This edited volume contains a series of papers, which address a key issue in public opinion research. The issue could not be more general or more important. It is this question: Do members of the mass public actually have opinions about political issues? The question has been debated from the earliest work on public opinion, but was raised in a particularly stark form in a seminal paper by Philip Converse published in 1964. As is well known, he suggested that most people did not have meaningful opinions on most issues, due to the instability of their responses across different waves of a panel survey. The contributors to this volume are leading authorities in the field of political psychology and public opinion research, and the papers collectively make an important contribution to addressing this question. Paul Sniderman and John Bullock in the concluding chapter summarize the existing research agenda on this issue in the following terms: “The dominant themes of two generations of research have been that citizens tend to be muddle-headed (the lack of constraint theme), empty headed (the non-attitude theme) or both” (pp. 337–38).

There have been two broad lines of attack in examining this question, and both are well represented in the book. One is methodological, and it suggests that if survey instruments and methods of analysis are improved, then a lot of the “empty headedness” would disappear. However, the article by William Van der Veld and Willem Saris uses a linear structural relations modeling approach to control for measurement error in panel data, and concludes that response instability can still be quite high, even when measurement error is controlled. This challenges earlier methodological work, which suggested that response instability was largely a function of such measurement error. Similarly, a paper by Peter Neijens sets out the Choice Questionnaire methodology, which has been developed largely by Europeans with the aim of improving the quality of survey instruments. In this approach, respondents are given written information before their opinions are elicited in order to inject deliberation into the process of opinion formation.

The second line of attack is to try to provide a theoretical account of nonattitudes and response instability. The Receive-Accept-Sample (RAS) model associated with the work of John Zaller is one influential theoretical approach, essentially arguing that response instability is a function of ambivalence. The suggestion is that in responding to survey questions, individuals haphazardly sample the multiple attitudes or “considerations” they might have in their minds and construct opinions “on the hoof.” Not surprisingly, this can produce response instability over time. In their paper, Milton Lodge and his associates also find that ambivalence is a key issue in influencing responses to survey questions because of the multidimensional nature of attitudes.

A third issue discussed is that of framing effects. As is well known, questions framed in different ways will produce attitude instability in respondents. This occurs not because they have no attitudes but because the information that people have is being processed differently in different contexts. Thus, framing effects are a logical consequence of multiple attitudes. Sniderman and Sean Theriault show that while these effects are large, they do not necessarily lead to biases in public opinion because the public is exposed to rival frames by different parties, which means that framing becomes part of the political contest.

Studies in Public Opinion addresses the agenda it sets itself very well, but it is not clear if this agenda is wide enough to answer the original question. At least three key issues are missing from the analysis. The first is the thorny problem, initially identified by Anthony Downs, that because it is rational for individuals to ignore things they cannot change, it is rational for them to be ignorant and have no stable opinions. In classical decision theory, actors acquire information for purposive reasons—they want to use information to acquire personal benefits. But in a representative democracy, only mass or aggregate opinions matter, and individual opinions are irrelevant for bringing about political change. Given this, why does anyone bother to spend his or her time thinking about political issues? This point provides a theoretical grounding for the original Converse critique. This volume needed to address the incentives for acquiring political information. Essentially, we need a theory of opinion incentives.

A second important point is that we know that people tell lies to survey researchers. In the British Election Study, we can validate if the respondents to our surveys actually voted or not, despite what they tell us in response to questions. We find consistently that upwards of about 5% of respondents claim that they voted, when in fact they did not. The methodological debates have focused almost exclusively on issues of reliability and have tended to ignore this important topic of validity. This is not surprising because the former is easier to get a handle on than the latter. There needs to be a research agenda that investigates the circumstances under which people will tell the truth as they see it, when they are responding to survey questions.

The third issue has received a lot of attention in the social psychological literature but has not been satisfactorily resolved. It concerns the relationship between attitudes and behavior. Under what circumstances will individuals behave in ways consistent with the opinions they express to survey researchers? Thus, for example, does the RAS model apply to voting? Notwithstanding the work of Martin Fishbein and others, we do not have a satisfactory theory of the relationship between attitudes and behavior, and so this is an important question that is neglected here.

Having made these points, I do think that this volume is a “must buy” for students of public opinion, since it makes a valuable contribution to an important topic.