Book Search:  


Google full text of our books:

Studies in Public Opinion:
Attitudes, Nonattitudes, Measurement Error, and Change
Edited by Willem E. Saris & Paul M. Sniderman

Book Description | Endorsements

COPYRIGHT NOTICE: Published by Princeton University Press and copyrighted, © 2004, by Princeton University Press. All rights reserved. No part of this book may be reproduced in any form by any electronic or mechanical means (including photocopying, recording, or information storage and retrieval) without permission in writing from the publisher, except for reading and browsing via the World Wide Web. Users are not permitted to mount this file on any network servers. Follow links for Class Use and other Permissions. For more information, send e-mail to

This file is also available in Adobe Acrobat PDF format


Willem E. Saris and Paul M. Sniderman

IN THE STUDY of politics and political choice there are long-running streams of studies, proceeding along a common channel, with the later studies drawing out and testing the implications of the earlier. Broadly, these works have followed one of two opposing strategies--pushing one argument to its limit or pulling together apparently irreconcilable arguments. Mancur Olson's work on collective choice and the problem of the "free rider" exemplifies the first strategy;1 Philip Converse's seminal work on the nature of belief systems in mass publics exemplifies the second.2

The first of Converse's master themes is the now famous "nonattitudes" hypothesis. Very briefly, and therefore roughly, the idea is that large numbers of the public do not hold any view on major issues of the day, but when asked their opinion, they express one anyway. These counterfeit attitudes are made up on the spot to avoid the embarrassment of appearing ignorant or the shame of seeming negligent. Converse dubbed these counterfeit responses "nonattitudes" and presented evidence suggesting that they are a pervasive feature of the political thinking of mass publics.

The second theme is the role of political sophistication. Two points have long been agreed upon. First, citizens are distributed along a gradient of awareness and understanding of politics. Second, the distribution of citizens along this gradient is skewed to the right: only a small proportion fall toward the tail of engagement with politics and political sophistication while a very high proportion fall toward the tail of political ignorance and inattentiveness. Converse drove home the lesson that the character of citizens' thinking about politics--the articulation of their positions on the major issues, their grip on fundamental political abstractions, even the stability of their opinions over time--is conditional on their level of political awareness and sophistication.

These two master themes--the pervasiveness of nonattitudes and the conditioning role of political sophistication--need not logically butt heads with one another: indeed, their conjunction yields the standard prediction that the likelihood of nonattitudes is inversely related to the level of political information and awareness. But the two themes tug in different directions. Focus on the pervasiveness of nonattitudes, and you emphasize the formlessness and incoherence of the political ideas of ordinary citizens; their indifference to matters of public concern and absorption in day-to-day concerns; the absence of deliberation or serious discussion of public affairs on their part; and an all-too-common failure to form any view whatever about central issues of the day. By contrast, focus on the structuring role of political sophistication, and you emphasize the organization of political belief systems, their crystallization, constraint, and stability, all of course conditional on political awareness and understanding. In a word, the first theme highlights the absence of political thought among citizens as a whole; the second, the conditions of their thinking coherently about politics.

Curiously, although the themes of nonattitudes and political sophistication were yoked together in Converse's original formulation, research on them proceeded independently of one another. The nonattitudes theme struck sparks first, rapidly generating a body of methodological research oriented to measurement issues.3 The political sophistication theme took hold nearly a decade later, ultimately spawning at least as large a body of substantive and methodological studies. But a small number of honorable exceptions notwithstanding, work on the problems of nonattitudes and political sophistication have whistled down parallel rather than intersecting tracks. The mission of this book is accordingly integrative: to examine from as many vantage points as possible, in as many ways as possible, how these two master concerns--the problem of nonattitudes and the role of political sophistication--bear on one another.

This work also aims to be integrative at another level. European researchers who have done work on both problems have followed the work of their North American colleagues; North American researchers have not followed the work of their European colleagues as assiduously, to put it generously. Accordingly, a major aim has been to bring these two communities into closer contact.4 The result is a work designed so that each can engage the other, and both their points of difference and commonality in approach to the core problems of nonattitudes and political sophistication can be brought into focus.

Each substantive chapter has a distinctive story to tell. But their cumulative contribution--the variety of ways, some obvious, others less so, in which they complement one another and so contribute to developing a theory of political choice--is our concern here. By considering the new studies reported in this book in the context of the larger bodies of research of which they are a part, we believe it is possible to develop a framework in which many--perhaps most --of the distinct models of public opinion choice can be coherently situated.


When Converse first presented the concept of nonattitudes, he proposed an explanatory mechanism. When citizens had not given thought to an issue, yet wanted to act as though they had, he suggested that they choose a position at random. The act of choosing, so conceived, is like flipping a coin. Citizens oppose a government action if it comes up tails; support it if it lands heads. For a generation of research, the metaphor of coin flipping served as an explanatory mechanism for nonattitudes.

But it was only a metaphor and never an entirely apt one. As a choice mechanism coin flipping presumes indifference: landing heads is just as satisfying, from the point of view of the person making the choice, as landing tails. But how plausible is indifference to the outcome as a motivational postulate? On Converse's story, people act as though they have an attitude even when they don't because they want to appear as properly conscientious citizens who have given thought to the leading issues of the day even when they have not. Making a choice at random is a satisfactory way out. And it might be a satisfactory escape route if they had to answer only one question. But what is an easy way out once is onerous and duplicitous done time after time and at odds with the standard (and usually warranted) premise that respondents try to answer in good faith.

Converse's seminal study of mass belief systems triggered a landslide of research. Some minimized the nonattitude problem, attributing the largest part of response lability to measurement error (Achen 1975; Judd and Milburn 1980). Others differentiated the issue, attempting to specify who is most, and who least, susceptible to nonattitudes (see, for example, Zaller 1992 and Petty and Krosnick 1995). It is the work of Zaller that has had the greatest impact. He formulated a theory of how people can so readily change their position from one occasion to the next as to appear to be answering randomly, yet on every occasion be responding sincerely. They can do so, Zaller argued, because most of them simultaneously have reasons to say yes and no. What they do, when responding to a question, is not to look for an attitude they have already formed and stored away in a file drawer. Instead, they sample from a miscellany of positive and negative considerations that happen to be salient when they are searching for a response. Depending on the proportion of positive to negative considerations, their answer is affirmative or negative.

Viewed through the lens of sampling considerations from memory, a new theory of public opinion comes into focus. The problem with public opinions is not, as Converse suggested, that citizens have too few ideas. It is, Zaller contends, just the opposite. Most people have too many ideas and, what is more, ideas on all sides of most issues. These contradictory ideas, on this explanatory story, are stored in memory without their consistency either with one another or with other ideas that people hold being checked, particularly by those who are less politically aware and involved. Though the proportion of relevant ideas may be roughly balanced between positive and negative in total, particularities of question wording and ordering direct their attention to the one or the other. So they take one course of action as willingly, as sincerely, as its opposite.


Sampling considerations from memory is thus the pivotal mechanism underlying nonattitudes according to Zaller. This mechanism can explain instability, but it does not explain why some people have more stable opinions, others less. Some researchers have argued that the choice mechanism is different for politically interested citizens; others favor a uniform mechanism. In chapter 1, following the lead of Tourangeau, Rips, and Rasinski (2000), Saris advances the argument that the theory of sampling considerations applies across the board. However, he contends that two factors cause the sampled considerations to be different for different people.

The first factor is people's political interest. Interested people collect more information and integrate this information in more fundamental considerations that can be standardly invoked across occasions. Paradigmatic examples are party identification and ideological outlook. There is thus an inverse relation between congruence of accessible considerations and political interest and awareness. The second factor conditioning the sampling process of accessible considerations is the task the people have to perform. People do not automatically recognize the relevance of a general principle they have in memory. The formulation of the task can condition the salience of deeper-lying considerations that serve as a basis for a coherent, stable response.

Van der Veld and Saris, in chapter 2, develop a tri-component formal model distinguishing stable considerations, unique considerations, and measurement error. In doing so they move away from Converse's black-and-white model, which assumes that people are either stable or not. They also move away from measurement error model of Achen, which has no separate provision for unique considerations at each specific point in time, assigning them instead to the error term, which therefore balloons, in the process giving the impression of striking stability to opinions. Van der Veld and Saris's formal model also differs from Zaller's, because it assumes that all people react in the same way, sampling considerations from memory, but differing in the amount of unique and stable considerations at their disposal. Completing the analytical story, Saris suggests, in chapter 1, that difference in amount of stable and unique considerations depends on the two parametric factors in our larger account: political interest and the task given to the people.

These two variables, interest and task, fill the gap left by Converse, Achen, and Zaller concerning the conditions that determine whether people provide stable or unstable answers and are seen as having an attitude or not. All the chapters have, of course, their own purpose and value, but in this introduction, we would like to concentrate on those aspects that speak to the plausibility of the theory concerning the two relevant factors: the task and political interest. We start with the effect of the task, then discuss the effects of political sophistication or awareness or political interest, then finally, by way of exploring further the role of political awareness and interest, consider bases of attitude stability.


The clearest case where one can see the effect of the task is the analysis of the structure of political argumentation by Sniderman and Theriault (chapter 5). A sizable body of studies on issue framing has now accumulated. Without exception, these studies report that large numbers of the general public can be moved from one side of an issue to the other depending on how the issue is framed. Sniderman and Theriault, however, observe that these studies have been confined to one-sided presentations. An argument is presented to evoke support for a policy, or opposition to it, but not both. But in real politics, of course, opposing candidates compete to put across their point of view. Sniderman and Theriault therefore add to the standard design of framing experiments a condition in which people are confronted with both pro and contra arguments. They then show that when exposed to conflicting arguments, people are significantly more likely to make a specific choice that is consonant with their general view of the matter than is the case under the one-sided situation common to framing experiments. This is a clear example that shows that the accessibility of relevant considerations is conditional on the structure of the choice.

Kriesi (chapter 8) also pays attention to the effect of characteristics of the choice options the people are presented with. He characterizes them as more or less familiar, complex, and constrained. Notice that this last element, the concept of constraint, is used in a sense quite different from its customary one in American political science. Here it refers not to the degree of connectedness between positions on an array of public policies, but instead to whether a policy measure has a direct bite on an individual or not, as, for example, the closing down of the centers of town for car traffic or the introduction of speed limits has a bite for car owners that it lacks for those who do not own a car. One would expect that the more familiar the people are with policies and the more constrained by them they are, the more stable their responses to them will be. In his analyses Kriesi indeed finds that stability is higher conditional on policy familiarity and constraint. But no comparable effect conditional on complexity is observed. This result is interesting because it confirms that the different choice options make a difference.

A third study that shows that the task plays an important role is that of Neijens (chapter 10). The Choice Questionnaire tries to tackle the problem that many people are not so interested in political problems and therefore do not have enough information. By providing them with the necessary information, the designers of the Choice Questionnaire try to fill this gap so that the choices of the people become better. In this context, "optimal" means that people make a choice in agreement with their evaluation of the consequences of different options. Accordingly, in the Choice Questionnaire, the two variables, task and effort, are manipulated to explore their contribution to more optimal choices. Neijens shows that the formulation of the task makes a substantial difference in the performance of the respondents. In the most complex form of the Choice Questionnaire people indeed make use of the necessary information. The result: a much larger group of the public makes choices consistent with their own evaluations of possible consequences.

The notions of consistency--and its conceptual rival, ambivalence--are both pivotal and elusive, as the chapters by Meffert, Guge, and Lodge (chapter 3) and by Steenbergen and Brewer (chapter 4) make plain. The former examines ambivalence toward candidates for public office; the latter, ambivalence about policy. The former study shows ambivalence to have a pair of key properties. The first corresponds to a familiar, albeit contested, claim; the second represents an original contribution. The familiar claim concerns the frequency of ambivalence. It is, they maintain, commonplace, showing up, in at least one of their estimates, in as much as 40 percent of the American electorate. The original contribution consists in establishing that the more inconsistent people's opinions and feelings toward presidential candidates, the less extreme their evaluations of them, the weaker their approval of them, and the less crisp their images of them. In a word, ambivalence is inversely related to attitude strength.

By contrast, though treating ambivalence the same conceptually, Steenbergen and Brewer seemingly reach very nearly the opposite substantive conclusions. In their view, rather than ambivalence being common, it is atypical, and rather than being inversely related to attitude strength, it is unrelated to attitude stability, constraint, or consistency, to all of which it should be negatively related, if it truly is inversely related to attitude strength.

This difference between the findings of Meffert, Guge, and Lodge on the one side and Steenbergen and Brewer on the other may turn just on a correct assessment of the facts of the matter. One may be closer to the way that things are across the board. That is not a conclusion that can be comfortably drawn either way, given the evidence on hand. It is perfectly possible there is one state of affairs for views about political candidates, another for public policies. The choice of candidates or parties in elections gets so much attention that even people with little political interest get a lot of information about the candidates or the parties. This may lead a large portion of the public to "ambivalence" as suggested by Zaller (1992), that is, an unordered set of contradictory considerations. On the other hand, for many matters of public policy, most people characteristically collect very little information. So for the most part those who make up an issue public specially focused on this matter of policy will obtain contradictory information. And given their special degree of engagement, they are in a position to integrate conflicting information. In short, they may have "ambivalence" as suggested by Alvarez and Brehm (1995), that is, an ordered set of pro and contra considerations bearing on the policy at stake. Just so far as this is so, it is yet another illustration of how the nature of the task--choices between candidates or courses of action--can shape the process of choice and thereby the stability and consistency of choices.


Focusing on the gradient of political awareness, Zaller (chapter 6) comes away with a new insight into the nature of electoral dynamics. His takeoff point is the now taken-for-granted finding that the less politically informed and sophisticated that citizens are, the less likely they are to form genuine attitudes about political matters. The inference that has seemed immediately to follow is that the less politically informed and sophisticated that citizens are, the more unsound and illogical their reasoning about political choices. By viewing successive elections as a choice sequence, however, Zaller puts the issue in a quite different light. Turning the standard argument on its head, he shows that the more politically sophisticated, by virtue of having stronger, more crystallized opinions that are more consistently interlocked one with one another, are more likely to be anchored in place politically. Hence they tend to support the same party from one election to the next, even when circumstances have changed (for example, the economy has gotten worse). By contrast, the less politically sophisticated, precisely because they have not formed and committed themselves to a consistent set of political ideas, tend to be more insecurely anchored, freer to respond to changes in objective circumstances, and therefore better able to reward those responsible for government when times are good and punish them when times are bad.

The implications of this result for a theory of electoral accountability merit serious attention.5 Zaller highlights the dependence of political accountability on citizen responsiveness to short-term factors. By virtue of being more responsive to short-term factors, it is less informed voters, not more informed voters, who enhance electoral accountability. Or more exactly, Zaller's findings show that they enhance accountability for judgments on one dimension--economic. The calculus of accountability taken as a whole involves a complex blend, with more informed voters contributing a stability component based on fundamental political orientations, less informed ones a dynamic component based on changing aspects of the political situation.

Danielle Bütschi (chapter 11) also concentrates on political interest--but with a new twist. Taking advantage of the Choice Questionnaire technique, she examines the effect of obtaining new information relevant to the choice at hand. She comes to the conclusion that it is mainly the people with less political interest, and therefore less stable opinions, who use the information in the Choice Questionnaire to reformulate their choice. This conclusion is of obvious importance on both empirical and normative grounds.

All of these studies confirm the effect of political interest and engagement on the choices the respondents make. It is all the more important, therefore, to work through the relations of a family of concepts and indicators bearing on the concept of political interest. This is the mission of the studies in part 4.


It is natural to expect that political interest and attitude strength and stability should be related. And, as one would expect, all of the contributions in part 4 that order respondents on the basis of some indicator of political interest, information, or sophistication show that the more interested and engaged respondents are much more stable in their preferences than the less sophisticated respondents. This indicates the importance of this ordering principle.

Yet there is an obvious problem. Different concepts have been used to make the same, or exceedingly similar, ordering distinction. Some scholars have suggested the use of political sophistication, others awareness, political interest, issue importance, knowledge of the issue, etc. Even more confusing is that social psychologists use all these terms as indicators for a latent concept "attitude strength" (Krosnick and Abelson 1992). One of the nice results of this book is that some general evidence to clear up this discussion about attitude strength has been brought together.

Krosnick and Petty (1995) define attitude strength in terms of two parameters--stability and crystallization. On their view, strong attitudes should be stable and should have a strong impact on judgments and behavior. Krosnick and Abelson (1992), however, argue that a slew of variables are treated as indicators of attitude strength. It is accordingly of general importance that Bizer, Visser, Berent, and Krosnick (chapter 7) show that these different variables do not measure the same attribute. It follows that they cannot be equivalent indicators of attitude strength. Bizer and his colleagues then make an argument that these variables should instead be viewed as possible causes of crystallization or stability and, thus, as indirect causes of attitude strength. In a set of analyses the authors look at possible causal relationships among the different variables in the set. In doing so they show, for example, that the variables "importance of an issue" and "knowledge of the issue" have different causes and have different effects on other variables. Their analysis leads to a kind of model (Figure 7.2 of Bizer et al.), which we would like to summarize here in Figure Intro.1.

It is straightforwardly plausible to suppose that there is a strong effect of attitude intensity on the stability of the attitude. Figure 1 is interesting, then, because it suggests which of the conditional variables customarily taken to be indicators of attitude strength would be the best predictor for the stability of responses. Naturally enough, the best predictor is the variable that is closest in the causal chain to stability. The farther away a variable is in the causal sequence from the variable "stability" the less the effect will be.6 According to this path model, then, the best predictor would be "attitude intensity." The second best would be knowledge. And trailing along would be political interest and issue importance, although a cautionary note should be sounded for issue importance since it has also a direct effect on attitude intensity.

Bizer's analysis thus provides an ordering of the conditional variables with respect to their effect on the attitude stability which, until now, has not been available. This result is also interesting because it gives an explanation for findings in two other studies on the effect of different conditional variables for attitude stability.

Taking the two in turn, Billiet, Swyngedouw, and Waege (chapter 9) advance the straightforward hypothesis that stability should be a function of political awareness: the more engaged citizens are in politics, the more stable over time their attitudes should be. They distinguish three indicators of political involvement or awareness--education, political interest, and political knowledge--and then estimate the stability of attitudes conditional on these three variables. Of the indicators of awareness that Billiet and his colleagues examine, political knowledge dominates education and political interest in predicting attitude stability. On the basis of the results of Bizer et al., we suggested that the effect of knowledge would be larger than the effect of interest. That the effect of education is also smaller is not surprising because that variable is even farther away for attitude strength. So this result of Billiet, Swyngedouw, and Waege is an interesting corroboration of the causal ordering suggested by us on the basis of the study of Bizer et al.

The second study comparing the effects of different conditional variables for stability is Kriesi's (chapter 8). Kriesi looks at the effect of political awareness, among other factors, on the stability of the responses. As we have noted, he uses a trio of variables, two familiar--knowledge and opinion strength or intensity--and one original--situational constraints, defined as the immediate relevance or "bite" of a policy for specific citizens. According to the hypothesis we have derived from the results of Bizer et al., we would expect that attitude strength or intensity would have the most effect on stability of the preferences. This is indeed what Kriesi reports on the basis of his analyses. He even shows that knowledge and constraint have no significant effect anymore in an analysis where the three variables are introduced together. In that case only the strength of the opinion has a significant effect. This does not mean that these variables have no effect at all but that the effects of these variables all go through the variable strength or intensity of the opinion. This is another corroboration of the model set out in the figure above.

All the chapters described so far show that personal characteristics have an effect on the way people generate their responses to policy questions. It becomes clear that people with more intense opinions also have more stable opinions. Whether that is a good sign or a bad one from a normative perspective is an interesting question. From an empirical standpoint, however, the studies in this book show that the conditions for opinion strength can be ordered with respect to their effect where attitude strength is the best predictor of stability, then follow knowledge and importance, then interest and other variables like education. This ordering of conditions is a new and simplifying contribution of this book to the field of nonattitude research.


Three questions run through this book: When do citizens form genuine attitudes about politics? What is the impact of political awareness and interest on the formation of political ideas? And what difference does the task, the structure of the problem presented to citizens, make to the choices citizens make? The three questions are not emphasized equally, but they recur repeatedly. And intuitively, the answers to one bear on the answers to the others. The last chapter, by Sniderman and Bullock (chapter 12), presents a theoretical framework to tie them together. It is admittedly a framework that proceeds from an unusual perspective. In the study of public opinion the overwhelming emphasis descriptively has been on the inconsistency of ordinary citizens' political ideas. By contrast, Sniderman and Bullock contend that if a theory of public opinion is to go beyond the problem of nonattitudes, it must provide an account of the conditions under which citizens achieve consistency. They accordingly distinguish a trio of ways the construct of consistency commonly has been understood, singling out one--maximizing congruence between alternative ways of dealing with a particular matter with one's general view of the matter. How, they ask, can ordinary citizens achieve consistency so understood?

A number of consistency-generating mechanisms--among them, on-line processing, core values, and judgmental heuristics--have been identified. But these mechanisms cannot provide a basis for a consistency theory of any substantial scope. Quite simply, given the minimal level of attention the public pays to politics, it is difficult to see how ordinary citizens are in a position to achieve consistency in political judgments and choices making choices on their own hook. But that, according to Sniderman and Bullock, is just the point. Citizens do not have to do all the work on their own. Political institutions--and above all, political parties--do the heavy lifting.

They accordingly focus on the idea of menu dependence. On this hypothesis, the choices that citizens make are a function of the set, or menu, of choices on offer. This menu is organized by political parties competing for electoral support. Stated broadly, citizens can coordinate their responses to political choices insofar as the alternatives they can choose among are coordinated by political parties.

By way of draping some evidentiary clothing on this hypothesis, Sniderman and Bullock explore three aspects of menu dependence: consistency for sets of issues; for issues taken one at a time; and for variations across issues. In the process they introduce a number of distinctions, for example, between consistency within and between political domains. This is very much a venture in looking forward. The evidence on hand is indirect and incomplete. And there are obstacles that it is not obvious now how to clear, above all, on what terms consistency maximizing should be reconciled with belief updating. But with all appropriate qualifications capitalized for emphasis, there is something to be said for acknowledging that citizens do not organize politics on their own: it comes to them already organized.

Return to Book Description

File created: 8/7/2007

Questions and comments to:
Princeton University Press

New Book E-mails
New In Print
PUP Blog
Princeton APPS
Sample Chapters
Princeton Legacy Library
Exam/Desk Copy
Recent Awards
Princeton Shorts
Freshman Reading
PUP Europe
About Us
Contact Us
PUP Home

Bookmark and Share