All of this brings up, of course, the critical question of whether respondents can ever tell us why they behave and think the way they do. If not, what, if anything, can pollsters do about it? What is the point of asking respondents anything about their thinking processes? Is it just a futile exercise in self-delusion? Should we go out of the polling business? Not anytime soon.

A first step in dealing with the issue, however, is to recognize that respondents' verbal self-reports about the causes of their behavior should not be taken at face value. Such self-reports may provide useful clues about influences respondents have been exposed to, such as media messages they mention on issues like the war in Iraq, as a plausible explanation or justification for their voting behavior. But only by experimentally manipulating and controlling exposure to these messages can we make firmly grounded causal inferences.

One way to simulate this causal process is by experimentally manipulating the context in which we ask questions within the survey interview. Prior to asking the standard question on presidential approval, for example, we could expose respondents to different sets of prior questions about the war in Iraq, versus Social Security reform, versus the war on terrorism and other media-generated content. We can then observe how these contextual manipulations affect the approval ratings for the president as well as the "reasons" respondents give for why they approve or disapprove of his performance. In this way we can begin to get at how respondents make causal inferences about their thinking and their behavior when we ask them why they do what they do.

Another methodology or questioning strategy for dealing with the problem is to use cognitive interviewing techniques (for instance, think-aloud, concurrent, and retrospective verbal probing techniques) to discover how respondents interpret and arrive at answers to survey questions about the reasons for their behavior, such as those used in the national exit polls: "Which ONE issue mattered most in deciding how you voted for president?" Respondents' answers to why-questions in public opinion polls thus become cognitive grist for the mill of analysis. Understanding the cognitive mist of "moral values" in the 2004 presidential election represents just such an opportunity for the public opinion analyst.

George Bishop is professor of political science and director of the Graduate Certificate Program in Public Opinion and Survey Research at the University of Cincinnati. He is also author of the recently published book, The Illusion of Public Opinion: Fact and Artifact in American Public Opinion Polls.

Additional reading