Conceptualizing Attitude Consistency

Although a great deal has been written about attitude consistency, little attention has been given to the theoretical foundations of the idea. In particular, remarkably little has been written about the problem of how to

choose the attitude questions that will be used in the analysis. There have been some ad hoc decisions to include or exclude variables, but there has not been any rigorous thinking on general rules for item selection. With the exception of Converse, who helped write the SRC questions in the 1950s studies, attitude consistency researchers have largely been limited to using whatever items SRC decided to put in the survey each time. Usually they have followed the "use all available items" approach to item selection.^[18] If the items in a consistency analysis make any difference--and obviously they do--there must be some theoretically justified basis for selecting them. I will now return to the foundations of consistency theory and try to see what that basis might be.

Since Fritz Heider (1944, 1946) first proposed balance theory, there has been a proliferation of closely related theories, which are collectively referred to as cognitive consistency theories. Because they are so similar, I focus on balance theory, emphasizing the elements of Heider's theory that are common to all consistency theories. More detailed reviews are available in Brown (1965), Kiesler et al. (1969), and McGuire (1966).

Balance theory deals with the relationships between a person and objects in the person's phenomenological world. Heider distinguished two basic types of relationships, which can be either positive or negative. The first type is that of liking or evaluating. Examples of such relationships include loving, approving, disliking, and hating. The second type is that of associating. Examples of associating include similarity, possession, causality, and proximity.

A balanced or consistent state is defined in terms of combinations of relations between a person and an object or among a person and two objects. In the case of a person and a single object, a balanced state exists when both evaluative and associative relations are either all positive or all negative. For instance, when one likes a public policy (positive evaluation) and it is enacted (positive association) or when one dislikes a policy and it is not enacted, one is in a balanced state. On the other hand, when one likes a policy that is not enacted or dislikes a policy that is, one is in an imbalanced state.

In the case of a person and two objects, a balanced state exists when all relations are positive or when two of the three relations are negative (see Figure 13). For instance, if one likes a presidential candidate, favors pro-environmental policies, and believes that the candidate also favors those

[18] Alternatively, this might be called the Mt. Everest method. Why use an item? Because it is there!

Figure 13.
Examples of Balanced and Imbalanced States

Note: + means positive association (e.g., liking, preferring).
- means negative association (e.g., disliking, opposing).

policies, one is balanced. Alternatively, if one opposes environmental policies, dislikes a candidate, and believes that the candidate favors those policies, one is again in balance. On the other hand, if one likes a candidate, opposes the environmental policies, but thinks that the candidate supports them, then an imbalanced state exists. In short, to be in a balanced or consistent state, one must like what one's friends like, dislike what one's friends dislike, like what one's enemies dislike, and so on.

According to Heider's model, balanced states are stable and resistant to change. Imbalanced states are unstable and are assumed to cause some sort of psychological tension that can be relieved only by changes among the relations toward a balanced state. Thus the psychological tension--referred to by other investigators as "stress," "discomfort," and "psychic pain"--is the dynamic force for attitude change.

The last central proposition of balance theory, which was pointed out by Osgood and Tannenbaum (1955) and Abelson and Rosenberg (1958), is what might be called the association condition: In order for an imbalanced set of attitudes to produce any tension, the individual must associate the various elements with one another. As long as the individual does not connect potentially conflicting ideas and perceptions with one another, there will be no conflict, and hence no tension and no attitude change. When an individual does bring conflicting ideas into association with one another, perhaps by simply thinking about them together, tension will be produced, and attitude change may result.

The association condition can be recognized as Converse's (1964, p. 208) dynamic case definition of constraint. When change in one attitude causes change in others, the attitudes are constrained. That, of course, is the association condition. So Converse's dynamic case definition is that attitudes are constrained when they are sufficiently closely associated with one another to make the attitude change dynamics of the consistency model relevant.

This also explains why consistency is a measure of political sophistication. People who know a good deal about politics, who think about the issues, and who understand the arguments that connect different issue positions--that is to say, people who are sophisticated--will be constrained. People who do not have the facts to connect the issues, or who do not think about politics so that they never get around to connecting the issues or who do not understand the arguments that connect the issues--that is to say, the unsophisticated--will not be constrained. Thus the theoretical foundations for Converse's work on attitude consistency stem from the work of Heider and others in cognitive consistency theory.

In his famous investigation of mass belief systems, Converse (1964) dealt with the nature of consistency, its sources, and a number of related characteristics. Although he established cognitive consistency theory as a basis for analyzing attitudes, his discussion was not very rigorous. He neither presented a set of theoretical propositions outlining his own version of consistency theory nor relied on any other published version of the theory. (In fact, Converse actually cited only a single consistency theory study [McGuire 1960], and he used it only to make a minor point about consistency stemming from logical relationships.) Instead, he presented a general discussion without treating any aspect of the underlying theory in detail. This vagueness caused many problems.

However, one can hardly blame Converse for not providing a detailed analysis. He did what he set out to do. He sought to assess the extent of ideological thinking in the mass public, not to forge a formal theory of belief systems. His analysis is a first attempt to extend consistency theory into the political sphere.^[19] He left for later investigators the problem of developing a more formal model. The problem is that no one fully developed that model.

I now turn to Converse's typology of the sources of constraint to show how his work and consistency theory fit together. Converse identified three sources: (1) logical—in the sense of formal logic, (2) psychological--in the sense of the rationality of a strong argument, and (3) social--in the sense of social pressures to accept certain opinions. Let me consider how we might define consistency for these three sources.

The nature of logical constraints is simple. A set of attitudes is consistent if it is consistent according to the rules of formal logic. That is, a set of propositions is consistent if both a proposition and its negation cannot be derived from the set (Mates 1972). For instance, it would be inconsistent to believe that (1) all men are mortal, (2) Socrates was a man, and (3) Socrates was not mortal.

As Converse pointed out, few political issues are subject to logical constraints. There are, however, exceptions. For instance, principles of civil liberties such as freedom of speech can be logically related to specific, concrete applications. If one favors absolute freedom of speech, then one should favor it for communists, fascists, and so forth. Most consistency

[19] This is a bit of an exaggeration. A few social psychologists examined political beliefs in their experiments. In addition, McPhee et al. (1962) examined attitude consistency a few years earlier. Their analysis is a more sophisticated, yet theoretically less useful analysis. So although Converse is not the first political scientist to examine attitude consistency, he is the first important one.

studies, of course, do not deal with examples of reasoning from the general to the specific, so logical constraints are rarely relevant.^[20]

Psychological constraints are more common among political issues, but harder to define rigorously. Heider's basic idea of bringing a set of evaluations and associations into balance applies, but because the sets to be balanced are sometimes long chains of reasoning, the problem is difficult. For instance, what do the principles of liberalism imply for our foreign policy in Nicaragua? Many arguments are possible, and evaluating them with the simple calculus Heider and his fellow consistency theorists offered is no easy task.

Coherent, persuasive arguments are like obscenity. We cannot define them precisely, but we know them when we see them. Although formulating a clear definition is probably impossible, one cannot criticize Converse on this score. After all, this imprecision in defining consistency is a well-known problem of nearly all consistency theories. Zajonc (1960, p. 285) provided an anecdote that illustrates the problem nicely: "Festinger once inquired in a jocular mood if it followed from balance theory that since he likes chicken, and since chickens like chicken feed, he must also like chicken feed or else experience the tension of imbalance."

There have been several efforts to develop formal methods for assessing arguments, but they have all been fairly narrow. The most important of these are Abelson and Rosenberg's (Abelson and Rosenberg 1958; Rosenberg and Abelson 1960). Although their "symbolic psycho-logic" is interesting, it is far more limiting than common notions about what constitutes a good argument. For instance, their model deals only with dichotomous variables. Seven point attitude scales do not work in their model. Thus, such efforts are of little use here. We must accept the obscenity standard.

Social constraints, which Converse maintained were more common than either logical or psychological constraints, are fairly similar to psychological constraints. The difference is that whereas psychological sources involve bringing a set of opinions into balance, social sources involve bringing both opinions and evaluations of people into balance.^[21]

[20] For an excellent discussion of the differences between "vertical" constraint (connections between general propositions and specific ones) and "horizontal" constraint (connections between two or more general propositions), see Conover and Feldman (1980).

[21] This is not what Converse said. I am restating his definitions in such a manner that they can be usefully related to cognitive consistency theory. The goal is to develop a basis for item selection within Converse's framework.

For instance, if an individual likes President Reagan, and the president advocates abolishing the Department of Education, then the individual should favor abolishing the department. As in the example, social constraints usually involve politicians or other influential public leaders.

Again, the definitions of balance and imbalance come directly from cognitive consistency theory (i.e., a set of three elements is balanced if all three relations are positive or if two are negative and the third is positive). Here, however, the chains of reasoning to be balanced are much shorter and easier to analyze. Why should a liberal oppose American involvement in Nicaragua? Because Ted Kennedy, Tip O'Neil, and other leading liberals oppose it.

In light of these definitions, I can now ask how one knows when a given pair of attitudes is consistent. Converse's answer is that to the extent that the attitudes are socially constrained, they will fall into a liberal-conservative ordering because that is how elites' beliefs are ordered. It follows that a pair of attitudes is consistent when both are liberal, both are moderate, or both are conservative.

In terms of balance theory, Converse's theory may be stated as follows: (1) Elites may be placed on the liberal-conservative dimension.^[22] (2) Elites advocate issue positions associated with their ideological positions. (3) The mass public learns issue positions from elites. (4) To the extent that individuals learn from elites, they will be consistent. Figure 14 illustrates the theory for the case of attitudes toward abortion and welfare spending. Liberal elites favor both abortion and spending more money on welfare (indicated by the pluses in the figure). Conservative elites oppose abortion and prefer spending less money on welfare. So to be consistent, the individual must adopt either the liberal or the conservative package of opinions.

The reasons Converse gave for emphasizing the social sources of constraint become clear when one considers this example. First, there obviously are no direct logical relationships between this pair of issues. That is, all combinations of attitudes toward abortion and welfare spending are logically consistent. Second, although there may be several good arguments (psycho-logical relationships) that connect the two attitudes, they are certainly not going to be obvious to many people. Few survey

[22] I am using the term "political elites" here in a broad sense. I refer to elected political officials, candidates, political columnists, and other political activists. In order to use this model as a basis for item selection, a researcher would have to develop a more precise definition. It is not important here.

Figure 14.
The Social Sources of Consistency

Note: + means positive association (e.g., preferring).
- means negative association (e.g., opposing).

respondents would be able to construct arguments connecting attitudes on those two issues. The social connections, however, are quite evident. Most liberals take one side, and most conservatives take the other on both issues.

Another way to think about this model is to consider the relationship between the two attitudes. The correlation between the attitudes is a spurious one. The attitude consistency model implies that correlations between attitudes are spuriously caused by outside forces--specifically, by political elites.

There are two important limitations on this model. First, it ignores other sources of consistency. Political elites are certainly the dominant sources of political messages, but they are not the only ones. There are psychological and occasionally logical sources of constraint. Moreover, there are other social sources of constraint. The Catholic church, for instance, currently opposes abortion and supports increased welfare spending. Devout Catholics who follow the church line would, therefore, be judged inconsistent.

The alternative sources of constraint provide alternative criteria with which to define and measure attitude consistency. For instance, one could investigate how consistently Catholics adhere to official church doctrine. Converse dismissed these alternative criteria in his discussion of idiosyncratic ideologies. Here I wish only to note that Converse's theory can be used to analyze many idiosyncratic ideologies just as well as it can be used to analyze mainstream ideologies.

A second limitation with the model is that it does not apply to all issues. All issues are not fought along liberal-conservative lines. An equally important point is that even among those issues that do have liberal and conservative positions, the strength with which the issue positions are associated with liberalism or conservatism may vary.

Although most controversial issues have liberal and conservative sides, some do not. Regional and sectional cleavages dominate several policy areas such as energy, water, and agriculture. As a case in point, the current farm crisis pits most senators and representatives from farm states against those from nonfarm states. Thus what to do about the rising tide of bankrupt farmers has niether a clear liberal nor a clear conservative position. Liberals and conservatives take both sides.

Some issue positions are more closely associated with liberalism and conservatism than others. That is, the extent to which an issue can be thought of as fitting into a liberal-conservative framework is a matter of degree. There are two ways an issue can fail to fit into the framework. First, differing mixes of liberal and conservative leaders can take opposing sides on the issues.^[23] The two extremes are (1) all liberals in one camp, all conservatives in the other and (2) half the liberals and half the conservatives in one camp and the other liberals and conservatives in the other camp. Thus the first problem is one of mixed messages, or item ambiguity. When political elites tell the public what the positions are, how clear or mixed is the message? In this case, the message is ambiguous to the extent that the sources are mixed. The less ambiguous the message (i.e., the less mixed the sources), the more easily the public will learn it.

Second, political leaders can vary the attention they give to an issue. By talking about an issue more and saying that it is important, politicians can go a long way toward teaching the public what they think. However, even if politicians disagree about an issue, they do not have to talk about it much. This is frequently the fate of more technical issues, for instance, what position to take in the trade-off between inflation and unemployment. Again, the clearer the message, the better the public will learn it. In this case, clarity is a function of how loud the message is, or how salient it is, rather than the extent to which it is mixed.

At this point, supporters (or former supporters) of The Changing American Voter should recognize that both of these variables--the ambi-

[23] For simplicity, I am assuming there are only two sides to every issue here. That is obviously not so, but it eases the explanation.

guity and salience of the messages--are environmental. Nie and his colleagues argued that parties became more distinct (i.e., the messages became less ambiguous) and the political rhetoric heated up (i.e., politics became more salient) in the early sixties. They used this argument to explain the 1960-1964 surge in consistency. Although this across the board surge did not happen, it does not follow that other, lesser changes did not occur; nor does it follow that the environmental causes of change that Nie et al. identified are not real. Of course the environment influences the public. What political leaders say has an impact. The impact is not as large as Nie et al. suggested. However, it is important for the problem of item selection. Nie et al. demonstrated that point when they showed that the size of government item ceased working in the late 1960s and early 1970s.

Few issues are completely unrelated to the liberal-conservative dimension, but conversely, few fit it exactly. Thus the practical problem for those who wish to measure attitude consistency is what these points reveal about item selection.