Revisiting Hiroshima in Iran: What Americans Really Think about Using Nuclear Weapons and Killing Noncombatants

In August 1945, 85 percent of the U.S. public told pollsters that they approved of President Harry Truman's decision to drop two atomic bombs on Hiroshima and Nagasaki.¹ In the more than seven decades since that time, however, U.S. public approval of the decision to use nuclear weapons against Japan has declined significantly. Seventy years after the end of the World War II, a majority of Americans no longer approved of Truman's decision: a 2015 poll found that only 46 percent of Americans still viewed the atomic bombing of Japan as “the right thing to do.”²

What accounts for this apparent decline in U.S. public support for using nuclear weapons and what is its significance? The answers to these questions are of more than just historical interest, for they can provide insights into whether the U.S. public would be a constraint against or a goad to encouraging a president to use nuclear weapons in international crises or conflicts in the future. The answers to these questions also help scholars and policymakers understand how Americans think about killing foreign civilians in war and how well they have internalized the principles of just war doctrine, such as noncombatant immunity, proportionality, and risk acceptance.

An initial review of U.S. public opinion on the decision to drop the bomb provides both a vivid reminder of the depth of hostility that Americans felt toward Japan in 1945 and a strong sense of declining support for the use of nuclear weapons since then. A University of Chicago National Opinion Research Center poll administered in September 1945 found that 44 percent of the public said the United States should have dropped the bombs “one city at [a] time”; 26 percent said that it should have dropped the bomb “where [there were] no people”; 23 percent said the United States should have “wiped out [all Japanese] cities”; and only 4 percent said they would have “refused to use” the bomb.³ A Roper poll published later that year in Fortune magazine similarly found that just 4.5 percent of the U.S. public believed that the United States should not have used atomic bombs at all; 13.8 percent believed that it should have first dropped a bomb on an unpopulated area; 53.5 percent maintained that it should have used the two bombs on two cities, as it did; and 22.7 percent believed that it should have dropped more bombs before Japan had the chance to surrender.⁴

Although the percentage of Americans approving of Truman's decision has fluctuated over time—in part because of the varied wording of survey questions, and whether alternatives to “approve” or “not approve” response categories were included in the polls—U.S. public support for dropping the bombs on Hiroshima and Nagasaki has been on a downward trajectory. For example, Harris polls in 1971 and 1982 found that 64 percent and 63 percent of Americans, respectively, deemed it “necessary and proper” to drop the atomic bombs on Japan. In contrast, a 1998 Roper poll found that only 47 percent thought that dropping the bombs was “a right thing,” 26 percent said that it was “a wrong thing,” and 22 percent were somewhere between the two extremes.⁵ Polls conducted by the Associated Press in March and July 2005 showed that, respectively, 47 percent and 48 percent of the public “approved” of the dropping of atomic bombs on Hiroshima and Nagasaki.⁶

The downward trend has continued in recent years. Indeed, when we replicated the 1945 Roper poll in a July 2015 poll (table 1), both of which included a “demonstration strike” option, we found that in 2015 less than 30 percent of the U.S. public supported dropping the two bombs (down from 53.5 percent in 1945), whereas 31.6 percent said they preferred the demonstration strike (up from 13.8 percent in 1945). The percentage of Americans who felt that the United States should not have used any atomic bombs more than tripled from 1945 to 2015, and the percentage of Americans who felt that it should have used more bombs before Japan had a chance to surrender dropped from 22.7 percent in 1945 to less than 3 percent seventy years later.⁷

Unlike these polls, which examined support for the use of the atomic bombs against Japan in 1945, other polls have focused on U.S. attitudes about contemporary conditions in which the United States might use nuclear weapons. The evidence here also suggests a general decline in support for using nuclear weapons. In 1949, for example, only 20 percent of the U.S. public agreed that “the United States should pledge that we will never use the atomic bomb in warfare until some other nation has used it on us first.”⁸ In contrast, a 2010 Chicago Council on Global Affairs survey found that 57 percent of the public agreed that “the U.S. should only use nuclear weapons in response to a nuclear attack by another nation” and that 20 percent agreed that “the U.S. should never use nuclear weapons under any circumstances.”⁹

Many scholars and policymakers—following the lead of Nina Tannenwald in her important book The Nuclear Taboo—have pointed to the decline in U.S. public support for using nuclear weapons as evidence of the gradual emergence of an ethical norm, a taboo, against the first use of nuclear weapons.¹⁰ Other scholars—most prominently, Steven Pinker, Neta Crawford, and Ward Thomas—have contended that the decline in support for using nuclear weapons is part of a much larger “humanitarian revolution” among the public, which has led to a broad acceptance of the just war doctrine principle of noncombatant immunity.¹¹ A third school of thought—featured most prominently in the work of Alexander Downes, Daryl Press, Scott Sagan, and Benjamin Valentino—has argued that the U.S. public has internalized neither the nuclear taboo nor a noncombatant immunity norm. These scholars contend that Americans appear to be willing both to support the use of whatever weaponry is deemed most effective militarily and to kill foreign civilians on a massive scale whenever such attacks are considered useful to defend critical U.S. national security interests and protect the lives of significant numbers of U.S. military personnel.¹²

In this article, we address these debates about U.S. public opinion concerning nuclear use and noncombatant immunity. We argue that the previous polls showing a decline in support for the atomic bombings in 1945 and contemporary uses of nuclear weapons are a misleading guide to understanding the real views of the American public about the use of nuclear weapons and the killing of noncombatants. Such polls failed to place respondents into a mind-set in which they are forced to make a trade-off between risking U.S. soldiers’ lives (if the United States does not use nuclear weapons) and killing foreign noncombatants (if the United States does use nuclear weapons). Yet, precisely this type of trade-off was at the heart of Truman's decision in 1945; and scenarios that entail this kind of trade-off are arguably among the most realistic and serious conditions in which a president and the public would be forced to contemplate the use of nuclear weapons in the future.

The first section of the article reviews the three main schools of thought about public attitudes toward the use of nuclear weapons and the killing of foreign noncombatants. The second section presents the findings from a survey experiment conducted in July 2015 in which a representative sample of the U.S. public was asked about a contemporary, hypothetical scenario designed to replicate the 1945 decision—the use of a nuclear weapon against a city in Iran in an attempt to end a war that the Iranian government had started in response to the imposition of U.S. economic sanctions. This type of thought experiment obviously cannot re-create the depth of urgency and emotion that many Americans felt in 1945. As such, it constitutes a conservative test of the public's willingness to use nuclear weapons and target large numbers of noncombatants in a war.

The findings demonstrate that, contrary to the nuclear taboo thesis, a clear majority of Americans would approve of using nuclear weapons first against the civilian population of a nonnuclear-armed adversary, killing 2 million Iranian civilians, if they believed that such use would save the lives of 20,000 U.S. soldiers. In addition, contrary to the noncombatant immunity norm thesis, an even larger percentage of Americans would approve of a conventional bombing attack designed to kill 100,000 Iranian civilians in the effort to intimidate Iran into surrendering (see table 2).

The third section of the article analyzes the reasons why so many Americans in our scenarios approved of using nuclear weapons. We present demographic data on the key characteristics of the supporters of nuclear use and demonstrate how our survey respondents explained their preferences about whether or not to use nuclear weapons. Here we present some novel findings showing that women support nuclear weapons use and violations of noncombatant immunity no less (and in some cases more) than male respondents. We also provide insights into how a belief in retribution and an ability to assign culpability retrospectively to foreign civilians allows individuals to rationalize the killing of foreign noncombatants. In the fourth section, we present the results of a related survey experiment constructed to determine whether Americans would support a diplomatic compromise in a war with Iran to avoid either the killing of U.S. soldiers or the killing of foreign noncombatants. In the concluding section, we outline a future research agenda to further understand these phenomena and present the implications of our findings for debates about just war doctrine and the risk of the use of nuclear weapons in future conflicts.

There are three main schools of thought about U.S. public opinion and the use of nuclear weapons against civilian targets. The first theory, advanced most clearly in the work of Nina Tannenwald, is that over time both U.S. leaders and the public have internalized a belief in a “nuclear taboo.” According to Tannenwald, “The taboo is not the behavior (of non-use) itself but rather the normative belief about the behavior … a shared expectation about behavior, a standard of right and wrong.”¹³ Tannenwald insists that “leaders and publics have come to view this phenomenon not simply as a rule of prudence but as a taboo, with an explicit normative aspect, a sense of obligation attached to it.”¹⁴

For Tannenwald, and many others, decreasing public support over time for dropping the atomic bombs in 1945 is clear evidence that a nuclear taboo exists. Tannenwald, for example, has argued that “the public's changing interpretation of the correctness of Hiroshima and Nagasaki over the years is perhaps explicable in the terms of a general delegitimation of nuclear weapons.”¹⁵ Peggy Noonan, special assistant and head speech writer for President Ronald Reagan, credits John Hersey's book, Hiroshima, for a universal moral repulsion against the bomb: “Hiroshima did a huge and historic thing. It not only told the world what happened when a nuclear weapon was used, it single-handedly put a powerful moral taboo on its future use. After Hiroshima, which sold millions of copies, no one wanted it to happen again.”¹⁶ Jeffrey Lewis similarly traces the origin of the nuclear taboo to Hersey's book and maintains that “[o]ver time, we've come to see nuclear weapons as Hersey saw them, as the ultimate expression of material and spiritual evil of total war. … The implication of this norm, of course, is that we can't actually use nuclear weapons.”¹⁷ Thomas Schelling, in his 2005 Nobel Prize speech—“An Astonishing Sixty Years: The Legacy of Hiroshima”—referred to and celebrated the “nearly universal revulsion against nuclear weapons.”¹⁸ In 2008, United Nations Secretary-General Ban Ki-moon also praised the existence of a nuclear taboo while lamenting that it had not led to nuclear disarmament: “Despite a longstanding taboo against using nuclear weapons, disarmament remains only an aspiration.”¹⁹

The nuclear taboo literature has shown that some U.S. leaders, during specific crises or wars, have felt that they were constrained by public opposition to the United States initiating the use of nuclear weapons. Tannenwald, for example, notes that Secretary of State John Foster Dulles commented in a 1953 National Security Council discussion about nuclear use in Korea that the public believed there to be a “moral problem” with these weapons and suggested that the Dwight Eisenhower administration actively “break down this false distinction” so that nuclear weapons would be viewed “as simply another weapon in our arsenal.”²⁰ William Burr and Jeffrey Kimball, and Scott Sagan and Jeremi Suri, have similarly demonstrated that President Richard Nixon felt constrained in his ability even to threaten to use nuclear weapons against North Vietnam, as part of his “madman theory” of nuclear coercion, and therefore had to resort to a “secret” (and less effective) nuclear alert in October 1969.²¹

Still, a number of weaknesses of the existing tests of the nuclear taboo thesis are worth noting. First, although there is historical evidence that some U.S. presidents felt constrained by U.S. public opinion in particular crises or conflicts, scholars have not systematically studied contemporary public opinion polls on nuclear weapons use in those periods. Thus, we do not know if these leaders’ perceptions of public opposition were accurate. Second, polls showing declining support for the bombing of Hiroshima and Nagasaki do not tell us whether the decrease in support reflects changes in the U.S. public's attitudes toward using nuclear weapons or changes in attitudes toward Japan and increasing temporal distance from the pressures of the war. It was much easier for Americans to support using nuclear weapons against an adversary in a brutal war in 1945 than it is to imagine using such weapons today against Japan, one of the United States’ closest allies. In addition, these polls rarely specified the known or expected numbers of civilians killed in atomic bombing attacks or whether they were targeted intentionally. Consequently, it was up to the respondents and their knowledge of the past to judge whether the use of nuclear weapons conformed with the principle of noncombatant immunity or proportionality (whether the costs of civilian deaths were proportionate to the benefits gained). And most important, existing polls have not forced survey respondents to contemplate the trade-off between the use of nuclear weapons, including the resulting loss of civilian lives in the adversary's nation, and the use of U.S. ground troops in an invasion and the resulting loss of American soldiers’ lives.

A second school of thought argues that an even broader norm has constrained the conduct of war in recent years. Proponents of this view assert that a phenomenon that Steven Pinker has called “the Humanitarian Revolution” has led to a deeply embedded and widely held norm against killing noncombatants, regardless of the choice of weapon. Pinker argues that the strengthening of “the better angels of our nature … can be credited for declines in violence.”²² He contrasts the common popularity of cruel criminal punishments and the brutality of soldiers in war in the past to more civilized norms today and suggests that both rational choice and emotional development have been at play: “The expansion of empathy may help explain why people today abjure cruel punishments and think more about the human costs of war…. The decline of violence may owe something to an expansion of empathy, but it also owes much to harder-boiled faculties like prudence, reason, fairness, self-control, norms and taboos, and conceptions of human rights.”²³ For Pinker, “The nuclear taboo emerged only gradually…. It began to sink in that the weapons’ destructive capacity was of a different order from anything in history, that they violated any conception of proportionality in the waging of war.”²⁴ Yet Pinker clearly considers the nuclear taboo part of a broader phenomenon, a change in what kind of killing in conventional and nuclear war is deemed morally and politically acceptable by most Americans. “By the 1990s,” he writes, “the only politically acceptable American wars were surgical routs achieved with remote-control technology. They could no longer be wars of attrition that ground up soldiers by the tens of thousands, nor aerial holocausts visited on foreign civilians as in Dresden, Hiroshima, and North Vietnam.”²⁵

Ward Thomas extends this argument, asserting that the noncombatant immunity “norm has in recent decades engendered sensitivity to noncombatant casualties that not only constrains states from targeting civilian populations per se but also creates pressures to minimize incidental casualties in general.”²⁶ Thomas argues that the dropping of the atomic bombs on Hiroshima and Nagasaki had “an ironic effect” in that “the postwar invigoration of the bombing norm [by which he means the noncombatant immunity norm] owes much to visceral moral reactions to the nightmare of World War II.”²⁷ Although he acknowledges that the noncombatant immunity norm is “far from absolute,” Thomas does assert that the experience of World War II bombing created a sea change in popular attitudes: “While unrestricted city bombing remained an awesomely destructive means of warfare, in the postwar world it seemed to offend, more than it had previously, people's ‘sense of what is right.”'²⁸

Similarly, Neta Crawford, in her work on the evolution of U.S. airpower operations, argues that “changes in the normative beliefs (beliefs about what is right and wrong) of elites and the general public about targeting civilians caused the United States to alter its bombing practices.”²⁹ Crawford also connects this explanation to the experience of World War II strategic bombing: “To the extent they adopted the principle of civilian immunity, it was part of a global change in views about human rights—viz., that there is something called human rights and they belong to civilians of all sides, even in war. The carnage of the two World Wars, of course, encouraged the development of human rights norms…. This greater respect for human rights also, incidentally, underpins the increased emphasis on force protection; it is no longer acceptable to use soldiers as mere tools whose lives may be ‘wasted’ in the thousands.”³⁰

The secondary point that both Pinker and Crawford make—that an increased commitment to human rights could lead to both an increased commitment to noncombatant immunity and a decreased willingness to sacrifice the lives of soldiers in war—underscores an inherent tension in the logic of their arguments. By not addressing whether these two factors might cancel each other out or whether one is much stronger than the other, they cannot predict or measure what the overall effect of such evolutionary norms might be.³¹ Even if there is a greater acceptance of the principle of noncombatant immunity among the American public, does this counteract the effects of a decreased willingness to accept fatalities among U.S. military personnel in war?³²

A third school of thought maintains that the American people have internalized neither a nuclear taboo nor an unconditional norm supporting the principle of noncombatant immunity. Instead, scholars in this school argue that concerns about winning wars and the desire to minimize the loss of lives of their nation's soldiers dominate public opinion about military operations. Alexander Downes, for example, examined all known cases of “civilian victimization” in interstate war from 1816 to 2003, defining civilian victimization as “a military strategy in which civilians are either targeted intentionally or force is used indiscriminately” and that produces at least 50,000 noncombatant deaths.³³ Drawing on quantitative data and historical case studies, he concludes that “states—including democracies—tend to prize victory and preserving the lives of their own people above humanity in warfare: desperation overrides moral inhibitions against killing noncombatants.”³⁴ Indeed, contrary to theories that emphasize that authoritarian regimes are more likely to target and kill foreign noncombatants, Downes found that electoral pressures on leaders in democracies make them “somewhat more likely than non-democracies to target civilians.”³⁵

Downes's work has three limitations, however, that reduce the applicability of its insights to the specific questions we explore here. First, Downes did not directly assess public support or lack of support for mass killings, leaving open the possibility that democratic leaders incorrectly assumed that public support for violence against foreign civilians was high. Second, his research included both cases in which the killing of civilians was intentional and cases in which it occurred as a result of collateral damage, which limits the degree to which his findings can be said to measure support or rejection of the principle of noncombatant immunity. Third, all of Downes's main case studies focus on events that occurred in the first half of the twentieth century, before many scholars believe that the nuclear taboo and norms of noncombatant immunity were firmly established.

A second major contribution to this school of thought is Christopher Gelpi, Peter Feaver, and Jason Reifler's 2009 study, Paying the Human Costs of War:American Public Opinion and Casualties in Military Conflicts, which focuses primarily on how changes in real or estimated U.S. military casualties over time influence public support for ongoing or potential U.S. wars.³⁶ Although their book does not investigate potential nuclear-weapons-use scenarios, it does include a brief analysis of how expected civilian casualties in an enemy country might influence U.S. public support for starting a war. They acknowledge that there is often an “unavoidable trade-off between their civilian casualties and our military casualties” and therefore ask “how does the public weigh that trade-off and what, to use an infelicitous metaphor from economics, is the exchange rate?”³⁷

Gelpi, Feaver, and Reifler conducted an original survey experiment measuring U.S. public support for an attack on North Korea to destroy its nuclear program. They found, unsurprisingly, that support for an attack was higher (60 percent) when no casualty estimates were mentioned, but dropped (by sixteen percentage points to 44 percent) when the survey informed participants that the attack would result in 2,000 U.S. military fatalities. Surprisingly, however, when a separate sample of the public was given a scenario that estimated that 2,000 North Korean civilians would be killed in the U.S. attack, an almost identical drop in U.S. public support for the attack was recorded. This finding could suggest that the U.S. public holds a cosmopolitan view about justice in war: that all human life has equal moral worth and therefore the lives of North Korea civilians should be valued as highly as the lives of U.S. soldiers.³⁸ Alternatively, because the experiments drew on separate representative samples of the U.S. public, the result could simply reveal that there is a significant drop in support for starting an attack whenever any casualties—U.S. military, enemy civilians, or allied civilians—are mentioned.³⁹ Moreover, the experiment did not include a condition in which the same respondents were asked about a direct trade-off between killing U.S. soldiers and killing North Korean civilians, nor did it ask respondents to consider the intentional targeting of North Korean civilians. All respondents, however, were asked: “In general, how would you rate the importance of limiting American military deaths as compared to limiting foreign civilian deaths? Is limiting U.S. military deaths much more important, somewhat more important, about equally important, somewhat less important, or much less important than limiting foreign civilian deaths?” Gelpi, Feaver, and Reifler report that 52 percent of the respondents answered “much more” or “somewhat more important”; 38 percent answered “about equally important”; and 6 percent said “somewhat less” or “much less important.”⁴⁰

The third example of this school of thought is the 2012 study by Daryl Press, Scott Sagan, and Benjamin Valentino, which utilizes a set of survey experiments to examine the degree to which the U.S. public has an aversion to using nuclear weapons. Their study demonstrated that a majority of the U.S. public based their preferences about the choice of weapons not on whether the weapon was nuclear or conventional, but rather on its relative effectiveness in destroying the target. When presented with a hypothetical scenario in which U.S. intelligence agencies informed the president that al-Qaida was suspected of building a nuclear weapon in a remote site in Syria, 18.9 percent preferred a nuclear attack, and 47.9 percent said they would approve of using nuclear weapons if the president had decided to do so, even when nuclear weapons were deemed to be of equal effectiveness to conventional weapons. When nuclear weapons were deemed to be twice as effective in destroying the target, however, 51.4 percent of the public preferred a nuclear strike and approval ratings increased to 77.2 percent.⁴¹ Moreover, the majority of respondents who opposed using nuclear weapons reported that they did so primarily for fear of setting a precedent that might encourage the use of such weapons in the future against the United States or its allies, not because of moral concerns. Press, Sagan, and Valentino conclude, “[F]or most Americans, the inhibitions against using nuclear weapons are relatively weak and decidedly not subject to a taboo.”⁴²

There were three limitations, however, of the Press, Sagan, and Valentino study. First, to measure the strength of the nuclear taboo per se (and avoid conflating the taboo with an aversion to killing large numbers of civilians), the experiment held the number of expected fatalities in the attack constant and relatively low, at 1,000 noncombatants killed. Although the results showed that most Americans were willing to use nuclear weapons against a terrorist target, the study could not assess the degree to which that support would drop if larger numbers of civilians in Syria had been expected to be killed in the attack.⁴³ Second, the nature of the enemy—in particular, a terrorist organization that had attacked the United States—may have influenced the public support for the U.S. nuclear attack. Many Americans may have been willing to use nuclear weapons against al-Qaida, but the study did not examine whether they would be willing to use nuclear weapons against a foreign state, as the United States did in World War II. Finally, unlike in World War II, when civilians and civilian infrastructure were the targets of many bombing campaigns, the civilians killed in the attack against al-Qaida were described in the study as unintended collateral damage.

To test the alternative theories concerning U.S. public opinion, nuclear weapons use, and noncombatant immunity, we designed a set of survey experiments focused on a contemporary conflict with Iran.⁴⁴ The experiments asked subjects to consider a scenario in which Iran attacked the United States first and in which the president was presented with two options: (1) to send ground troops to capture Tehran, which would lead to large-scale U.S. military fatalities; or (2) to attack a major Iranian city, deliberately killing civilians, in the effort to shock the Iranian government into accepting unconditional surrender. The three theoretical approaches described above lead to different hypotheses about how a representative sample of the U.S. public would respond if asked about the potential first use of nuclear weapons against Iran, a country that Americans have negative feelings about today.⁴⁵ Tannenwald argues that the nuclear taboo has “qualities such as absoluteness, unthinkingness, and taken-for-grantedness.”⁴⁶ If the nuclear taboo thesis is correct, therefore, the majority of Americans should oppose the hypothetical U.S. nuclear attack against Iran. If Pinker, Ward, and Crawford are correct and the U.S. public has deeply internalized the noncombatant immunity norm, then the majority of Americans should also oppose the use of both nuclear and conventional bombing when it is designed to kill a large number of Iranian civilians. If the arguments that Americans prioritize winning the war and saving the lives of U.S. soldiers are valid, however, the majority of Americans should support the use of nuclear weapons or conventional weapons against Iranian civilians in the scenarios described in the survey experiment. If this final thesis is correct, moreover, it suggests that there has been relatively little change in U.S. public opinion about using nuclear weapons or killing noncombatants since 1945.

We contracted a leading political polling firm, YouGov, to administer our survey experiments to a sample of U.S. citizens older than 18. The polling was conducted from July 23 to July 30, 2015, among 780 respondents selected from individuals who registered to participate in YouGov internet surveys, with the data weighted by YouGov to reflect the demographic composition of the U.S. public. YouGov utilizes a technique called “sample matching” to approximate a representative sample.⁴⁷ Starting with a panel of opt-in participants, YouGov draws a stratified sample matched to the key demographic characteristics of the U.S. population.⁴⁸ This sampling technique is relatively new compared to traditional equal probability random sampling, but it is becoming increasingly popular for use in academic research applications, and its performance has been shown to meet or exceed that of surveys based on more traditional telephone polling techniques.⁴⁹

Unlike public opinion polls that inquire about actual wars, a survey experiment allows us to construct hypothetical scenarios in which we can hold relevant facts about the war constant (e.g., the causes of the war or the identity of the enemy) while varying only one aspect of the conflict (e.g., the number of noncombatants killed or the kind of weapons used). This design enables us to isolate the effects of different levels of expected noncombatant deaths and the effects of conventional versus nuclear bombing on the U.S. public's approval of attacks.

All the stories used in our experiments are presented in the online appendix.⁵⁰ Each respondent read a mock news article that reported that the United States had placed severe sanctions on Iran in response to allegations that the Tehran government had been caught violating the terms of the 2015 Joint Comprehensive Plan of Action (colloquially known as the Iran Nuclear Deal). In response, Iran attacked a U.S. aircraft carrier in the Persian Gulf, killing 2,403 military personnel (the same number killed in the Pearl Harbor attack, though that is not mentioned in the story). As described in the mock news story, “U.S. forces retaliated immediately with large-scale airstrikes that destroyed all of Iran's nuclear infrastructure, air defenses, and all Iranian Air Force bases and planes. When Iran rejected the United States’ demand for the ‘immediate and unconditional surrender’ of the Iranian government, the President ordered a ground invasion by U.S. Marines and Army forces designed to destroy the Iranian military and replace the government in Iran.” The story then reported that the invasion eventually stalled after several months of fighting and 10,000 U.S. military fatalities.

Subjects next read that the Joint Chiefs of Staff had presented the president with two options to end the war. The first option was to continue the land invasion to capture Tehran and compel the Iranian government to surrender. The second option was to “shock” the Iranian government into accepting unconditional surrender by dropping a single nuclear weapon on Mashhad, Iran's second-largest city. The article stated that a weapon would be targeted “directly on Mashhad … in the effort to undermine civilian support for the war and pressure the Iranian government to surrender.” Thus, it was clear that the bombing was a direct violation of the principle of noncombatant immunity and had no military target. Holding the estimated number of U.S. military fatalities in a continued ground invasion at 20,000, we varied the number of Iranian noncombatants expected to be killed in the attack on Mashhad from 100,000 to 2 million to measure the degree to which the U.S. public would be willing to use nuclear weapons and violate the principle of distinction (or non-combatant immunity) in a scenario similar to the Pacific War in 1945. We also included a third condition in which 100,000 Iranian noncombatants were estimated to be killed in a conventional bombing attack. The purpose of this condition was to determine how the use of nuclear versus conventional bombing would affect U.S. public attitudes about targeting large numbers of noncombatants. This condition thus provides a direct measure of the depth of the public's support for the noncombatant immunity norm compared to its aversion toward the use of nuclear weapons.

Because the U.S. military today is so much smaller than it was in World War II, we could not realistically replicate the conditions of August 1945 in our news story. In 1945, estimates of U.S. military fatalities in an invasion of Japan varied significantly—ranging from as low as 20,000 to as high as 300,000—depending on estimates about where the invasion would occur and how long the war was expected to last before Japan surrendered, as well as who was making the calculations.⁵¹ We chose an estimate of 20,000 U.S. military fatalities in the march to Tehran in our stories for two reasons. First, it was a more realistic number for a potential war with Iran today. Second, we compared the overall size of the U.S. military today (approximately 1.5 million) to that of the U.S. armed services at the end of World War II (approximately 12 million), and calculated that 20,000 military fatalities today would fall into the proportional midrange of estimated fatalities in 1945.

For each of these scenarios, we posed the following questions to respondents: “Given the facts described in the article, if you had to choose between launching the strike against the Iranian city or continuing the ground war against Iran, which option would you prefer?” and “Regardless of which option you preferred, if the United States decided to conduct the strike against the Iranian city, how much would you approve or disapprove of that decision?” We asked about the personal preferences of respondents to help us understand their beliefs about what is ethical and appropriate military behavior, which reflects how well or poorly they have internalized the principles of just war doctrine. We also asked about respondents’ willingness to approve nuclear and conventional bombing attacks, because that is a better measure of their willingness to support presidential decisions concerning such attacks. Approval rates may also be more relevant from a policy perspective, given that political leaders usually focus on estimates of future public attitudes, presumably influenced by leaders’ explanations of their decisions and the prospects of success, and not the public opinion preferences expressed before the action.⁵² The “approval” version of the question, moreover, more directly resembles the kinds of questions asked in other surveys about President Truman's decision to drop the atomic bombs in 1945.

As shown in figure 1, a majority of the U.S. public indicated that they would approve the bombing of Mashhad across all three conditions (columns B, D, and F). Indeed, Americans explicitly preferred the air-strike option against Mashhad to the ground assault on Tehran in two of the three conditions (columns A and E). Only in the condition in which 2 million Iranian civilians would be killed did a slight majority (52.3 percent) prefer the ground attack: the preference for a nuclear strike against Iranian civilians drops from 55.6 percent (column A) to 47.7 percent (column C) when the number of expected Iranian civilian fatalities rises from 100,000 to 2 million. Although this difference is not statistically significant, it suggests that there may be some degree of aversion to killing extremely large numbers of civilians. Approval rates, however, remain higher, at close to 60 percent, in both conditions (as seen in column B and D).

The percentage of Americans who approved of the bombing attack against Mashhad remained relatively stable regardless of whether conventional or nuclear weapons were used, and regardless of whether the attack was estimated to kill 100,000 or 2 million civilians. The absence of statistically significant differences in these latter results across different conditions is precisely what is substantively and politically significant. That nearly 60 percent of the U.S. public would approve of a nuclear attack on Iran that would kill 2 million civilians suggests that the decreasing level of support found in recent polls about Truman's decision to drop the bombs in 1945 is a misleading guide to public opinion about nuclear use today.

That approval ratings were higher than preference ratings in both of our nuclear survey conditions (the difference was only statistically significant in the 2 million Iranian deaths condition, columns C and D) suggests that there may be a small “rally around the bomb” effect, like the common “rally around the flag” phenomenon seen after most U.S. decisions to use military force.⁵³ Some Americans are willing to approve of a presidential decision to use nuclear weapons even in situations in which they did not personally prefer a nuclear response. These results do not appear to be influenced by the perception that the more destructive attack would likely be more effective. When we asked a follow-up question about whether the air strike was likely to lead to unconditional surrender, the percentages of respondents who answered positively were statistically indistinguishable across all conditions.

The conventional bombing scenario also produced noteworthy results. In column E, 67.3 percent of respondents preferred the conventional bombing attack that deliberately killed 100,000 Iranians, an increase of 11.7 percent from the percentage of respondents who preferred the nuclear attack in column A (p = 0.07). This result suggests that although these Americans may have been influenced by an aversion to using nuclear weapons against civilians, they were not averse to intentionally killing civilians with conventional weapons. That such a large majority of respondents preferred a conventional bombing attack on a city that would kill 100,000 civilians over a ground assault that would kill 20,000 U.S. soldiers is strong evidence against the theory that a robust noncombatant immunity norm has been internalized by the U.S. public today.

To explore the reasons why some Americans are more willing to use conventional weapons than nuclear weapons, we asked all of our subjects two additional questions. First, “Regardless of which option you preferred, how ethical or unethical do you think it would be if the United States decided to conduct the strike against the Iranian city in the situation described in the article?” Second, subjects were asked whether they agreed or disagreed with the statement that “launching the strike against the Iranian city would set a precedent that would encourage our adversaries to launch similar attacks against America or our allies in the future.” Our results indicate that a combination of normative and practical concerns help explain the higher support for the conventional air strike. Among those who opposed the nuclear strike, 65.1 percent judged it unethical. Among those who opposed the conventional strike, however, the percentage that judged it immoral declined only slightly, to 61.6 percent (the difference was not statistically significant). This suggests that anti-nuclear norms add little to the aversion to killing civilians. A larger percentage of those opposed to the strike, however, agreed that it would set a bad precedent both in the nuclear condition (76.1 percent) and in the conventional condition (76.8 percent).⁵⁴

The main conclusions of these survey experiments are clear. The majority of the U.S. public has not internalized either a belief in the nuclear taboo or a strong noncombatant immunity norm. When faced with realistic scenarios in which they are forced to contemplate a trade-off between sacrificing a large number of U.S. troops in combat or deliberately killing even larger numbers of foreign noncombatants, the majority of respondents approve of killing civilians in an effort to end the war. Protecting the lives of U.S. troops was a higher priority than preventing the use of a nuclear weapon or avoiding the large-scale conventional bombing of an Iranian city.⁵⁵

These survey experiments produce a kind of “historical resonance,” causing us to consider aspects of U.S. history upon which scholars have not reflected sufficiently. Such historical evidence both provides indirect support for the external validity of our experiments and generates new insights into the past. Polling evidence from World War II, for example, suggests that U.S. public support for killing enemy civilians on a large scale with chemical weapons increased when subjects were cued that using such weapons would save the lives of U.S. soldiers. In September 1944, for example, only 23 percent of the U.S. public said they would approve of “using poison gas against Japanese cities” if it meant “an earlier end of the war in the Pacific”; yet support surged to 40 percent in a June 1945 poll in which the prompt stated that a poison gas attack “against the Japanese” might “reduce the number of U.S. soldiers killed or wounded.”⁵⁶

Evidence from polls taken during the 1991 Gulf War also supports this observation. Two surveys by Gallup, identical except for the phrase “if it might save the lives of U.S. troops,” prompted the respondents to consider the U.S. use of an atomic bomb against Iraq. On January 27, 1991, the public was asked: “Do you favor or oppose using tactical nuclear weapons against Iraq if it might save the lives of U.S. troops?” On February 14–17, 1991, Gallup posed a similar question without mentioning the trade-off: “Do you favor or oppose the U.S. use of tactical nuclear weapons in the Persian Gulf War?” Support for using nuclear weapons was 45 percent when the trade-off was mentioned, but only 28 percent when it was not.⁵⁷ Caution should be used in interpreting these polls, however, because the findings might have been influenced by the fact that the second poll was taken toward the end of the war, when a larger percentage of the U.S. public was confident that the United States was going to win, making the use of nuclear weapons seem less necessary.⁵⁸ In addition, both poll questions explicitly referred to “tactical nuclear weapons,” which might have influenced the findings. Future survey experiments could examine the interactions among the desire to protect U.S. troops, the perceived probability of winning the war, and the use of different types of nuclear weapons (tactical weapons used against enemy troops versus bombing of a city, for example).

One way to begin explaining the reasons behind the surprising public attitudes revealed in our surveys is to examine the variation in the preferences of different demographic groups in our sample.

Figure 2 presents the average preferences across the three air-strike scenarios (100,000 Iranian deaths/nuclear weapons, 2 million Iranian deaths/nuclear weapons, and 100,000 Iranian deaths/conventional weapons) for five major demographic groups—party identification, age, education, race, and gender—and one additional characteristic, whether the respondent indicated that he or she approved of or oppossed the death penalty. Consistent with many other studies about public opinion and the use of force, we found that Republicans (69.5 percent) were much more “hawkish” than Democrats (48.4 percent) averaging across the three conditions.⁵⁹ Americans aged 60 and older were also significantly more likely to prefer the air-strike options (70.5 percent) than Americans younger than 60 (51.6 percent), which is also consistent with other surveys of public opinion and nuclear weapons use.⁶⁰ All three of these differences remained statistically significant when all the variables in figure 2 were included in a logistic regression.⁶¹

Because these findings are consitent with results from other studies on public opinion on the use of force, they add confidence in the external validity of our survey experiments. There are two surprising findings, however, that deseve more scrutiny. First, why is there so little difference between the preferences of men and women regarding nuclear use? Second, why is support or opposition to the death penalty correlated so strongly with nuclear-use preferences?

The finding that women were no less willing than men to support using nuclear weapons or conventional weapons in a manner that violated noncombatant immunity principles was particularly surprising. Women have been found to be less hawkish than men in most studies on the effect of gender on support for going to war or using nuclear weapons. Lisa Brandes's 1994 study examined all polls from 1945 to 1988 regarding support for the use of nuclear weapons and found an average 9.1 percent “gender gap” between women and men regarding approval of U.S. nuclear weapons use.⁶² Richard Eichenberg's broader 2003 study analyzed 486 surveys conducted from 1990 to 2003 and found that women are less supportive of the use of military force in general and that the gender gap between men and women concerning support for a war increases when likely or current U.S. military casualties are mentioned in the polling questions.⁶³ A rare finding to the contrary is reported in Deborah Brooks and Benjamin Valentino's 2011 study, which demonstrated that U.S. women are more hawkish than men in supporting the use of military force in two scenarios: when the war was described as a humanitarian intervention and when it had the support of a multinational organization, such as the United Nations, rather than when it was the product of a unilateral national decision.⁶⁴

We therefore disaggregated our respondents by gender to examine how male and female preferences for the air strike varied across our three scenarios. The findings in figure 3 demonstrate that women were no less, or even more, hawkish than men regarding their support for killing large numbers of foreign civilians to avoid the deaths of U.S. soldiers. Female respondents supported using the atomic bomb at virtually the same percentages (55.9 percent in column B, 54.3 percent in column D) regardless of whether 100,000 or 2 million Iranian noncombatants were killed. That a higher percentage of U.S. women (54.3 percent, column D) preferred the atomic attack that killed 2 million Iranians than did U.S. men (39.8 percent, column C) is particularly surprising, although given the relatively small cell sizes resulting from dividing our sample in two, this difference is not significant at conventional levels (p = 0.128). In addition, women's support for the air strike, unlike men's, did not increase significantly in the conventional condition. The interaction between gender and nuclear/conventional conditions, however, was only marginally significant (p = 0.079). Nevertheless, this finding provides some suggestive and surprising evidence that the aversion to nuclear weapons use may be stronger among men than women.

This finding about gender also provides another example of historical resonance, for it shines a new light on American public opinion during World War II. The approval rate in August 1945 for dropping the atomic bombs on Japan among U.S. women (83 percent) was virtually identical to that of American men (86 percent).⁶⁵ Further research could focus on identifying the conditions under which women may be no less willing, or even more willing, than men to support the use of military force in ways that violate the principle of noncombatant immunity and kill large numbers of foreign civilians.

The second particularly noteworthy finding for our study concerns the strong positive correlation between whether or not respondents preferred the air-strike option against Iran and whether they approved the use of the death penalty for convicted murderers in the United States. Across the three scenarios, Americans who approved of the death penalty were more than twice as likely to prefer the air strike option (67.3 percent) than were Americans who opposed the death penalty (31.5 percent).⁶⁶ This finding is consistent with Peter Liberman's study of the sources of U.S. public support for the use of torture and his findings about belief in retribution and elite support for the 1991 Gulf War.⁶⁷ Rachel Stein has also found that democracies that have maintained the death penalty at home are more likey to intiate military conflicts abroad. She attributes this correlation to beliefs in retribution and revenge that are prevalent among some democratic publics.⁶⁸

What could be the linkage between death penalty support and approval of nuclear or conventional strikes designed to kill Iranian civilians? Our surveys reveal a strong pull of retribution and a tendency for individuals to rationalize the killing of others by claiming that it was their fault. For example, averaging across all three of our main conditions, a large majority (83.9 percent) of those who preferred the air-strike option agreed with the following statement: “Since Iran's leaders started the war, they are morally responsible for any Iranian civilian deaths caused by the U.S. strike described in the news story.” Among respondents who preferred the ground-assault options, only 52.8 percent agreed with that statement. A large majority (68.5 percent) of the respondents who favored the air-strike options also agreed with the statement that “[b]ecause the Iranian civilians described in the story did not rise up and overthrow the government of Iran, they must bear some responsibility for the civilian fatalities caused by the U.S. strike described in the news story.” In contrast, among those who preferred the ground-assault options, only 36.6 percent agreed with that statement.⁶⁹

In an open-ended question, we asked respondents to write down the single most important reason they supported the air strike. The results revealed that utilitarian calculations about saving U.S. lives and ending the war promptly, as well as assigning culpability retrospectively, may have motivated subjects’ preference for the air strike. The majority of respondents in all three conditions who supported the bombing attack cited saving American lives or ending the war quickly as the reason for their support. That was not unexpected. What was surprising was the number of Americans who suggested that Iranian civilians were somehow culpable or were less than human. Justifications for the atomic attack that killed 100,000 Iranians included: “Hate the thought of being the only nation again to launch a nuclear strike, but they started it and we have lost enough lives”; and “Iran lied again about their nuclear facilities, launched an attack on an American warship that killed and hurt 4,000 Americans, and nothing short of complete and total victory over the lying savages would be acceptable to me.” Justifications for killing 2 million noncombatants included: “They were dancing in the streets and partying in favor of what happened to the twin towers”; “They hate us and they will violate any agreement”; “Islam is a religion of fanatics [and] they need to be taught a lesson”; “Air strike would wipe out any military opponents who will be hiding among civilians”; and “Kill the cockroaches.” Answers given by those who voted for the conventional attack that was estimated to kill 100,000 Iranian civilians included: “The article said that Iran attacked the U.S. ship first. The U.S. is not the aggressor here”; “Iran is ruled by radicals who have no regard for innocent life”; and “My first choice would be to decapitate the snake—wipe out the capital—wipe it clean—every living thing.”

The experiments of Stanley Milgram and Albert Bandura focused attention on how often and easily individuals rationalize their violence against others by retrospectively insisting to themselves that the victims deserved to be punished. Milgram concluded after his famous obedience experiments “that many subjects harshly devalue the victim as a consequence of acting against him…. Once having acted against the victim, these subjects found it necessary to view him as an unworthy individual, whose punishment was made inevitable by his own deficiencies of intellect and character.”⁷⁰ Bandura similarly notes that “blaming one's adversaries … can serve self-exonerative purposes. In this process, people view themselves as faultless victims driven to injurious conduct by forcible provocation. Punitive conduct is, thus, seen as a justifiable defensive reaction to belligerent provocations.”⁷¹

We found evidence for this phenomenon in our experiments. We asked all respondents the following question: “Please rate your feelings toward the following countries, with one hundred meaning a very warm, favorable feeling, zero meaning a very cold, unfavorable feeling, and fifty meaning not particularly warm or cold. You can use any number from zero to one hundred. The higher the number the more favorable your feelings are toward that country.” Averaging again across all three treatments, we found that the subjects who approved of the air strike rated Iran 19.3 out of 100, more than 15 points lower than subjects who opposed the strike.⁷² We do not know from this evidence, however, which way the causal arrow moves—did respondents support the air strikes because they have more negative feelings about Iran compared to ground war supporters, or do air-strike supporters prefer that option for other reasons and then retrospectively justify their decision by feeling more negatively about Iran? We can, however, see evidence of how the extent of killing in an air strike retroactively influences the respondents’ feelings about Iranians when one compares the “thermometer scores” of the two different sets of respondents who approved of the different nuclear attacks. The 55.6 percent who preferred the air strike that killed 100,000 Iranians gave Iran a 19.4 out of 100 mean score, whereas those who opposed the strike rated Iran 28.6. In the condition in which 2 million Iranian civilians would be killed by the strike, the 47.7 percent who preferred the air strike gave Iran a 14.5 mean score, and those who opposed it gave the country a score of 35.7. Thus, the difference between the thermometer ratings of those who supported and opposed the strike more than doubled, from 9.2 points in the 100,000 deaths condition to more than 21 points in the 2 million deaths condition. This effect was marginally significant (p = 0.082). The prospect of killing more noncombatants appears to have both polarized attitudes toward Iran and intensified the hostility of those who supported the strike as a way of justifying the large number of deaths.

This kind of assignment of “retroactive culpability” is common. We can see it in fiction from the pre-nuclear age and in the history of the first days of the nuclear age. Leo Tolstoy, in his scathing description of Czar Nicholas in the novel Hadji Murat, described this phenomenon well: “Nicholas frowned. He had done much evil to the Poles. To explain that evil he had to be convinced that all Poles were scoundrels. And Nicholas regarded them as such and hated them in proportion to the evil he had done them.”⁷³ President Truman, when responding to a letter criticizing his “indiscriminate bombing” of Japanese cities, also appears to have justified his decision by assigning culpability to all Japanese and dehumanizing the victims: “Nobody is more disturbed over the use of Atomic bombs than I am but I was greatly disturbed by the unwarranted attack by the Japanese on Pearl Harbor and their murder of our prisoners of war. The only language they seem to understand is the one we have been using to bombard them. When you have to deal with a beast you have to treat him as a beast.”⁷⁴

According to international law and traditional just war doctrine, only civilians who are making a direct contribution to a government's war effort can be treated as combatants and thus liable to attack. Mere political support for a government is not sufficient to make a civilian a legitimate military target, and certainly the failure of a people to overthrow a government is an even less legitimate excuse. There is some support among revisionist just war theorists for including foreign civilians’ indirect responsibility for contributing to their government's war effort through political support as a legitimate consideration in attack decisions.⁷⁵ Nevertheless, no one argues that leaders and the public should assign culpability retroactively to justify the killing of civilians. Some of our respondents, however, appear to have done just that: they viewed Iran and Iranian civilians as Czar Nicholas did the Poles and “hated them in proportion to the evil that [they] had done them.”

Whenever one is confronted with two morally repugnant choices, it is prudent to consider whether other possible courses of action are available, including revising one's objectives. Michael Walzer, in his classic Just and Unjust Wars, criticized the bombing of Hiroshima and Nagasaki from precisely this perspective: “Americans would accept nothing less than unconditional surrender. The war aims of the American government required either an invasion of the main islands, with enormous losses of American and Japanese soldiers and of Japanese civilians trapped in the war zones, or the use of the atomic bomb. Given that choice, one might well reconsider those aims.”⁷⁶

Would the U.S. public accept less than unconditional surrender in a war with Iran that the Tehran government had started with a “surprise” attack on a U.S. warship? To explore this question, we constructed an experiment designed to examine how the U.S. public would react if it was also presented with the option of changing the terms of surrender from unconditional surrender to allowing Ayatollah Ali Khamenei to stay as a figurehead spiritual leader without political authority (see figure 4). A different representative sample of Americans read the same story with the identical ground-assault costs (20,000 U.S. soldiers) and nuclear air-strike costs (100,000 Iranian civilians) as in the earlier story, but with the following paragraph added:

It is valuable to understand whether the U.S. public would support or approve of the dropping of the bomb if there was an expectation that the Iranian government would surrender under this diplomatic compromise option. All subjects who read this story were then asked, “If you had to choose between the three options described in the article, would you prefer to launch the strike against the Iranian city, to continue the ground war against Iran, or to offer to allow the Ayatollah to stay on as a spiritual leader without any political power?”

This experiment attempted roughly to mirror the 1945 history in which President Truman insisted on the unconditional surrender of Japan in the Potsdam Declaration, despite having received advice from Secretary of War Henry Stimson that some kind of guarantee that Emperor Hirohito would not be put on trial or punished after the war would be needed to get Japan to surrender.⁷⁷ This advice proved prophetic when the Japanese government, after the destruction of Hiroshima and Nagasaki and the Soviet entry into the war, signaled that it was ready to surrender but only on the condition that surrender did not prejudice “the prerogatives of his majesty (Emperor Hirohito) as a sovereign ruler.”⁷⁸ Truman immediately sought to signal the Japanese government through the carefully crafted letter drafted by Secretary of State James Byrnes on August 11 that the emperor would, in fact, not be subject to war crimes trials and would be permitted to stay on as a figurehead leader in Tokyo. The Byrnes letter, while insisting that the “authority of the emperor” would be “subject to the Supreme Commander of the Allied Powers,” also promised that “the ultimate form of government of Japan shall, in accordance with the Potsdam Declaration, be established by the freely expressed will of the Japanese people,” which was understood by both Hirohito and Truman as a signal that the emperor would not be put on trial.⁷⁹ On August 14, after having been briefed on the contents of the Byrnes letter, and a few days after both the Soviet Union's entry into the war and the atomic bombing of Nagasaki, Hirohito finally decided to publicly accept the Allied peace terms.⁸⁰ On August 12, when a member of Japan's imperial family asked Hirohito if the war would continue if the kokutai (the imperial national polity) could not be preserved, the emperor replied “of course.”⁸¹

President Truman did not discuss the Byrnes letter in his public address announcing the end of the war in which he insisted that the final Japanese surrender constituted “a full acceptance of the Potsdam Declaration which specified the unconditional surrender of Japan.”⁸² Truman understood the public view of Hirohito's culpability: a poll taken in June 1945 showed strong support for punishing the emperor for his role in the war: indeed, 11 percent of the U.S. public favored “keeping him in prison for the rest of his life,” and 33 percent favored executing him.⁸³ It is impossible to know what would have happened had the United States softened its unconditional surrender terms earlier. Truman did not express any remorse about the failure to signal the Japanese about the fate of the emperor until after both bombs had been dropped, simply stating in his diary on August 10: “They wanted to make a condition precedent to the surrender. Our terms are ‘unconditional.’ They wanted to keep the Emperor. We told'em we'd tell'em how to keep him, but we'd make the terms.”⁸⁴ Only Secretary of War Stimson, who had earlier advocated for softening the unconditional surrender terms, would later express his belief “that history might find that the United States, by its delay in stating its position, had prolonged the war.”⁸⁵

The results presented above suggest that the majority of the U.S. public has not foresworn nuclear weapons or rejected the targeting of civilians during war since World War II. Might the public be more open today, however, to the kind of deal that Truman and Stimson were willing to make in 1945? Would the U.S. public support such a diplomatic compromise today, if it could avoid either an atomic attack on Iran or a costly ground assault on Tehran to end a war?

Recall from figure 4 that when respondents read that the president was considering only two options in the first experiment—and asked which they preferred—55.6 percent (column A) preferred launching a nuclear strike that would result in 100,000 Iranian civilian deaths, whereas 44.4 percent (column B) preferred a ground attack that would result in the deaths of 20,000 U.S. soldiers. When respondents were presented with a third option (the deal involving the Ayatollah Khamenei), however, the percentage preferring the nuclear strike decreased to 40.3 percent (column C), and only 18.6 percent (column D) now preferred the conventional ground assault. In short, 41.1 percent (column E) of the public was willing to permit the ayatollah to remain as a spiritual leader in Iran if that meant ending the war without further U.S. military fatalities or Iranian civilian deaths. It is important to emphasize that the majority of those now supporting the diplomatic option appear to have shifted from the camp that had preferred the conventional ground-attack option in the scenario without the diplomatic option.

Most important, 40.3 percent (column C) of the public still preferred the nuclear attack that would kill 100,000 Iranian civilians rather than compromise on unconditional surrender. This finding is particularly disturbing. That 55.6 percent of the U.S. public preferred and 59.3 percent approved of killing 100,000 Iranian civilians to save 20,000 U.S. soldiers may not be surprising to many. It suggests, after all, that Americans place a much higher value on saving U.S. soldiers’ lives than on sparing the lives of foreign noncombatants. However, the finding that more than 40 percent of Americans are willing to kill 100,000 Iranians to avoid permitting Ayatollah Khamenei to stay in a position of spiritual leader without political authority after the hypothetical war suggests that something other than pure utilitarian assessments of the value of American versus Iranian lives influenced the assessment of options in this scenario.⁸⁶

On the one hand, when asked, for example, if they agreed with the statement, “If the U.S. agrees to allow the Ayatollah to stay on as a spiritual leader without any political power, other adversaries will doubt the credibility of the United States to carry out its threats in the future,” 70.1 percent of those who preferred the nuclear strike said yes. On the other hand, when asked whether they agreed with the statement that “[b]ecause the Iranian civilians described in the story did not rise up and overthrow the government of Iran, they must bear some responsibility for the civilian fatalities caused by the U.S. strike described in the news story,” 77.9 percent of those who preferred the strike answered yes. Moreover, 40.2 percent of those who preferred the nuclear strike also said they thought the deal would end the war. So, those subjects did not choose the nuclear strike based on a perceived lack of effectiveness of the diplomatic option.

These findings warn us, as Liberman has noted, that “[p]eople tend to downplay or underestimate their own retributiveness.” “Survey respondents,” he writes, “are far more willing to justify their support of the death penalty in terms of a ‘life for a life’ or ‘punishment fits the crime’ than as ‘retribution’ or ‘vengeance,’ even though these are identical concepts. Out of wishful thinking or social desirability bias, it seems, people habitually exaggerate the instrumental purposes of punishments that they actually favor for retributive reasons.”⁸⁷

Some respondents’ answers to the question of why they supported the nuclear attack, despite the existence of a diplomatic option to end the war, are more candid and reveal the depth of feelings of retribution and hatred. Reasons for supporting the nuclear attack in this condition included answers such as: “They visously [sic] attacked America, and starting [sic] a war by killing many Americans and no more Americans should have to die for a someone else's hateful acts”; “Wipe them out, all leaders and followers”; “Bomb the hell out of them; They're all barbaric animals anyways, dirty muslim [sic] lives are less valuable then [sic] American lives”; “Don't stop until they are dead”; and “They attacked us first, therefore, show no mercy. But, the bomb should be dropped on Tehran.”

Recall that 22.7 percent of the U.S. public in 1945 said that they wished that President Truman had dropped many more bombs before Japan had a chance to surrender. Few Americans today express such views about Japan, but it would be falsely reassuring to believe that such sentiments are anachronistic, produced by the long and deadly fighting in the Pacific War. Our experiments reveal a darker truth. As in 1945, a significant portion of the U.S. public today would want to use nuclear weapons against an enemy that attacked the United States even when presented with a diplomatic option to end the war. For this significant portion of the U.S. public, there is no atomic aversion. Atomic attraction prevails.

The novel survey experiments described in this article, by re-creating the trade-off between an atomic attack and the risk to the lives of U.S. soldiers that the United States faced at the end of World War II, produced many surprising findings. The U.S. public's willingness to use nuclear weapons and deliberately kill foreign civilians has not changed as much since 1945 as many scholars have assumed. Contrary to the nuclear taboo thesis, a majority of Americans are willing to support the use of a nuclear weapon against an Iranian city killing 100,000 civilians. Contrary to the theory that Americans accept the noncombatant immunity norm, an even larger percentage of the U.S. public was willing to kill 100,000 Iranian civilians with conventional weapons. Women are as hawkish as men and, in some scenarios, are even more willing to support the use of nuclear weapons. Belief in the value of retribution is strongly related to support for using nuclear weapons, and a large majority of those who favor the use of nuclear weapons against Iran stated that the Iranian people bore some of the responsibility for that attack because they had not overthrown their government.

Future research will be necessary to determine whether these findings hold only for Americans or whether they are generalizable to the citizens of other countries. The U.S. public may be an outlier with regard to supporting nuclear weapons use, as it is an international outlier with respect to support for the death penalty. A 2016 public opinion poll in the United Kingdom, for example, found that approval ratings for President Truman's decision to drop the atomic bombs on Japan are much lower than those in the United States (28 percent in the United Kingdom compared with 45 percent in the United States).⁸⁸ It would be especially useful to focus on public attitudes about nuclear weapons use among key U.S. friends and allies. It would be valuable to know, for example, how the Israeli, the British, or the French public views potential uses of nuclear weapons in scenarios, such as those analyzed above, in which they are contemplating trade-offs between saving their own soldiers versus killing foreign noncombatants. Future survey experiments could provide new information not only about allied publics’ views regarding potential uses of their own nuclear arsenals, but also about whether these publics would support or oppose the use of U.S. nuclear weapons in specific scenarios, which could be an important consideration for U.S. leaders contemplating nuclear use.

Further research will also be needed to determine whether the U.S. public maintains these priorities under other wartime or crisis scenarios. It is possible that the widespread hostility that the American public harbors toward Iran helped produce these findings, and in a war against other foreign powers against which there is less hostility, there would be less willingness to use nuclear weapons or kill noncombatants. Still, it seems likely that the U.S. public would be hostile toward the population of any foreign state that is at war with the United States. Yet it is also possible that the U.S. public would feel differently about killing foreign civilians in a war that the United States initiated instead of one sparked by an enemy surprise attack. It would also be valuable to know how strongly U.S. public attraction to nuclear use in our experiments was influenced by the fact that the United States’ adversary in the war was an Islamic state in the Middle East. One could explore these effects through new experiments that vary the religious and racial identity of potential victims of a U.S. nuclear attack.⁸⁹

We have been careful in these experiments not to “prime” the respondents in ways that might bias the results. For example, in our stories, we did not mention the possible environmental effects of nuclear weapons. We described the victims as “civilians” rather than “innocent women and children.” We mentioned only “immediate deaths and long-term fatalities” from the nuclear attack, and did not describe the gruesome details of fatal burns or radiation sickness. We did not raise the possibility that an attack targeted against a city as a “shock strategy” would violate both the laws of armed conflict and U.S. nuclear weapons employment guidance, and thus would also likely be opposed by many senior U.S. military leaders.⁹⁰ Nor did we expose subjects to cues from political elites who opposed the bombing.⁹¹ All these factors could influence public support for or opposition to the use of nuclear weapons or violations of the noncombatant immunity norm in the real world and should be studied in the future.

In the final analysis, our survey experiments cannot tell us how future U.S. presidents and their top advisers would weigh their options if they found themselves in a conflict in which they faced a trade-off between risking large-scale U.S. military fatalities and killing large numbers of foreign noncombatants. Nevertheless, these surveys do tell us something unsettling about the instincts of the U.S. public concerning nuclear weapons and noncombatant immunity. When provoked, and in conditions where saving U.S. soldiers is at stake, the majority of Americans do not consider the first use of nuclear weapons a taboo, and their commitment to noncombatant immunity in wartime is shallow. Instead, a majority of Americans prioritize winning the war quickly and saving the lives of U.S. soldiers, even if that means killing large numbers of foreign noncombatants.

Both just war doctrine and the laws of armed conflict require leaders and soldiers to make active efforts and accept risks in war to avoid the deaths of foreign civilians. Michael Walzer's “doctrine of double intention,” for example, requires that soldiers not just intend to attack only legitimate military targets, but that they also take active measures to minimize unintended collateral damage, including accepting at least some risk to themselves.⁹² With respect to international law, Article 57 of the Geneva Protocol I demands that all signatories “take all feasible precautions in the choice of means and methods of attack with a view to avoiding, and in any event to minimizing, incidental loss of civilian life.”⁹³ Although the “principle of feasible precaution” may require military professionals to take some personal risk to protect noncombatants or require political leaders to take some risk of the loss of soldiers to protect foreign noncombatants, there is no agreement on the proper risk ratio. How many soldiers’ lives on one's own side is it appropriate to risk to reduce the risk of the unintended killing of foreign noncombatants?⁹⁴

We were not surprised by the finding that most Americans place a higher value on the life of an American soldier than the life of a foreign noncombatant. What was surprising, however, was the radical extent of that preference. Our experiments suggest that the majority of Americans find a 1:100 risk ratio to be morally acceptable. They were willing to kill 2 million Iranian civilians to save 20,000 U.S. soldiers. One respondent who approved of the conventional air strike that killed 100,000 Iranian civilians candidly expressed even more extreme preferences regarding proportionality and risk ratios, while displacing U.S. responsibility for the attack onto the Iranian people: “I would sacrifice 1 million enemies versus 1 of our military. Their choice, their death.”

U.S. political leaders have, in some important cases in the past, been aware of public sentiments regarding retribution and revenge and have used the threat of public pressure in favor of nuclear attacks to add credibility to thinly veiled nuclear threats. President George H.W. Bush, for example, wrote to Iraqi President Saddam Hussein in January 1991 that “the United States will not tolerate the use of chemical or biological weapons…. The American people would demand the strongest possible response.”⁹⁵ Secretary of State James Baker amplified the message in a meeting with Iraqi Foreign Minister Tariq Aziz: “If the conflict starts, God forbid, and chemical or biological weapons are used against our forces, the American people would demand vengeance. We have the means to exact it.”⁹⁶ Although scholars now know that the Bush administration had already decided not to use nuclear weapons to respond to any Iraqi chemical or biological weapons attack, Saddam Hussein did not know that and took the threat of U.S. nuclear weapons use seriously.⁹⁷ Our survey experiments demonstrate that such public pressures to use nuclear weapons are not fanciful and should be taken seriously by both U.S. leaders and any foreign government contemplating war against the United States. Indeed, these experiments suggest that pressures for escalating violence, including a public demand for vengeance and pressure to use nuclear weapons, extend beyond scenarios in which the United States is responding to nuclear, chemical, or biological attacks.

Past surveys that show a very substantial decline in U.S. public support for the 1945 dropping of the atomic bombs are a misleading guide to how the public would react if placed in similar wartime circumstances in the future. It is fortunate that the United States has not faced wartime conditions in the nuclear era in which U.S. political leaders and the public had to contemplate such grave trade-offs.⁹⁸ Today, as in 1945, the U.S. public is unlikely to serve as a serious constraint on any president who might consider using nuclear weapons in the crucible of war.

The authors thank Barton Bernstein, Joanne Gowa, David Holloway, Maral Mirshahi, Andrew Reddie, Adam Roberts, Sarah Sadlier, Todd Sechser, Rachel Stein, Nina Tannenwald, Allen Weiner, and William Wohlforth for helpful comments on earlier drafts of this article. They also thank participants in seminars in which they presented this work at Stanford University, Yale University, the University of Virginia, the George Washington University, the University of California, Berkeley, the Massachusetts Institute of Technology, and the London School of Economics. Funding for this project was received from the MacArthur Foundation, Stanford's Freeman Spogli Institute, Stanford's McCoy Family Center for Ethics in Society, and the Naval Postgraduate School's Project on Advanced Systems and Concepts for Countering Weapons of Mass Destruction.