"Public opinion," Napoleon wrote, "is an invisible power, mysterious and irresistable. Nothing is more mobile, nothing vaguer, nothing stronger." In modern times, of course, we have a highly developed technique to demystify public opinion, render it highly visible, and even more irresistable. That technique is the public opinion poll.

Right now public opinion polls find that President Clinton enjoys a high public approval rating for his job performance, and though a huge majority doubts his character and trustworthiness, by a large majority the public says it does not want to see him removed from office. This, despite having told pollsters for months that if perjury in a legal proceeding was proved it would be a serious offense warranting removal from office. What gives? Has the public really changed its mind? Does Clinton have some kind of Svengali powers over the American public? Is fluoridated water finally having the effect the John Birch Society warned us about?

Several explanations are offered. Political cynics--and exasperated Republicans--suggest that Clinton’s drag-it-out strategy has simply worn down the people, who wish the non-stop scandal would simply go away. Social psychologists speak about "cognitive dissonance," which is the ability of individuals to keep two contradictory ideas in mind at the same time and be utterly untroubled about it. In this case, people tell pollsters that while they would never let their daughters work at the White House, they like the job our president is doing. Still others point to the common sense idea that the strong economy is influencing public opinion.

All of these explanations have some merit, but they accept the premise that opinion polls should be taken at face value. This is not to suggest that the polls are wrong or inaccurate, or that they have been "cooked." Rather, the problem with opinion polls on abstract questions is that they have been "precooked." By this I mean that most of the public does not have deeply formed views on many issues, so when a pollster calls up on the phone, he must offer a menu of answers from which the respondent must choose. This is where things get sticky.

Public opinion polling is very good at measuring or predicting the public mind about binary (either-or) questions, i.e., are you for Candidate A or Candidate B; do you favor or oppose Proposition X? These kind of polls are quite good at predicting the outcome of election contests. But when you move to a level of abstraction beyond this it is difficult to tell what people really think.

For example, consider a pair of questions that pollster Richard Wirthlin has been asking about the environment every year since 1981. The statement "continuing environmental improvements must be made regardless of cost" commands nearly 80% agreement. But the statement "economic growth should be sacrificed for environmental quality" gets less than 17% agreement. The difference in the result is rather jarring when you ponder that the questions are substantively identical. It is difficult to know from this poll what people really think about environmental policy; however, it is worth noting that most environmental ballot propositions that cost real money have been losing in recent elections.

Perhaps this is only to point up again the importance of nomenclature. Out on the West Coast the California Trial Lawyers Association decided three years ago to change its name, because trial lawyers have rightly acquired a high degree of public disapproval. A poll at about the time of the name change found that 31% of respondents "don’t trust at all" the California Trial Lawyers Association. But only 17% said they "don’t trust at all" the Consumer Attorneys of California (which is the new name for the trial lawyers). Same old shysters; different result.

This is not to suggest that polls and pollsters are bosh, but rather to reinforce George Orwell’s famous dictum that he who controls the language controls the outcome of the political debate. People surveyed in polls about issues are most likely to choose those precooked answers that conform to what people think they are supposed to think. Most people wish to choose the most "sophisticated" answer, which may or may not reflect what they really think about the issue, if they think about the issue much at all. Issue polling is really more of a test of how people respond to the way the questions are phrased and arranged. This is useful for corporate marketing, perhaps, but not for a great democracy.

So it is foolish for political leaders to take their lead on many issues, whether impeachment or abortion, from the tea leaves of conflicting and ever-changing polls. It is like trying to steer a ship by looking at its wake in a tossing storm. The real leader will understand where he or she wants to take the ship of public opinion, and seek to define the terms of our public discourse that will determine where the ship will end up.

Steven Hayward is an adjunct fellow at the Ashbrook Center for Public Affairs at Ashland University.