Social Communication and Cognition: A Very Preliminary and Highly Tentative Draft - Appendices

Cognitive Dissonance: Reexamining a Pivotal Theory in Psychology - Eddie Harmon-Jones 2019

Social Communication and Cognition: A Very Preliminary and Highly Tentative Draft

Leon Festinger

In order to understand and explain communication behavior in persons, it is necessary to separate various kinds of communication since the whole area of communication is probably as all inclusive as human behavior. We shall here concern ourselves with one conceptually defined subarea under the rubric of communication, which seems to be important. Specifically, we shall present a theory, together with some supporting data, to explain communication oriented toward acquiring or supporting one’s cognition. We shall state the theory in a series of hypotheses along with derivations which can be made from them.


Here we will deal with the ways in which persons acquire cognition. Much of what we say under this heading may seem trivial and well known, but it is necessary to state these things as precisely as possible in order to develop our theory coherently.

Hypothesis 1

There are two major sources of cognition, namely, own experience and communication from others.

Acquiring cognition through one’s own experience is, of course, the most direct way. We can think of the other source, that is, communication from others, as being an indirect way of acquiring cognition.

Frequently, cognition is acquired by a combination of both. Thus, for example, a child may learn from his mother that fire is dangerous and will hurt if touched. This cognition will undoubtedly be reinforced the first time the child gets burnt. One can raise the question of where, in this denotation of two sources of cognition, one would place such things as reading a newspaper, listening to a radio, hearing a political speech, or seeing a sign along a road that says “15 miles to Peoria.” These are all sources of cognition, none of them direct, but there seems to be a big difference among them. On logical grounds, according to our division, we would have to say that in each case the cognition that was acquired about the environment was acquired indirectly, through communication from others. Practically no one, however, will react to the traffic sign as anything other than fact, while many will react to what the politician tells them as quite different from fact. (We will ignore here the cognition acquired about the politician himself from listening to him. This particular cognition is of course acquired directly through experience.) We see then, that in the case of acquiring cognition indirectly, there are factors which will affect the impact which the communication has on the cognition of the recipient. We will deal with this more in detail later.

Hypothesis 2

The impact of direct experience will exert pressure on the cognition to conform to the experience.

In other words, there will be forces acting on the person to have his cognition correspond to reality as he experiences it. The result of this will be that, in general, persons will have a correct picture of the world around them in which they live. This is, of course, not surprising since the organism would have a hard time surviving if this were not the case.

Hypothesis 3

The strength of the impact of indirect experience (communication) to make the cognition conform will vary with the relationship between the communicator and recipient.

To make this hypothesis specific enough to be useful we must specify something more about the dimensions of relationship between the communicator and recipient which are relevant and the direction of the effect on impact of the communication. We will do this by stating two subsidiary hypotheses.

Hypothesis 3.1

The greater the “trustworthiness” of the communicator, the greater will be the impact on the cognition.

Trustworthiness here means a complex of things which, in the future, might better be separated. We will, however, spell out some of the things which affect it. To the extent that the communicator is seen as in the same situation as the recipient and consequently likely to experience things from the same point of view, he will be seen as trustworthy. Also, to the extent that the communicator is seen as performing an impartial service for the recipient, he will be seen as trustworthy. A road sign, for example, is seen as a communication emanating from an impartial servant and is hence regarded as trustworthy. A union member will regard a statement of fact or of opinion or of interpretation as more trustworthy if it comes from another union member (a person in similar circumstances) than if it comes from an executive of the company.

Hypothesis 3.2

The stronger the attraction on the recipient toward association with the communicator, the greater will be the impact of the communication on cognition.

This hypothesis is related closely to others which will be stated later, and so no detailed explanation will be given at this point.


It has frequently been stated that cognition steers behavior. This is quite true and is important for an understanding of the directedness of behavior in an organism. For an understanding of cognition formation and the communication processes which determine and result from cognition formation, it is also important to understand that behavior steers cognition. In the following hypotheses we will state our theory of how this takes place.

Hypothesis 4

There exists a tendency to make one’s cognition and one’s behavior consonant.

In order to explain this hypothesis, which is basic to the theory, we must spend some time in giving definitions of the terms we have used. First of all, although it may seem obvious, let us state more specifically the distinction between cognition and behavior. By cognition we wish to designate opinions, beliefs, values, knowledges, and the like about one’s environment, including oneself in this environment. By behavior we wish to designate actions of the person and reactions which he has. Actions would include driving a car, reading a book, being a man, residing in a certain place, and the like. Reactions would include things like being afraid, being hopeful, being anxious, and others.

It is also necessary to define consonance and the lack of consonance which we will call dissonance. There are three possible relations which can exist between items of behavior and items of cognition, namely, consonance, dissonance, and irrelevance. (These same three relations may also exist among items of cognition or among items of behavior. To avoid confusion we shall not deal with these at this point but shall return to them later.) A relationship of irrelevance exists if a particular item of cognition has absolutely nothing to do with a particular item of behavior. Thus, for example, an irrelevant relation exists between having the opinion that elementary schools are overcrowded and the behavior of playing golf on a nice sunny Saturday morning. Such irrelevant relations produce no pressures on persons and we may, for the rest of the paper, ignore them.

A relationship of consonance exists between a particular item of cognition and some item of behavior if, holding the motivation the same, this behavior would follow upon this cognition in the absence of other cognitions and in the absence of restraints. Thus, for example, the knowledge that construction crews are working on a certain street would be consonant with the behavior of taking an alternate route in driving to work. The opinion that it is going to rain would be consonant with the behavior of carrying a raincoat. The belief that thieves are around would be consonant with feeling afraid when walking home alone on a dark night.

A relationship of dissonance exists between an item of cognition and an item of behavior if, under the same conditions described in the paragraph above, a different behavior would follow upon this cognition. Thus, for example, the belief that it is going to rain is dissonant with the behavior of going on a picnic. The opinion that some other person is a very excellent and very careful driver would be dissonant with having a fear reaction while driving with him in the ordinary course of events.

One may, of course, raise the question as to why dissonances ever arise. There are many circumstances in which dissonances are almost unavoidable, and it will help our discussion later to list now the various ways in which dissonances can occur.

1. A change occurs in the situation. A given behavior may have been consonant with cognition before a change occurred in the situation. This new set of circumstances impinges upon the person’s cognition either directly or indirectly, and the new cognition is, at least momentarily, dissonant with the existing behavior.

2. Initial direct contact with a situation. A person’s cognition may have been formed from communication with others. The first direct experience may impinge on the cognition so as to produce dissonance with the existing behavior at least temporarily.

3. New communication from others. This can function in the same way as the above two to introduce a new cognitive item which is dissonant with existing behavior.

4. Simultaneous existence of various cognitive elements. It is probably a usual state of affairs that there are several relevant cognitive items, some of which are consonant with a given behavior and others of which are dissonant with that same behavior. Under such circumstances, it may not be possible for the person to find a behavior which eliminates all dissonances.

We can now return to an elaboration of Hypothesis 4. This hypothesis states then that, if a state of consonance exists it is an equilibrium, that is, no forces to change the relation are acting. If a dissonance exists, there will be forces set up to eliminate the dissonance and produce consonance. We will list the various ways in which these forces can act as subsidiary hypotheses.

Hypothesis 4.1

Given a dissonance between an item of cognition and an item of behavior, there will be a tendency to change the behavior so as to make it consonant with the cognition.

When this tendency is strong enough to produce actual changes in behavior, we will observe the kinds of things which have generally been treated as problems of learning or adaptation. For our present focus of interest, we will not elaborate on these problem areas but will rather concern ourselves with those situations where the tendency to change one’s behavior is not strong enough so that the existing behavior persists.

Hypothesis 4.2

Given a dissonance between an item of cognition and an item of behavior there will be a tendency to change the cognition so as to make it consonant with the behavior.

In general there are two ways in which the cognition can be changed, assuming we are dealing with persons who are in sufficient contact with reality, so that Hypotheses 1, 2, and 3 hold. One of these is to actually act on the environment so as to produce a situation where the veridical cognition will be consonant with the behavior in question. Of course, such action would only be successful in producing consonance in circumstances where the person has control over the environment. We will leave this without further elaboration since, again, it is not our main focus of interest.

Probably the major way in which cognition is changed so as to make it consonant with behavior is by selective exposure to either direct or indirect impact from the environment and actively seeking communications which will change the cognition in the desired manner. Thus, for example, a person who is afraid of riding in airplanes may avidly read and remember every account of an airplane disaster which he comes across and may avoid hearing about or reading about safety records and the like. Or let us imagine a person who has bought a new car just prior to the introduction of some new improvement. He may very actively try to persuade his friends that this new improvement is useless and will not work and adds unnecessarily to the expense of the car. If he succeeds in persuading them, he will then have support for a cognition consonant with his possession of a car which does not have this new improvement.

Hypothesis 5

If a consonance exists there will be resistance to changes in behavior or cognition which would introduce dissonance.

Hypothesis 6

If a dissonance exists there will be resistance to changes in behavior or cognition which would increase the magnitude of the dissonance.

To summarize the statements involved in Hypotheses 4 through 6, we may say that tendencies operate to avoid increases and produce decreases of dissonance. These tendencies, when equilibrium does not exist, will manifest themselves either in changes of behavior or in changes of cognition. In order to make the theory more usable, it is necessary to state some of the conditions which will determine whether the behavior or the cognition changes.

Hypothesis 7

Behavior or cognition will change in the presence of a dissonance, if the strength of the dissonance is greater than the resistance to change of either the behavior or the cognition in question.

Hypothesis 8

Whether the behavior or the cognition changes will be determined by which has the weakest resistance to change.

The last two hypotheses will have meaning only to the extent that we can specify resistance to change and the determinants of the strength of resistance to change of both behavior and cognition. We shall consequently proceed to do this.


We will state two hypotheses which are in essence an attempt to state some of the sources of resistance to change.

Hypothesis 9

The resistance to changing behavior which is dissonant with cognition will be directly related to the strength of the motivation which this behavior satisfies, the amount of effort or pain or loss involved in changing the behavior, and the number of cognitive elements which are consonant with the behavior.

Thus, for example, a person who has strong motivation toward having status and power might behave as if he had them, even though his cognition was dissonant with this behavior. The stronger the motivation, the more resistant would the behavior be to change. Under these circumstances, provided the cognition is less resistant to change, the person will find ways to change his cognition rather than change his behavior.

An example of another source of resistance might be a person who has recently bought a new car and later knowledge tends to make his cognition dissonant with this behavior. Changing the behavior, that is, selling the car and getting another, might involve financial loss and might also involve admitting that he had been foolish to have made the original purchase. There will, consequently, be a certain resistance to changing the behavior.

The last factor mentioned in the hypothesis is, of course, clear. If the behavior in question is already consonant with many cognitive elements, it will be much more resistant to change than if there are no consonances between cognitive elements and this item of behavior.

Hypothesis 10

The resistance to changing an element of cognition which is dissonant with some behavior will vary directly with the number of behavior items with which this cognitive element is consonant, with the importance of those behavior items, and with the strength of the impact of the environment, directly or indirectly, which supports this cognition.

The last of the factors listed in the above hypothesis is relatively clear. If the paper on which I am writing this is white, and I continue to see it as white, it will be difficult to change this cognition. Or if someone believes that modern art is decadent and all his associates tell him that this opinion is correct, it will be rather resistant to change.

Also, if an element of cognition is consonant with many items of behavior in which the person engages, it will be more resistant to change than an element of cognition which is not consonant with any behavior. The more important a particular behavior, the more resistance will there be to changing an element of cognition which is consonant with it. Importance of the behavior would depend upon the strength of the motivation which the behavior satisfies.

At this point we would like to digress slightly to deal with a concept which has been used frequently by others but which we have not mentioned, namely, consistency among cognitive elements. Thus, for example, various persons have maintained, and it seems plausible, that new cognitions are absorbed by persons so that they are consistent with what already exists in the cognition. This has been maintained about opinions, attitudes, perceptions, and the like. We have, however, not defined anything so far about relations among cognitive elements. Such definition is necessary, however. We will maintain here that any two or more cognitive elements which are consonant with the same behavior items and dissonant with the same behavior items are consonant with each other. Such a set of cognitive elements, which are consonant with one another, in this sense is a consistent cognitive system. The first part of Hypothesis 10, in light of this definition, could be stated as follows:

Corollary 10. The resistance to changing an element of cognition which is dissonant with some behavior will vary directly with the number and importance of the cognitive elements in the consistent system of which this particular element is a part.


Thus far we have dealt only with the relation of cognition to behavior which already exists. There are many situations in which a person has decided or is forced to do something, but there are still alternatives which are available to him. We will here state a number of hypotheses dealing with the relationship between cognition and, at the moment, nonexistent behavior.

For the purposes of the following discussion let us distinguish three kinds of situations with respect to behavior in which a person might find himself.

1. A given realm of behavior may be entirely irrelevant for a person. This would mean that he does not engage in any of these behaviors nor is there a likelihood that he may. Thus, for example, the whole realm of behavior related to cars may be completely irrelevant to a poor farm worker in India. We will refer to this as an irrelevant realm of behavior.

2. A person may not engage in any number of possible specific behaviors, but there is the possibility, likelihood, or even certainty that at some time in the future he will engage in at least one of them. Thus, for example, a person may have accepted a job in a different city from the one in which he now lives. This means that he is going to have to select a neighborhood in which to live, find a house to buy or an apartment to rent, and the like. All of these behaviors are ones in which, at some future time, he will engage in. We will refer to this as relevant future behavior.

3. The person engages in some behavior or has committed himself to some specific course of action. This is the kind of situation which we have been discussing above, and we will not dwell on it again. It is clear that this situation, which we will call relevant present behavior, and the situation of relevant future behavior may exist simultaneously. The person in the example above is in both of these situations. His decision to accept the new job has committed him to a specific course of action and is, hence, present relevant behavior. At the same time, it has involved him in a situation of relevant future behavior.

Hypothesis 11

There will be no active seeking out or active avoidance of cognition related to an irrelevant realm of behavior.

A person in this situation may be a passive recipient of such cognition, but there will be no initiative on his part. Thus, for example, most persons do not exert effort to find out or to avoid finding out how far from the earth the moon is.

Hypotheses 12

In the situation of relevant future behavior there will be active seeking out of cognition relevant to each of the possible future behaviors.

If there are a variety of possible behavior items, one or more of which the person may engage in in the future, this person will actively seek cognition relevant to each of them. Thus, for example, if a person has decided to buy a new car, but has not yet decided what kind to buy, he will actively seek information about each of those which he regards as possible purchases. Of course, once a decision has been made, this type of cognition seeking stops. From then on, the person will seek information consonant with his behavior and avoid information which is dissonant with it, as stated in Hypothesis 4. This partial reversal of information seeking behavior and the result of acquiring mainly consonant cognition after the decision results in what Lewin has called “freezing of decisions.”


There is surprising little data extant in the literature which is relevant to the above set of hypotheses. The data which we have been able to find is not always such as to be completely trustworthy. We will present it for what it is worth along with the statement of the major derivations from the hypotheses.

Derivation A

If a person’s action or reaction with respect to some event is dissonant with his cognition, he will communicate with others, the content of the communication being consonant with his action or reaction.

Derivation B

Such communications (the specific content) will be widespread if many persons have the same initial cognition and the same dissonant action or reaction.

There are data of a sort relevant to these derivations from two studies of rumors in India. Prasad (1950) systematically recorded rumors which were widely current immediately after the earthquake in the province of Bihar in India on January 15, 1934. The quake itself was a strong and prolonged one, felt over a wide geographical area. Actual damage was quite localized, and for a period of time communication with the damaged area was poor. We are, then, dealing with communication among persons who felt the shock of the quake but did not see any actual damage or destruction. While the study does not report anything about the specific reactions of these persons to the quake, it is probably plausible to assume that these persons, who knew nothing about earthquakes or their causes, had a very strong reaction of fear to the violent and prolonged quake which they felt. We will also assume that this reaction of being afraid persisted in them for some time after the shock was over. While the shock was going on, this fear reaction would be consonant with their cognition. But when the shock was over—the next day or even the day after that—when they could see no difference in anything around them, no destruction, no further threatening things, their cognition became dissonant with this reaction of fear which persisted. If this interpretation of the reactions of these persons is correct, then Derivation A would lead us to expect communication which would be consonant with fear reduction, namely communication, which would make it “appropriate” to be afraid. According to Derivation B, we would expect these communications to be widespread since many of the persons would have had the same fear reaction.

Actually, the vast majority of the rumors which Prasad recorded were what one might call “fear provoking” rumors. The following are a fair sample of these rumors as illustrations.

· The water of the River Ganges disappeared at the time of the earthquake, and people bathing were imbedded in the sand.

· There will be a severe cyclone at Patna between 18 and 19 January. (The earthquake was on January 15.)

· There will be a severe earthquake on the lunar eclipse day.

· A flood was rushing from the Nepal borders to Madhubani.

· 23 January 1934 will be a fatal day. Unforeseeable calamities will arise.

· There will be a Pralaya (total deluge and destruction) on 26 February 1934.

It is clear that a goodly number of rumors arose which predicted that more disasters were shortly to come. This cognition is, of course, consonant with the reaction of being afraid. The data, interpreted in this way, tend to support our derivations. This support is, however, weak because so much has to be read into the situation to make an assumption about the reaction of persons, and because there are many possible other explanations of why these particular kinds of rumors arose in such a situation. It is fortunate, however, that there is another study which can serve as sort of a control to compare with the one just discussed.

Sinha (1952) reports a careful collection of reports and rumors following a terrible disaster in Darjeeling, India. The author states, “There had been landslides before but nothing like this had ever happened. Loss of life and damage to property were heavy and extensive. . . . In the town itself houses collapsed and victims lay buried under the debris. . . . Over a hundred and fifty persons lost their lives in the district, about thirty of them in the town itself. Over a hundred were injured. More than 200 houses were damaged and over 2000 people were rendered homeless.” In other words, it was a disaster easily comparable to that of an earthquake. The author states, “There was a feeling of instability and uncertainty similar to that which followed the Great Indian Earthquake of 1934,” (p. 200).

We may then regard this aspect of the situation to be sufficiently comparable to allow us to compare the rumors which arose in this situation with those which arose following the 1934 earthquake. There is, however, one important difference which enables us to regard this study as a control. While the rumors following the earthquake were collected from persons who had no direct experience with the destruction, the rumors which Sinha reports were collected from persons in Darjeeling who did experience and see the destruction. In other words, we may again assume that the persons in this area had a strong fear reaction to the landslide all around them. However, there was evidence of the destruction, and consequently, their cognition was consonant with this fear reaction. In this situation then, we would not expect any rumors which predicted further disasters.

Actually there was a complete absence of rumors predicting further disasters. Some of the rumors represented slight exaggeration of the actual damage, and some rumors are even of the hopeful variety. The following is a selection of rumors to illustrate the general kind that existed.

· “Many houses have come down on the A-road.” (p. 201; Only one house had actually collapsed on this particular road.)

· Widespread belief that there had been a slight earthquake, which had helped in producing the damage. (There had been no earthquake.)

· “It has been announced that the water supply will be restored in a week.” (p. 203)

· “It will take months before the water supply is resumed.” (p. 203)

· “There are extensive floods in the plains. . . . Many bridges have been washed away.” (p. 204)

The remarkable thing about these rumors is the lack of serious exaggeration and even the presence of a few which are hopeful. The contrast with the rumors following the earthquake reported by Prasad is quite dramatic.

Since it seems that the two studies are comparable except for the fact that, in one instance, rumors were collected which circulated among persons not on the scene of destruction, and, in the other instance, the rumors circulated among persons on the scene of destruction, this difference in the nature of the rumors tends to give further support to the derivation.

Derivation C

If cognition is consonant with behavior, and events occur which would tend to make the cognition dissonant with the behavior, there will be communication whose content reaffirms the consonant cognition and denies the dissonant cognition.

There are data relevant to this derivation from a study by Sady (1948) of rumors among the Japanese in the relocation centers during the second World War. The data are not ideal for our purposes since, just as in the case with the previous studies cited, there is relatively little information given as to the actions and reactions of the people in the various specific situations in which the rumors arose. The result is that, in order to interpret the data with reference to this derivation, we are forced to make various guesses about the reactions of the persons involved. These data are, however, the best that we have been able to discover.

Because of this problem of guessing the reaction of the residents of the relocation centers, we shall deal only with rumors that circulated near the beginning of the establishment of the camps or just prior to the closing of the camps. In both of these instances we can be reasonably sure about some of the reactions of the residents. When they were first sent to the camp, they were very upset and anxious. They had been uprooted suddenly, and the major reaction, which persisted for some time, was fear and anxiety. Once in the camps, however, there was a tendency for cognition to become dissonant with the reaction of fear and anxiety. There were consequently many rumors which would tend to prevent this dissonance from arising. Rumors that many people were dying because of the heat and that their bodies were taken away secretly at night, or that the site for the relocation center had been deliberately chosen so that as many as possible of the evacuees would die, and many others of the same type were widely current. The following specific instance will illustrate the process.

During the first summer at the Poston Camp, temporary clinics operated before the regular hospital was ready. When the hospital was opened, these temporary clinics were closed, and a 24-hour home call service for emergencies was instituted. The change was, of course, an improvement in the medical services offered in the camp. This, according to our interpretation, would tend to introduce cognitions dissonant with the persisting reactions of fear and anxiety. In spite of (or perhaps to counteract) the fact of the introduction of the 24-hour home call service, the story circulated widely and was widely accepted that doctors would not make any more home calls. No matter how serious the case, the patient, they said, would have to go to the hospital before seeing a doctor. Thus, the change in medical services, through the rumor, was accepted into the cognition as a change for the worse and thus was consonant with their fear and anxiety.

Toward the end, when persons started to be resettled, the prevalent reactions, according to Sady, were again fear and anxiety, but this time about the problems of resettlement. There was apparently a considerable fear of how they would be treated by the communities on the outside. There were again quite a number of rumors which provided cognition consonant with this fear and which counteracted events which tended to produce dissonant cognitions. The following specific example will illustrate this.

The father and son of a family in one of the camps left to inspect their farm and to make arrangements for returning the entire family to their original home. A few hours after they left, the rumor spread that they had been beaten up on the way and that one of them had been taken to the hospital. Administration personnel from the camp contacted the father and son and discovered the story was false, that they had been given an excellent reception. This was made public but the rumor persisted. The father and son then returned to camp, and the whole family left for their farm. Several letters were received from them telling about the good treatment they were receiving. Nevertheless, rumors continued to circulate, such as that they were having difficulty shopping and had become discouraged about staying on their farm. Again the rumors served to counteract dissonant cognition and preserve the consonance with their reactions of fear.

There is one other, rather dramatic, illustration of this process of avoiding dissonance from this same study. During the war, some Nisei and some Issei requested repatriation to Japan, the Nisei renouncing their citizenship. While the majority of the Issei in the relocation centers believed and hoped that the war would end in a negotiated peace, most of those who had requested repatriation firmly believed that Japan would win the war and explained news of Japanese reverses as American propaganda. The renunciation of citizenship and the request for repatriation were rather irrevocable decisions, and at the time these decisions were made, the cognition of these persons was consonant with their decision. This group who had requested repatriation continued to believe that Japan had won the war even after the surrender. After having seen the newspapers and photographs attesting to the surrender of Japan, the great majority of the Japanese in the camps accepted the evidence. Those who had requested repatriation, however, held fast to their belief that Japan had won and continued to dismiss the evidence as American propaganda. This belief persisted all the way back to Japan on the boat. It was not until after landing in Japan that this belief was finally dispelled. The following Associated Press news story describes the situation.

Nippon Times, December 2, 1945

“Bitter Disappointment Marks Return Home of Nisei Who Wished They Had Stayed in U.S.”

Why 95% of those who came back to Japan on the ship with me thought that Japan had won the war: They thought it just a bunch of American propaganda that Japan surrendered and they believed that they were being brought back to Japan because Japanese had won the war and were compelling the Americans to transport them.

We will now discuss some data relevant to Hypotheses 11 and 12. Baxter (1951) reports a study in which a number of persons were interviewed periodically during the election campaign of 1948. Particular attention was paid to the collection of data concerning discussions with others about the election. The panel of respondents was interviewed in June 1948 for the first time. Among other things, they were asked whether they were doing anything for their party in the present election, how interested they were in that election and whether or not they had talked politics with anyone that month. Table A.1 presents the data based on these three questions.

TABLE A.1. Percentage of Respondents Who Talked Politics in June

% Talked politics

Total #

Doing something for party

High interest in election



Low interest in election



Doing nothing for party

High interest in election



Low interest in election



An examination of the figures makes it clear that those who are doing something for their party talk more about politics than those who are not doing anything. If doing something for the party in the election campaign can be viewed as a behavioral commitment, then the interpretation would seem indicated that, having committed themselves behaviorally, there are pressures to talk politics and so make their cognition consonant with this behavior. Those who have not committed themselves behaviorally, with the election still so far in the future, show much lower incidence of seeking or giving opinions concerning politics. The possible interpretation that those who are doing something for the party are simply more interested in the election is ruled out by the fact that there are very similar proportions of interested and uninterested persons in both classifications.

This interpretation is supported more strongly if we examine this data more closely. To the extent that high interest in the election can be interpreted as indicating how important the person thinks the election is for himself, we would expect that the highly interested persons would talk politics more frequently than the less interested persons. This is indeed true for those who are not doing anything for the party, in other words, those not behaviorally committed. For those who are behaviorally committed, however, there is no difference at all in the frequency with which the highly and less interested persons discuss politics. In fact there is a slight difference in the opposite direction. It is possible that those who are behaviorally committed, and yet have low interest in the election, talk as much or even more than the others in order to make their cognition consonant with their behavior of working for the party.

Table A.2 shows the percentage among those who did talk politics in June who talked frequently. It is clear that all the differences are in the same direction even to the reversal between the high and low interest groups who are committed to working for the party.

TABLE A.2. Percents of Those Talking Politics in June Who Talked Often

% Talking often

Total #

Doing something for the party

High interest



Low interest



Doing nothing for the party

High interest



Low interest



According to Hypotheses 11 and 12 we would, of course, explain the low percentages of persons talking politics in June among those highly interested but not working for the party by the fact that any decision (as to voting) was still a long way off. We would then expect that these percentages would increase enormously as the election drew near and the future implied behavior became salient. Table A.3 shows the data for the percent in the various classifications who talked politics in October, just prior to the election.

TABLE A.3. Percents Talking Politics in October

% Talking often

Total #

Doing something for the party

High interest



Low interest



Doing nothing for the party

High interest



Low interest



Again, all of the differences are in the same direction as they were in the previous tabulations, including the reversal between the interest groups who were working for the party. The only difference now is that the two high interest groups talk almost as much. In other words, the immediacy of the election has had the anticipated effect.

In the preceding three tables which we have discussed, we cannot, because of the small numbers of cases in the group who are not very interested in the election and are working for the party, be very confident about the tendency for them to talk even more than those who are highly interested in the election. We can, however, be quite confident that among those committed to working for the party, the factor of interest in the election makes no difference in their tendency to talk politics.

Let us now turn our attention to the kind of data which is more usually obtained in studies of public opinion and mass media. The hypotheses and derivations which we have stated should be relevant to this type of material. Unfortunately, once more, we run up against the obstacle that there are practically no studies in which sufficient data were gathered to make a test of these hypotheses unequivocal. Typically, the elements of data which are missing are those concerning the behavior and reactions of the persons, the degree of commitment to the specific behavior or reaction which exists. We will discuss below a selection of material where plausible assumptions can be made regarding these missing items of data and see to what extent the results fit the theory we have presented.

There are many instances in the literature of reported relationships between information about or awareness of some item and some other variable which can usually be called “interest in the matter.” Interest in something is, of course, a very vague term and does not usually refer to anything unambiguous. To the extent that interest is a measure of how important the problem is to the person, we would expect this relationship from our theory. That is, the more important an implied future behavior was, the stronger would be the tendency to acquire information relevant to it. Sometimes, however, the variable of “interest” seems to be interpretable as a reaction or behavior, and in such cases, the data should conform to the predictions from the theory. We will give a few examples of such instances. It must be remembered that in most cases the data simply relate two variables with no indication of the direction of causality. That is, it is possible that a relationship between interest and amount of information could be found because once the person acquires the information, for whatever reason this may have occurred, it stimulates his interest in the problem. Probably this kind of thing always happens to some extent. In our interpretation, however, we will not dwell on this direction of causality but shall explore how the data can be interpreted from the other causal direction.

We will first examine a number of items of data from a report by the Survey Research Center, University of Michigan (1948) of a survey concerning awareness of a recent Cancer Society campaign and attitudes toward cancer.

Respondents in this survey were asked which diseases they considered “most dangerous.” Table A.4 shows the relationship obtained between whether or not cancer was named as a “most dangerous disease” and the awareness of the cancer campaign.

TABLE A.4. Relation of Awareness of Cancer Campaign to Choice of Cancer as a “Most Dangerous” Disease


Cancer named

Cancer not named

Very high












Very low



Not ascertained





Before interpreting this relationship let us speculate briefly about the meaning it has for a person to name cancer as one of the “most dangerous diseases.” It will help in this speculation, of course, to know something about the reasons persons gave for feeling it was so dangerous. Actually, 74% of those who named cancer as one of the most dangerous diseases gave as a reason that it is incurable or that it is fatal. Let us imagine, then, that naming cancer as a most dangerous disease indicates some fear of cancer, or at least indicates the presence of an implied future behavior, namely, something must be done to avoid it. If this is true, then the relationship obtained with awareness of the campaign would be consistent with our theory. A campaign by a cancer society may be expected to provide information about things to do to prevent cancer and may also be expected to provide cognition consonant with a “fear of cancer.” We would then expect persons who are afraid of the disease or who have an implied future behavior to expose themselves to the campaign and hence be more highly aware of it than persons who are not afraid of it or have no implied future behavior. It can readily be seen that, interpreted in this way, such data support the theory. It can also readily be seen that an enormous amount of interpretation and conjecture is necessary in order to interpret the data at all. This, unfortunately, is true of almost all of the data in the literature.

Those data which we have had to infer or conjecture are precisely the ones that would have to be supplied to make a test of the theory rigorous. There is much material of this type which we could present and discuss, but there is little point in doing this since in all cases the same problems will obtain.


The theory which we have developed in the preceding pages has direct implications concerning who exposes himself to what. In other words, the theory predicts certain things about the composition of voluntary audiences. Specifically, it states the following things:

1. People with no present behavior or implied future behavior relevant to a given topic should not be found among purely voluntary audiences.

2. People who would expect to obtain cognition dissonant with present behavior should not be found in purely voluntary audiences.

3. Persons who expect to obtain cognition consonant with present behavior should be in voluntary audiences.

4. People with implied future behavior should expose themselves voluntarily irrespective of the bias of the communication.

Let us make clear, of course, that this does not refer to exposure for purposes of amusement or pleasure such as listening to a play or a dance band on the radio. That part of such a program that deals with propaganda, advertising, or educational material is communicating to an involuntary audience, that is, an audience that was lured there for some other purpose. By a purely voluntary audience, we mean one which exposed itself voluntarily to the particular material in question.

One can find frequent remarks in the literature that tend to confirm these implications of the theory. It is often stated by persons writing on the subject that people listen to things they already agree with and do not expose themselves to things they disagree with. However, in going through the literature one begins to wonder where they reached this conclusion because there is almost a complete absence of data concerning it. For example, Klapper (1949) states, “this phenomenon of self-selection might well be called the most basic process thus far established by research on the effects of mass media. Operative in regard to intellectual or aesthetic level of material, its political tenor, or any of a dozen other aspects, the process of self-selection works toward two manifestations of the same end: every product of mass media (1) attracts an audience which already prefers that particular type of material, and (2) fails to attract any significant number of persons who are either of contrary inclination or who have been hitherto uninterested.” He presents no data in support of this, however, except to quote Lazarsfeld (1942) as follows: “even so called educational programs are not free from this tendency. Some time ago there was a program on the air which showed in different installments how all the nationalities in this country have contributed to American culture. The purpose was to teach tolerance of other nationalities. The indications were, however, that the audience for each program consisted mainly of the national group which was currently being praised. There was little chance for the program to teach tolerance, because . . . self-selection . . . produced a body of listeners who heard only about the contributions of a country which they already approved.”

Lazarsfeld, in turn, presents no data to support this impression. The one instance he mentions does, of course, support our derivations. Certainly being a member of a specific nationality group in America is an irrevocable behavior. Cognitions that this nationality group is important in American culture would be consonant with being a member of the group. Consequently they listen to a broadcast which provides this consonant cognition.1


Baxter, D. (1951). Interpersonal contact and exposure to mass media during a presidential campaign (Unpublished doctoral dissertation). Columbia University, New York, NY.

Klapper, J. (1949, August). Effects of the mass media (A report to the director of the Public Library Inquiry). New York, NY: Columbia University, Bureau of Applied Social Research.

Lazarsfeld, P. (1942). Effects of radio on public opinion. In D. Waples (Ed.), Print, radio, and film in a democracy (pp. 114—158). Chicago, IL: University of Chicago Press.

Prasad, J. (1950). A comparative study of rumours and reports in earthquakes. British Journal of Psychology, 41, 129—144.

Sady, R. R. (1948). The function of rumors in relocation centers (Unpublished doctoral dissertation). University of Chicago, Chicago, IL.

Sinha, D. (1952). Behaviour in a catastrophic situation: A psychological study of reports and rumors. British Journal of Psychology, 43, 200—209.

Survey Research Center, University of Michigan. (1948, December). The American public discuss cancer and the American Cancer Society campaign: A national survey. Ann Arbor, MI: Author.


From Social Communication and Cognition: A Very Preliminary and Highly Tentative Draft (Unpublished manuscript), by L. Festinger, 1954. Copyright 1954 by Leon Festinger. Reprinted with permission.

1Festinger reported the results of a study by Childs in the original version of this paper. Because the editors could not find the study to which he referred, the discussion of this study was eliminated from the present version.