How much foreign policy disagreement is there, really, between the two largest parties in America? The rhetoric of presidential campaigning sometimes makes the differences seem stark. But many Democrats are at pains to emphasize that some central features of the Bush Administration's war on terror are likely to remain in place in the event of a "regime change" at home in next year's election. This may include even the controversial notion of preemptive or preventive military action (for simplicity's sake, I use the two interchangeably here). At a recent Washington, D.C., mini-conference on transatlantic relations, a former Clinton diplomatic appointee said that preemption can be legitimate and necessary and gave an example to illustrate: If we had acted against Hitler in 1938, he said, no one would argue with that. In private conversations after his remarks, the same example was invoked by two other liberal foreign policy thinkers, in order to make the same point. They wanted to show that they, and by extension a liberal Administration, would be tough when it came to emerging threats to U.S. national security.

This is no surprise, given that Americans are likely in no mood to be surprised again as they were on September 11^th, 2001. But saying that one would have acted against Hitler before he invaded Poland does very little to prove a willingness to act preventively. Instead, invoking that example shows how much of a gap exists between people who say that and people who really are committed to a policy of occasional preemption.

Uncertainty is the reason the Hitler example serves so poorly. The point of preemptive or preventive action is that to avert disasters that would or might well happen in the absence of the action. Those who carry out preventive action will never be absolutely sure what would have happened had they not acted. So it proves very little to say what one would have done in historical scenarios in which no-one acted and we are certain what happened as a result. One can call for action against Hitler in 1938 in the full and certain knowledge that it would have prevented world war and genocide. But such conditions of certainty are not the circumstances in which decisions of prevention have to be made, ever. These decisions are made, instead, in conditions of uncertainty.

The recent debate over preemptive and preventive action has generated lists of possible standards under which such actions might be justified. These concern the imminence of the threat, the degree to which it might be deterred by non-military means and other matters. But no such list can adequately address the uncertainty which plagues all future-oriented action. This uncertainty occurs in two quarters: in any one decision-maker's minds, and then amongst decision-makers. First, each decision-maker is bound to be uncertain just how likely the threat is to materialize, its scope if it did, and what actions are justified in against it. Second, even if one decision-maker decides that a threat is sufficient to justify preventive action, not all others will necessarily have arrived at the same judgment under uncertainty. These differences can be domestic. Michael Oren's Six Days of War shows that the Israeli cabinet was unsure about a possible Arab attack in 1967, when deciding whether to strike first. Disagreement can also be international, as leaders of some countries may not agree that a threat is sufficiently serious. So a decision-maker must deal with his or her own uncertainty and then decide what to do if others conclude differently.

A real test of willingness to take preventive military action has to take uncertainty and its effects into account. I can think of at least two tests which do that. The first is whether a person endorses an actual case of preventive or preemptive action in history, one in which we will never know what would have happened had the action not occurred. One obvious modern example is the Israeli strike against Saddam's Osiraq nuclear reactor complex in 1981. We will never know for sure whether Saddam would have used it to create weapons-grade plutonium, succeeded in producing a working bomb and used such a bomb. Those are among the reasons the Israelis were roundly condemned at the time. Do would-be preemptors or preventers endorse this action, or any other they might care to name?

A second test is whether a person is willing to say they would have acted under conditions which describe Hitler in 1938 but which do not include what came later. In 1938, Hitler was a dictator with an insidious secret police. (But that did not set his regime apart from most of the time.) He had written a book describing a takeover of Europe, was maneuvering diplomatically very threateningly, and was engaging in massive military buildup. (But he had not so far used violence against a single neighboring state.) He spoke in shockingly racist terms, said Jews were evil and alluded to purging them from German society, and had moved to deprive Jews of civil and political right. (But there as yet existed no proof of a Nazi plan to murder Jews or anyone else.) It seems safe to say that had some states overthrown the Nazis on the basis of such facts, there would have been a chorus of voices - both domestic and foreign, both left-liberal and isolationist conservative - decrying them as adventuristic. Are the would-preemptors and preventors saying they would have intervened knowing only this much and despite that criticism?

The value of these tests is clear: they at least begin to establish general conditions under which the speaker advocates preventive action, conditions which centrally include inevitable uncertainty. When it comes to these sorts of tests, we know for a fact how the Bush Administration's foreign policy team comes down. We still don't know where many liberals come down on this question, and we won't found it out from the Hitler example, at least as it is usually invoked.

This doesn't apply only to deterring wars or terrorist attacks. In her Pulitzer prize-winning 2001 book "A Problem from Hell," Samantha Power called for a risk-taking interventionism designed to prevent genocides. The book was well received in many quarters not normally associated with advocating military action. But such a policy would have to confront uncertainty, too. Power sidestepped the issue in her book by focusing only on cases in which no-one acted and genocides then occurred. But endorsing action in such cases is a very poor test of a person's willingness to intervene prospectively. It offers little or no insight into what they would do if ethnic massacres in a given country were increasing, but no one could know if they would gain momentum, as sometimes happens, or just peter out, as also happens. We have the same problem when trying to decide what a liberal Administration would be inclined to do in the face of an acute, emerging national security threat.

Gerard Alexander, an associate professor of politics at the University of Virginia, is author of The Sources of Democratic Consolidation (Cornell University Press, 2002).