have already done so, fifteen who would do so if at least thirty-five have, and fifty-five who would join if at least fifty have shown their true colors. As stated, the majority culture is stable. Imagine, however, that five of the most conformist individuals leave or die and are replaced by five non-conformists. In that case, the majority would unravel. The twenty-five non-conformists would create the conditions for ten more to join them; the resulting thirty-five would attract fifteen more, thus generating the requisite threshold for the remaining fifty to join. Instead of referring to the process as the unraveling of conformism, we may also see it as the snowballing of non-conformism. We shall observe a similar dynamic in collective action (Chapter 23).
Conformism may unravel in many other ways. The little child in Andersen's tale is reflected in the line-matching experiments: when a single confederate stated the veridical opinion that D was the closest in length to B, the conformism all but disappeared. For another example, consider the widespread belief in both England and France prior to the Reformation that the king could heal scrofula by touching the sick person. The Reformation undermined this belief, since Catholics in France and Anglicans in England now were compelled to explain why the evidence in the other country was spurious. But recognizing the possibility of large-scale collective error turned out to be dangerous, since the allegedly invalid proofs used to support the belief in the other country were not very different from the ones invoked in one's own.
Another mechanism for unraveling is the publication of an opinion survey. Prior to the 1972 referendum on Norwegian entry into the Common Market (as it was called then), the government, the main opposition parties, and the major newspapers were all massively in favor of entry. Although, as the referendum showed, there was a popular majority against entry, each individual opponent would have been led to believe himself or herself a member of a small minority had not the opinion polls indicated otherwise. Without the polls, the outcome of the referendum would in all likelihood have been different. Some of those opposed to entry would have abstained from voting, since the outcome would have been seen as a foregone conclusion. Also, the movement that was formed to persuade the undecided would have remained small and uninfluential. In the period between the introduction of universal suffrage and the rise of opinion surveys, the scope for pluralistic ignorance about political matters must have been considerable.
Elections, too, may reveal a state of pluralistic ignorance. Before the first semi-free elections that were agreed upon in the Round Table Talks between the Communist Party and the opposition in Poland in 1989, most people on both sides of the negotiations as well as foreign observers believed that the Communists would obtain enough votes to allow them to stay in power. Although a majority of the population was opposed to the regime, they may have believed that they formed a minority. In fact, it is unlikely that the Communists would have agreed to the institutional reforms had they not been confident of sufficient support.7 On June 4, the opposition swept the elections, the regime collapsed, and other countries in the region followed suit. I return to some aspect of this snowballing process, within and across countries, in Chapter 23.
Rumors, fears, and hopes
Another tale by Hans Christian Andersen, “There Is No Doubt About It,” illustrates how “one little feather may easily grow into five hens” through successive exaggerations. The study of rumor formation and propagation is not, to my knowledge, very far advanced. With some exceptions, psychologists have not made much progress on the issue, partly because laboratory studies cannot create the tense and dense atmosphere in which rumors are born and spread. To illustrate, let me quote a letter from 1798 about the rumors of a conspiracy in England to assist the French: “I am quite disgusted and discouraged at the difficulty of knowing what is the truth of facts that are even passing almost under one's own eyes. I seldom venture to report serious reports at more than second hand at the farthest – but now I find that even too far & that little less is safe than the testimony of your own eyes – nor do I know that even they ought to be trusted in the first moments of fear acting upon prejudice.” The last four words offer a remarkable insight, but hardly one that could be confirmed in the laboratory.
Also, in my opinion, psychologists rely too heavily on speculative ideas about the function of rumors. While rumors may indeed serve the need of finding meaning and order in the universe (Chapter 9), other alleged functions seem more doubtful. Economists tend to view rumors as “rational herding” or “informational cascades.” While this approach may sometimes be appropriate, there is little doubt that most rumors have an irrational component. Hence I shall rely mostly on the work by historians, notably on the pathbreaking study of the “Great Fear” of 1789 by Georges Lefebvre.
I shall not try to define rumors, except to note that they are not necessarily false, but, when true, true only by accident. Because of their causal origin in hopes and fears, they cannot be characterized as “justified true beliefs.” I shall classify them into two categories, optimistic and pessimistic, related to, respectively, wishful and counterwishful thinking.
Optimistic rumors
rumors in the French countryside in the spring and summer of 1789 that the convocation of the Estates-General implied that Louis XVI had already decided to satisfy the demands of the people;
rumors among Virginian slaves in 1829 that the state constitutional convention had their liberation as its main goal;
rumors among Galician peasants in 1845 “that their final emancipation was impending, and that, indeed, a decree freeing them had been signed by the emperor but filched by the gentry”;8
rumors among Napoleon's followers about his return after each of his two defeats in 1814 and 1815;
rumors after the French revolutions of 1830 and 1848 about impending tax reductions;
rumors among the opponents of Napoleon III about his impending fall after an assassination attempt in 1858;
rumors that tens of thousands of Russian soldiers had joined the Allied troops in August 1914;
rumors in America after Pearl Harbor that the Japanese had oil and food reserves for six months only, and that revolutions in Japan and Germany were imminent;
rumors about rising stock values (“irrational exuberance”).
Pessimistic rumors
rumors in France under the ancien régime about famines created by speculators;
rumors in France under the ancien régime about the creation of a tax on children;
rumors in the French countryside in the spring and summer of 1789 about roving “brigands” who were out to cut the grain before it was ripe (“the Great Fear”);
rumors during the War of 1812 that a slave revolt had erupted in the District of Columbia and Maryland;
rumors among Napoleon's opponents about his return after his defeat in 1815;
rumors about a socialist leveling in the wake of the 1848 Revolution;
rumors about a massive invasion of Germany in March 1848 by impoverished French workers, plundering, burning, and killing;
rumors among the partisans of Napoleon III about his impending fall after an assassination attempt in 1858;
panics in financial markets;
rumors in India in 1935 about an impending earthquake;
German rumors about francs-tireurs shooting at German soldiers from the rooftops when Germany invaded Belgium in 1914;
Belgian rumors about German atrocities when Germany invaded Belgium in 1914;
rumors in Moscow in the 1930s that gangs of former kulaks threw rubbish, nails, wire, and broken glass into the food to cripple workers who ate it;
American rumors in September 1942 that crab meat packed by the Japanese contained ground glass.
The coexistence of optimistic and pessimistic rumors with regard to the same event – the return of Napoleon I, the fall of Napoleon III – is striking. It is worth asking whether these might mutually fuel each other.
Although the impression from these enumerations that pessimistic rumors are the more frequent could be misleading, a systematic analysis confirms this bias. A study of 1,089 war-related rumors gathered in the United States in September 1942 found that 65 percent had their origin in anger, 25 percent in fear, and only 2 percent in hope. Rumors in the first category did not target the enemy, but allies or domestic groups. They included the rumors that Churchill blackmailed Roosevelt into provoking a war with Japan, that the British sabotaged their own ships in American ports so that they would not have to put out to sea, and that American Catholics were trying to evade the draft.9 Since both the anger-inspired and the fear-inspired rumors were intrinsically unpleasant, their preponderance is surprising. Why do we so easily believe the worst?10
To this quantitative symmetry between optimistic and pessimistic rumors we can add a qualitative asymmetry: rumors inspired by hope are less likely to be used as premises for action. With the important exception of rumors in financial markets, optimistic rumors do not seem to affect behavior. To my knowledge, none of those who hoped for and believed in Napoleon's return left their work to join up with him. By contrast, those who feared and believed in this event took their precautions, such as cutting the grain before it was ripe, hiding their valuables, getting married to avoid being drafted, or withdrawing their deposits from the savings bank. The unfounded rumors of a slave revolt during the War of 1812 caused many members of the militia to flee to protect their homes, thus contributing to the American defeat. In a striking study of the impact of rumors linking contraceptives with cancer and other serious diseases in the Dominican Republic, unfavorable rumors from trusted sources encouraged many people to discontinue contraception, but rumors favorable to contraception did not increase the likelihood using it. I find these two asymmetries puzzling. In individuals, wishful thinking is both common and, when it occurs, often used as a premise for behavior. I do not know why collective wishful thinking tends to be neither.
The dynamics of rumors can illuminate some aspects of this problem. I shall discuss in turn the origin, propagation, and amplification of rumors.
The origin of rumors is facilitated by the universal tendency to explain events in terms of intentional agency or objective teleology (Chapter 9). In particular times and places, the formation of rumors can also be facilitated by a preexisting prejudice, schema, or script. In France around 1789, they involved roving bandits, callous grain speculators, a hated seigneurial class, and an oppressive tax collector. A small village may have been more favorable to rumor formation than a larger agglomeration, where skeptics might provide a reality check. Lack of education, too, might favor credulity. In Germany in 1914, the script of francs-tireurs in the Franco-Prussian War 1870–1 was easily triggered.
These are, however, only facilitating conditions. It remains to ask, as Lefebvre did: “How did one take the step from suspicion to affirmation?” His answer, which relies on the presence of “some individuals bolder than others,” is sketchy, but we can flesh it out with examples from his work. Triggers could be accidental or deliberate. The former included the reflection of a setting sun in a window mistaken for a fire, a misunderstood remark overheard in passing, a joke taken too seriously, and movements of animals mistaken for those of men.11 The latter included a hawker offering sensationalist rumors to attract buyers, an egocentric trying to make himself important, or somebody just trying to stir up trouble.12 As these examples indicate, the prospects for a “general theory of rumor formation” are dim. Yet the fact that the Great Fear, with no foundation in reality, erupted simultaneously and independently in seven parts of France shows that some ill-understood systemic forces were at work.13
A case of a state-initiated rumor was the decree issued by the French contrôleur-général Orry in 1745, when, as part of a larger statistical effort, he instructed local officials to spread rumors (semer des bruits) about an impending tax increase and an impending conscription. The texts to which I have access do not allow me to determine the intention behind this request or its effects. Orry may have been testing the waters, to determine whether these reforms would have been acceptable to the population, since he also asked the officials to assess the resources that might be found to increase the royal revenue. In other words, he might have tried to determine whether these increases were both objectively feasible and subjectively acceptable. Leaders of the Soviet Union, who had an extreme fear of popular unrest, also initiated rumors to test out public reactions to contemplated measures, such as deflation or dismissals.
The propagation and confirmation of rumors take many forms. Word of mouth is probably the main mechanism, at least until the arrival of the internet. Montaigne offered what may have been the first analysis of the mechanisms of rumor transmission:
The distance is greater from nothing to the minutest thing than it is from the minutest thing to the biggest. Now when the first people who drank their fill from the original oddity come to spread their tale abroad, they can tell by the opposition which they arouse what it is that others find it difficult to accept; they then stop up the chinks with some false piece of oakum … At first the individual error creates the public one: then, in its turn, the public error creates the individual one. And so it passes from hand to hand, the whole fabric is padded out and reshaped, so that the most far-off witness is better informed about it than the closest one, and the last to be told more convinced than the first. It is a natural progression. For whoever believes anything reckons that it is a work of charity to convince someone else of it; and to do this he is not afraid to add, out of his own invention, whatever his story needs to overcome the resistance.
This mechanism turns on the need for the rumormonger to preempt skepticism by filling in missing details of the story. In other cases, the recipient of rumor may be reluctant to express skepticism. Some individuals may be motivated to accept accounts of the danger lest they be accused of cowardice. To display incredulity during the Great Fear was both to invite accusations of serving the counterrevolution by putting the people to sleep in the midst of danger and to risk offending the amour-propre of those who called the alarm. Concerning the rumors of Belgian francs-tireurs, their accuracy appeared incontrovertible once they had been used as a premise for bloody reprisals. How else, in fact, could the German atrocities be justified? (Recall Seneca: “Those whom they injure, they also hate.”) And when rumors were passed on by wounded soldiers sent home from the front, who would dare to contradict them?
Montaigne's argument assumes that belief in a rumor is a condition for passing it on. Field observations and experiments suggest, however, that this need not be the case. In a study of the transmission of rumors in a neighborhood community, it was found that five of seven individuals who believed the rumor relayed it, and that nine of twenty-three who did not believe it also passed it on. In an experimental study of “revenge rumors” in the workplace, it was found that more credible rumors were not transmitted more frequently. When reflecting on the rationality of rumors, therefore, we should distinguish between the rationality of the belief content of the rumor and the rationality of the strategic use of rumors. In American presidential campaigns, cooked-up rumors to destroy an opponent are common. Franklin Delano Roosevelt survived the rumor that he was suffering not from infantile paralysis, but from syphilis, and Barack Obama the rumor that he was a Muslim or not an American citizen. However, the “swiftboating” rumor may have contributed to the defeat of John Kerry, and Michael Dukakis may have lost the election partly from the rumor that he had seen a psychotherapist (Ronald Reagan referred to him as “an invalid”). While those who transmitted these rumors may not have believed them, they certainly tried to make the “end users” – the electorate – not only believe them, but use them as premises for their votes.14
Lefebvre summarized part of his explanation of the Great Fear by saying that “the people scared itself” (le peuple se faisait peur à lui-meme). The rumor that brigands were approaching caused the mobilization of troops, which other peasants from a distance mistook for brigands. When a village rang the church bells, the detachments sent out by neighboring villages were mistaken for enemies. In 1848, a warning shot of a cannon in one French village was interpreted in neighboring villages as the din of battle. When rumors of an impending invasion of French paupers reached Germany in March 1848, road workers on the French side of the Rhine crossed the river in a hurry to return to their homes and families. Others, watching from a distance, may have thought they were the French approaching.
The amplification of rumors is well documented. Hans Christian Andersen's story of how one feather turned into five hens does not exaggerate the magnifying effect of rumor. After the insurrection of workers in Paris in June 1848, two men who were observed sitting by the side of a country road in Normandy became ten, three hundred, six hundred in the telling and retelling, until finally one could hear that three thousand “levelers” (partageux) were looting, burning, and massacring. Thirty thousand soldiers were sent out to counter the threat. An investigation revealed that one of the two was mentally ill and that the other was his father, who was in charge of him. In the same period, a peasant invented a fantasy to scare a child; soon thereafter more than a thousand men were in arms to defeat the non-existent “brigands.” One mechanism behind such exaggerations is the widespread tendency to dramatize: of two narratives that fit the facts more or less equally well, many people will choose the one that is highest in drama content. Another is that nuances in probability assessments and danger assessments easily get lost in the retelling and perhaps that people round up rather than down.
Informational cascades
Rumors can also arise by entirely rational belief formation, through a mechanism known as “informational cascades.” Suppose that each individual in a group has access to some private information about some matter. All form their beliefs sequentially, each one relying on his or her private information and on the beliefs expressed by their predecessors (if any) in the sequence. Each villager, for instance, might have some private evidence about the presence of brigands in the vicinity and use it, together with what he has heard from others, to form the opinion he then passes on. Voting by roll call can have the same dynamic, if the matter at hand turns only on beliefs about factual issues and not on preferences. Each member of an assembly will rely not only on her own information but also on what is revealed by the vote of those who precede her in the roll call. For a third example, consider a journal referee for a paper who learns that (but not why) a referee for another journal had favored rejection.
In these situations people use the conclusions of the belief formation process of others as indirect inputs to their own belief formation, without knowing the direct inputs (the private information) others used to form their conclusions. It may then happen that rational individuals end up with false beliefs, although they would have reached the correct conclusion had each of them had access to the “raw data” of their predecessors and not only to their conclusions. In the referee example, the second reader might, if he had read the first report, have spotted bias or faulty reasoning. Yet if he only knows the conclusion of the first report and the fact that the first journal is highly respectable, he should rationally take into account the negative opinion of the first referee together with his own assessment. If the latter is favorable but only slightly so, he may end up recommending rejection. A third reviewer with a strongly favorable personal opinion might also (rationally) favor rejection if she learns that two previous referees did so. Yet the outcome may be suboptimal (relative to the aims of the scholarly community), since the second and third referees favored publication and the first may have been only mildly against it.15 If the reviewers had read the article in the inverse order, the conclusion would have been different (“path-dependence”).
Bibliographical note
The line-matching experiments, first carried out by Solomon Asch, are described in any textbook on social psychology, for example, E. Aronson, The Social Animal, 9th edn (New York: Freeman, 2003). The moving-light experiment is described in R. C. Jacobs and D. T. Campbell, “The perpetuation of an arbitrary tradition through several generations of laboratory microculture,” Journal of Abnormal and Social Psychology 62 (1961), 649–58. The data on Israeli and Palestinian preferences for a return to the 1967 borders are from J. Shamir and M. Shamir, “Pluralistic ignorance across issues and over time,” Public Opinion Quarterly 61 (1997), 227–60. The reference to the pressures on Pasternak is from S. Fitzpatrick, Everyday Stalinism (University of Chicago Press, 1999), p. 198. For drinking on campus, see D. A. Prentice and D. T. Miller, “Pluralistic ignorance and alcohol use on campus: some consequences of misperceiving the social norm,” Journal of Personality and Social Psychology 64 (1993), 243–56. For the students exposed to the incomprehensible article, see D. T. Miller and C. McFarland, “Pluralistic ignorance: when similarity is interpreted as dissimilarity,” Journal of Personality and Social Psychology 53 (1987), 298–305. The unraveling scenario relies on T. Kuran, Private Truths, Public Lies (Cambridge, MA: Harvard University Press, 1995). The observation about the effect of the Reformation on the belief in the king's power to heal is due to M. Bloch, Les rois thaumaturges (Paris: Armand Colin, 1961). The letter (by Hannah Greg) citing the effects of “fear acting upon prejudice” in rumor formation is cited from J. Uglow, In These Times (London: Faber and Faber, 2014), p. 229. The main studies of rumor formation on which I have drawn are G. Lefebvre, La grande peur de 1789 (Paris: Armand Colin, 1988); F. Ploux, De bouche à oreille: naissance et propagation des rumeurs dans la France du XIXe siècle (Paris: Aubier, 2003); R. Cenevali, “The ‘false French alarm’: revolutionary panic in Beden, 1848,” Central European History 18 (1985), 119–42; M. Bloch, “Réflexions d'un historien sur les fausses nouvelles de guerre,” Revue de synthèse historique 33 (1921), 13–35; and C. Prochasson and A. Rasmussen (eds.), Vrai et faux dans la Grande Guerre (Paris: Éditions La Découverte, 2004). The rumors among the Galician peasantry are mentioned in L. Namier, 1848: The Revolution of the Intellectuals (Oxford University Press, 1946), p. 15. A detailed catalogue and analysis of rumors in ethnic riots are found in D. Horowitz, The Deadly Ethnic Riot (Berkeley: University of California Press, 2001), pp. 76–8. For rumors in the stock market, see A. M. Rose, “Rumor in the stock market,” Public Opinion Quarterly 15 (1951), 61–86. The study of rumors during World War II is R. Knapp, “A psychology of rumor,” Public Opinion Quarterly 8 (1944), 22–37. The asymmetry between “bad-luck” and “good-luck” superstitions is noted in D. Hand, The Improbability Principle (New York: Scientific American, 2014), p. 248. The claim that propaganda may aim at convincing part of the enemy population that it is treated worse than others is from D. Krech and R. Crutchfield, Theory and Problems of Social Psychology (New York: McGraw-Hill, 1948), p. 411, cited after F. Heider, The Psychology of Interpersonal Relations (Hillsdale NJ: Lawrence Erlbaum, 1958), p. 289. The rumor about kulaks putting broken glass in the food of workers is reported in S. Fitzpatrick, Everyday Stalinism (University of Chicago Press, 1999), p. 45. An exceptionally detailed catalogue of rumors in India following the 1934 earthquake is J. Prasad, “A comparative study of rumours and reports on earthquakes,” British Journal of Psychology 41 (1950), 129–44. The comments on state-initiated rumors in the Soviet Union draw on the entry “Rumors” in I. Zempsov, An Encyclopedia of Soviet Life (New Brunswick, NJ: Transaction Publishers, 2001). The comments on the French decree of 1745 draw on F. de Dainville, “Un dénombrement inédit au XVIIIe siècle: l'enquête du Contrôleur général Orry – 1745,” Population 7 (1952), 49–68. The study of the transmission of beliefs in a neighborhood community is L. Festinger et al., “A study of a rumor: its origin and spread,” Human Relations 1 (1948), 464–86. The study of revenge rumors is P. Bordia et al., “Rumor as revenge in the workplace,” Group & Organization Management 39 (2104), 363–88. An introduction to the mechanism of informational cascades is S. Bikchandani, D. Hirshleifer, and I. Welch, “Learning from the behavior of others: conformity, fads, and informational cascades,” Journal of Economic Perspectives 12 (1998), 151–70.
1 In Chapter 5 I discussed how observing others can trigger behavior similar to theirs by the quasi-moral norm of fairness, whereas being observed by others may trigger similar behavior through the fear of disapproval. In belief formation, observation by others can also produce conformity through fear of disapproval, whereas conformity produced by the observation of others occurs either by learning or by dissonance reduction.
2 In the numerical example used to illustrate Bayesian learning in Chapter 13, each new piece of confirming evidence brings about a smaller increase in probability than did the previous one. This “decreasing marginal value of new information” is a quite general phenomenon.
3 Actually the moving-light experiment was doubly public, leaving scope for insincerity. Yet the ambiguous nature of the task presumably facilitated sincere, or quasi-sincere, adoption of the exaggerated belief. I assume that the reason for the procedure was that the study of successive generations would have been hard to do in a singly public condition.
4 Tocqueville asserts that relations among occupational groups in the French ancien régime had this structure: “Our forefathers lacked the word individualism, which we have forged for our own use, because in their day there was no individual who did not belong to a group and who could regard himself as absolutely isolated. But each of the myriad small groups that made up French society had no thought for any group other than itself. There was, if I may put it this way, a sort of collective individualism, which prepared people for the true individualism that we have come to know … Had anyone been able to plumb their inner depths, moreover, he would have discovered that these same people, being quite similar to one another, regarded the flimsy barriers that divided them as contrary to both the public interest and common sense and, in theory at any rate, already worshiped unity. Each clung to his own condition because others distinguished themselves by theirs, yet all were prepared to merge into a single mass provided no one stood out in any way or was elevated above the common level.”
5 It may be true, however, both that most students do not drink and that most friends of most students drink, if those who drink have more friends than those who do not.
6 The extinction of false beliefs in the moving-light experiment is also a form of unraveling, due to the fact that in each generation subjects use their own “raw data” to adjust the estimated distance somewhat downward compared to what they hear from others.
7 I conjecture that the Ministry of the Interior conducted opinion polls that confirmed this belief. Yet in a semi-totalitarian society citizens may be reluctant to express their true opinions, even if assured of anonymity.
8 By contrast to these three cases of unfounded optimism, the “If only the King knew” syndrome, implying a conflict between the good ruler and his evil advisers, did not exist under Stalin, at least not in the 1930s. An historian of the period writes that “No rumors are known to have circulated in the villages concerning Stalin's benevolent intentions or his role in rescuing mice from the predatory behavior of little cats. Instead, the peasant mice seem to have stuck to their view that a cat is a cat, only some cats are bigger and more dangerous than others.” Rumors, too, are somewhat subject to reality constraints (Chapter 7).
9 I do not know whether these rumors arose spontaneously or were encouraged by the enemy. The second idea is consistent with the statement by two social psychologists that “One of the deadliest weapons in the arsenal of psychological warfare is propaganda aimed at convincing some segments of the enemy group that they are suffering more hardships or are gaining fewer benefits than other members of that group.”
10 One author affirms that “more superstitions seem to be associated with bad luck than with good” and suggests that the asymmetry may be “an evolutionary consequence of the need for caution: you're more likely to survive if you can spot potential threats.” The asymmetry is indeed puzzling, but I am not sure that the explanation is correct. The tendency to avoid spurious dangers, associated for instance with unlucky numbers, can hardly enhance survival. The author does not discuss the question I raise concerning rumors: are people more likely to act on superstitions associated with bad luck than on those associated with good luck?
11 In his study about false rumors in World War I, briefer but even more penetrating than Lefebvre's book, Marc Bloch refers to an incident in which the mishearing of the name of the hometown Bremen of a German prisoner of war as the name of the French town Braisne led to a rumor about a German spy who had been established in France since before the war.
12 As noted, I am not persuaded by Festinger's claim that people in India came up with rumors to justify their anxieties.
13 The rumor about Russian troops joining the Allies in August 1914 probably also originated simultaneously and independently in France and England, the only difference being that in France these troops were rumored to be in Marseilles, whereas the English thought they were in Scotland.
14 I do not know of a single instance in which a candidate's handlers have tried to generate a positive rumor about him or her. Believing the worst without evidence seems to come more naturally than believing the best without evidence.
15 Various conformity-reducing practices may be understood in this perspective. When (as used to be the case in Norway) the internal grader at university exams sends the student's answer to an external grader, he does not pass his own grade along. For the same reason, when seeking a second medical opinion, one should not tell the second doctor what the first said.