Public Opinion on Nuclear Energy and Nuclear Weapons: The Attitudinal Nexus in the United States

Does a lack of enthusiasm for nuclear energy among the US public relate to connections with nuclear weapons? The United States is projected to decline as the global leader in nuclear power production due to stagnation in the industry. Alongside this trend, US public skepticism of nuclear energy has remained high for decades. Policymakers, industry, and pro- and anti-nuclear advocates must understand these attitudes given electricity needs, energy sector competition, and climate change. Existing scholarship explains public opinion on nuclear power by comparing its economic and environmental desirability with other energy sources. Yet, a critical area of public opinion remains understudied: the connection between nuclear energy and nuclear weapons. Scholars have theorized such a relationship in the public consciousness, but the premise has not been systematically investigated. Two studies examine this theoretical linkage. Study 1 uses a nationally representative US survey experiment (n = 904) to validate that public attitudes toward civilian and military nuclear technologies are indeed related. Study 2 uses an exploratory survey experiment (n = 1,003) to show that persuasive information about each technology affects opinion about the other. These studies provide evidence of psychological linkage. In fact, attitudes toward nuclear weapons may even drive those on nuclear energy.


Introduction
According to the International Atomic Energy Agency, 442 nuclear power reactors in 30 countries provide over 2,560 terawatt hours of electricity annually [1]. 1 This figure accounts for approximately 10% of global electrical power [2]. Despite the 2011 Fukushima Daiichi nuclear disaster in Japan, 53 reactors are under construction in 19 countries [1]. 2 Meanwhile, pro-nuclear groups advocate for nuclear energy as a solution to human-induced climate change and shrinking fossil fuel supplies [3][4][5]. Major environmentalist groups and various experts disagree, arguing that rapid nuclear expansion would heighten accident risks and long-term waste disposal problems [6][7][8][9][10].
In the United States, the domestic nuclear industry has faced opposition for decades [11][12][13]. Still, the country remains the world's largest nuclear energy producer and consumer. The 96 US commercial power reactors produced approximately 807 terawatt hours of electricity in 2018-roughly 19.3% of US electricity and 24.5% of global nuclear electricity generation that year. 3 However, the United States had 111 commercial power reactors in 1990, 15 more than today. Several facilities are also expected to close in coming decades as their Nuclear Regulatory Commission licenses expire; only two plants are under construction [1]. In addition to constraints posed by climate change and energy sector competition, US energy consumption is expected to increase substantially by 2050 [14]. Policymakers, industry, and pro-and anti-nuclear advocates will have to debate the role of nuclear power in the US energy mix.
As these players discuss nuclear energy, public opinion will be important alongside economic and technical considerations. American public attitudes mirror the downward trend in reactor construction and have remained lukewarm since the 1970s amid environmental, security, and economic critiques [11,[15][16][17][18]. Vocal opposition arose again after Fukushima [19,20]. While support has rebounded somewhat, most polls indicate a majority of US residents oppose building commercial power reactors and producing nuclear energy [21,22].
Informative, but generally dated, public opinion studies point to construction costs ( [23]; see also [24] for an updated treatment) and safety concerns such as waste storage [12,15,25,26] as drivers of skepticism. Others highlight the influence of the environmentalist movement on public opinion and institutional decision-making [12,25,27].
Notwithstanding these rationales, a critical area of public opinion remains understudied: the relationship between nuclear power and nuclear weapons. This tie has been central to policy in the nuclear age, as technologies used in nuclear power generation can contribute to nuclear weapons development. Today, nine countries possess nearly 15,000 nuclear weapons [28], which could be used with catastrophic effects. Public concerns about links between nuclear energy and the spread of nuclear weapons to additional states also remain salient [29][30][31]. Yet, little academic attention has been devoted to possible attitudinal connections. Scholars have theorized a relationship between attitudes toward nuclear weapons and nuclear power [32,33] with predominantly qualitative empirical studies [34,35]. While this research is informative, a paucity of quantitative and experimental analyses obscures the nature of attitude formation on nuclear energy in the United States.
We present two survey-based studies examining causal connections between public opinion on civilian and military nuclear technologies. 44 The results display evidence of linkage and suggest that attitudes toward nuclear weapons may drive attitudes toward nuclear energy. Our findings present an opportunity to examine the nature and origins of these attitudes and carefully consider their implications.

Literature review
Scholars of politics, public opinion, psychology, and environmental studies have investigated US attitudes toward nuclear energy for decades. We survey historical trends alongside theorized drivers. The literature thoroughly considers public opinion, with one puzzling omission: the interaction of attitudes toward nuclear energy and nuclear weapons.

Historical US public attitudes on nuclear energy
Current public indifference and skepticism toward nuclear power contrasts with earlier optimism. Following World War II, US policymakers sought to promote commercial nuclear power globally, distinguishing the "peaceful atom" from the weapons used in Hiroshima, Nagasaki, and the Cold War arms race with the Soviet Union [34,35]. This push, spearheaded by the Atomic Energy Commission, met with great success. US scientific cooperation agreements oversaw export of nuclear materials and energy production technologies to 30 countries [36,37]. Initially, public favorability toward domestic nuclear power dovetailed with increasing US exports. Opinion polls in the 1950s and 1960s showed remarkably high approval of nuclear power [25,[38][39][40]. Attitudes regarding construction near a poll respondent's residence were also positive, with multiple surveys showing majority or plurality approval [15].
Opinion surveys conducted through the mid-1970s show rapid declines in support from 1975 onward [15, pg. 30, Figure 3.4; 41, pg. 375, Figure 2]. Support particularly declined following the Three Mile Island (1979) and Chernobyl (1986) accidents. Although in-principle support for building nuclear plants remained high until the 1980s, major polls even showed signs of increasing opposition pre-dating Three Mile Island [15]. This shift was particularly dramatic when polls asked about local construction, a different topic than general support for nuclear energy. By 1976, Gallup and Harris polls indicated a majority of US respondents opposed siting nuclear plants "within five miles" of their locale, with opposition reaching 60% by 1979 [15,42,43]. Opposition soon expanded into a broader coalition of citizens concerned about nuclear power. Since the 1980s, nearly every survey by a major polling organization has shown a majority or plurality of Americans oppose nuclear energy [41,44].
More recently, a Pew survey conducted two weeks after Fukushima showed 52% of the US population opposed "the increased use of nuclear power," with only 39% in favor. Shortly thereafter, a Washington Post-ABC poll indicated 64% of Americans did not support construction of additional reactors in the United States [19,44,45]. But consistent with trends observed after Chernobyl and Three Mile Island [15,[46][47][48][49][50][51], approval rebounded to pre-accident levels the following year: 44% in favor versus 49% opposed [52].
US public attitudes remain uncertain several years after Fukushima. Results released in 2016 by the Nuclear Energy Institute, which lobbies the US government for the nuclear industry, showed 67% public approval of nuclear energy [53]. Surveys promoted by pro-nuclear groups [53][54][55][56] conflict sharply with declining support observed in mainline polls. Contemporaneous polls by Gallup and Pew in 2016 showed 54% opposition to domestic nuclear energy use [21,22]. Other surveys from preceding years yielded comparable results [57,58].

Theorized drivers of downward trends
Existing research on US public opinion toward nuclear power points to three potential drivers of downtrending support. First, high costs of nuclear energy relative to fossil fuels may have decreased its economic desirability. In particular, the discovery of extensive US natural gas reserves made that fuel source substantially cheaper than nuclear power. As these reserves became available to consumers, they reduced the relative appeal of nuclear fuel for ensuring energy security [31,59]. Today, nuclear energy remains considerably more expensive than competing fossil fuel sources per kilowatt hour of electrical power generation [60].
Second, negative perceptions of safety-especially regarding accidents and radioactive discharges-decreased interest from utilities, insurers, investors, and consumers. These attitudes may have been influenced by opposition to spent fuel reprocessing and a focus on proliferation risks during the administration of President Jimmy Carter [61]. As noted, support had begun to decline prior to Three Mile Island and Chernobyl, a phenomenon some attribute to the environmentalist movement of the late-1960s [12,25,62,63]. Movement actors opposed siting of plants, promoted referenda against new construction, and challenged the Atomic Energy Commission and its reporting on nuclear risks in highly publicized congressional hearings. Research suggests these actions weakened the Congressional Joint Committee on Atomic Energy that defended the nuclear industry and destabilized broader institutional support for the Atomic Energy Commission [12,25,27,35].
Third, concerns regarding waste storage may have ruled out nuclear energy as a safe, long-term fuel option. While some experts maintain that nuclear waste storage is technically feasible, 1970s polls revealed a dubious public [15]. More recent studies indicate this remains the case. Adamantiades and Kessides [64, pg. 5161] contend that while various "solutions for the safe storage of spent fuel exist and seem to be accepted by most technical specialists and observers...institutional obstacles remain formidable, and the public's aversion to having a repository in its 'backyard' continues" (see also [65]). However, public concerns are reasonable given a lack of scientific consensus on solutions and the costly time scales over which waste remains dangerously radioactive [59]. Historical management and engineering failures at waste sites, leading to leaks, serve to underline issues with expensive geologic disposal projects (see [66,67] on leaks; see [68] on costs). Ramana [69] describes how these failures have driven public concern, along with mistrust of the authorities responsible for ensuring safety of waste disposal. Rather than a symptom of "not in my backyard" attitudes, Ramana [69, pg. 7] contends that opposition ultimately "stems not just from a desire to avoid personal harm, but also from the ethical conviction that no community should be subjected to such risks."

Missing explanations and new directions
The above explanations do not completely account for observed changes in public opinion. For instance, most findings used to justify these explanations stem from surveys that do not investigate underlying sources of attitude formation. Studies also often assume that attitudes shift solely in response to information regarding nuclear power's costs relative to other energy sources. This precludes systematic examination of the possibility that attitudes toward nuclear energy may be influenced by attitudes toward nuclear weapons. If such attitudes are indeed linked, ignoring this perspective could yield ineffective policymaking.
It is important to remember that nuclear technologies are dual-use and offer opportunities for energy production and weapons development [29][30][31]. Early civilian nuclear energy programs, like those of the United States and the Soviet Union, were byproducts of classified initiatives to build atomic weapons. Low-enriched uranium consisting of 3-5% isotope U-235 can fuel a nuclear reactor, but a highly-enriched level of 90% U-235 is considered "weapons-grade." A critical mass enriched to this higher level can sustain a fission chain reaction to power a bomb. Further, radioactive decay of enriched uranium in a nuclear reactor produces plutonium (typically isotope Pu-239). Reprocessing spent reactor fuel can recover unused plutonium to serve as fissile material for nuclear weapons [59].
Enrichment and reprocessing thus enabled the Cold War nuclear arms race and risk of cities being destroyed in minutes by nuclear-tipped ballistic missiles. 5 Numerous studies have addressed the psychological effects of the arms race on publics [32,34,35,[72][73][74]. Elder generations of Americans, after all, grew up with "duck-and-cover" drills preparing for the use of nuclear weapons. Substantiating these anxieties, scholars have chronicled and publicized close calls and near-accidents involving nuclear weapons from the 1962 Cuban Missile Crisis to the modern era [75][76][77]. Many environmental groups opposing nuclear energy have also engaged in activism that some argue spurred arms control and established taboos against nuclear weapons [78][79][80][81][82]. Historical polls cataloged by the Roper Center for Public Opinion at Cornell University indicate that US support for nuclear weapons in the 1950s dissipated as the Cold War dragged on. More recent polling reveals that-with nine nuclear-armed states-86% of Americans want action on nuclear disarmament [83]; 65% support permanently banning US nuclear weapons testing [84].
Civilian nuclear energy technologies also are potential facilitators of nuclear weapons proliferation to new states. Nuclear power programs can be instrumental in the accumulation of technical knowledge [37] and diversion of fissile materials [85] for use in parallel military nuclear weapons programs. Indeed, beyond the nine nuclear-armed states, an additional 26 have explored development of nuclear weapons [86]. Recent scholarship has shown that these countries may leverage civilian nuclear programs as a threat to build nuclear weapons [87][88][89]. Considering the dual-use dilemma, the spread of this so-called "nuclear latency" may combine with the risk of nuclear weapons use to cast a long shadow over public opinion on nuclear energy.
A potential link in American public attitudes toward civilian and military nuclear technologies could be impactful, especially given fear and dread often associated with nuclear energy [90]. Given the intertwined history of nuclear weapons and nuclear power, it is imperative to dissect their connections in the US public consciousness. Are they seen as two sides of the same coin? Do discussions of reactors evoke images of mushroom clouds?
Prior studies suggest, but do not explore, the possibility of psychological linkage mediating attitudes toward nuclear energy [25,61,[91][92][93][94]. Plous [95] even offers laboratory studies measuring attitudes toward both technologies together. However, these findings have limited reliability since subjects were not randomly assigned to treatment. Instead, subjects were assigned to study groups based on previously expressed nuclear attitudes. Consequently, these groups were imbalanced on key demographic covariates known to correlate with nuclear attitudes, such as gender and political identification-substantiating selection bias concerns.
Other work has advanced theoretical and historical findings about this relationship. Pahner [32], for instance, examines the connection psychologically. The historical and archival analyses provided by Boyer [33], and especially by Weart [34,35], explore this relationship extensively. Weart [35] specifically argues that public fears of nuclear power derive from imagery that painted nuclear weapons and nuclear war as forbidden, apocalyptic violations of nature. These accounts do not, however, systematically test their observations through original survey research or extensive analysis of public opinion research. Both Boyer [33] and Weart [35] cite existing national polls on each technology, but neither links polls on one technology to those about the other.
Polling gaps are obvious when examining historical US surveys. Fig. 1 displays results from a systematic search of the Roper Center iPOLL Databank for questions pertaining to nuclear power, nuclear weapons, and their intersection. It is telling that an essential resource for high-quality public opinion polling on nuclear topics contains barely any questions considering military and civilian nuclear applications together. Over seven decades of available data (3,949 questions), nearly 60% of questions relate exclusively to nuclear weapons, compared to approximately 39% about nuclear energy. Questions asking about both technologies account for just 1.4% of polling.

Methods and data
Our first study (Study 1) to investigate the attitudinal linkage between nuclear power and nuclear weapons was based on questions we  5 There is, more generally, an interdependence between civilian nuclear energy investment and nuclear-armed states' weapons development programs [70], as well as the production of nuclear-powered submarines that may carry ballistic missiles [71].
included in the 2016 Cooperative Congressional Election Survey. This was a nationally representative internet survey of voting-age US citizens conducted by the international polling firm YouGov from September 28-November 7, 2016. 6 Our questions ( = n 904) asked subjects to indicate attitudes toward nuclear energy and nuclear weapons. Attitudes were measured using two separate attitudinal "batteries"-a series of questions gauging subjects' views with respect to each technology. These questions are presented in Table 1. Each battery contained four questions about attitudes toward the given technology in different contexts: 1) continued research and development (R&D); 2) living near a nuclear site; 3) use of the technology to support US interests; and, 4) environmental impacts of continuing R&D. Question ordering within each battery and the order of the batteries themselves were randomized to avoid ordering effects.
Responses to each battery were reported attitudes toward the technology under discussion. Subjects expressed responses to each question on a seven-point Likert scale, ranging from "very negative" (1) to "very positive" (7). 7 We preserved the content and wording of questions as much as possible across batteries, allowing reliable assessment of cross-domain attitude correlations. In addition, we randomized battery order for identification of treatment effects of ordering on correlations and attitudes. We also calculated average correlations via the 4 × 4 pairwise matrix of Pearson correlation coefficients for all attitudes. Table 2 presents mean attitudes and their differences, pooling over question order (i.e., including subjects who saw the weapons battery first and those who saw the energy battery first). Bootstrapped standard errors for each mean estimate are reported in parentheses. On average, subjects' attitudes were consistent across technologies. There was one minor exception pertaining to nuclear R&D. The extent to which subjects were more positive toward nuclear energy R&D than nuclear weapons R&D was statistically significant (p < .050) but not substantively large.

Results
Responses were generally neutral toward both technologies across the four activities, presenting a more nuanced picture of attitudes than previous polls asking respondents to indicate support with a single answer. 8 Average responses regarding nuclear R&D and use of nuclear technologies to promote US security hovered around the midpoint of our Likert scale. Subjects were also only marginally concerned about environmental impacts. At first glance, neutrality might seem to suggest indifference to nuclear technologies, were it not for the notable exception of opposition to living near a nuclear site-regardless if military or civilian. Preferences against living near a site suggest subjects are concerned with nuclear technologies insofar as they raise "not in my backyard" concerns that have been salient for decades [105,106].
Importantly, respondent attitudes were strongly correlated across civilian and military technologies. Table 3 presents the 4 × 4 matrix of Pearson correlation coefficients, computed using all complete pairwise observations of attitudes. Nuclear energy topics appear as rows, while corresponding nuclear weapons topics appear as columns. Values on the diagonal represent the correlation of attitudes within contexts but across technologies. The results indicate strong, positive correlations among all attitudes, especially within context areas. The correlation between attitudes toward living near a civilian or military nuclear site was particularly pronounced.
Cross-domain correlations were also strong across topics. For Domain order and row-ordering of contexts within domain were randomized.  [96]. Surveys are nonetheless known to have limitations that can affect the credibility of results, including dishonesty in selfreported responses, framing effects [97], and non-response bias [98]. However, our design and analysis limit these concerns. For instance, to address demographic representativeness, we employ survey weights provided by YouGov. Effect estimates relying on random assignment of question order also allow us to measure credible causal effects. For additional discussion of similar and alternative methodological approaches to our own in the domain of energy policy research, see [99]. 7 These batteries do not exhaustively present all contexts related to the two technologies-such as nuclear waste storage-and contain slightly varied phrasing. They nonetheless provide an informative measure of attitudes and have high validity: the estimated Cronbach's = 0.87 for the power questions, and = 0.83 for the weapons questions. This suggests that a subject's responses to the batteries would likely predict their attitudes in other relevant contexts.
instance, positive attitudes toward continued nuclear weapons R&D were predictive of positive attitudes about nuclear energy use, living near a nuclear power plant, and the environmental impacts of nuclear energy. Of course, correlation alone does not show that attitudes are causally interdependent. Yet, these results do suggest an interdependence, as if attitudes regarding military and civilian nuclear technologies each originate from a shared belief system [107,108]. The randomized battery order in Study 1 allowed us to investigate this possibility: If attitudes toward one technology drive attitudes more generally, then subjects primed to think of that technology first should display a stronger association between technologies. A change in question ordering could also affect attitudes themselves. Our analysis suggested that considerations about nuclear weapons likely drive the attitudinal relationship between technologies. For good measure, we first assessed this possibility using bivariate and multiple ordinary least squares regressions to estimate whether showing respondents the weapons battery first affected attitudes toward each technology. The results showed that battery order had no effect on attitudes in any model specification. This is not unexpected, as battery order alone constitutes a weak treatment. We would not expect the question ordering to systematically alter public attitudes toward technologies. However, order did affect the average correlation between attitudes, implying that respondents who saw the weapons-related questions first drew a stronger association between the two technologies (Table 4). 9 Table 4 shows the pairwise correlations of all questions, conditional on weapons battery being seen first (orange) or second (blue). The average correlation of "weapons first" responses (values in the black-dashed box in the lower left of  10 and the result is significant (p < .050). There was no effect from placing nuclear power questions first, suggesting that attitudes on nuclear weapons may be the primary driver of an association across technologies.
We observed additional evidence of a relationship between nuclear attitudes in other correlates from the 2016 Cooperative Congressional Election Survey. Table 5 shows that respondents who voiced stronger support for environmental regulation-usually Democrats-were likely to view both nuclear technologies negatively. The overall importance that respondents assigned to environmental issues (column 1 of Table 5) was the most effective predictor of nuclear attitudes. Among specific policy preferences (columns 2-5 of Table 5), support for giving the Environmental Protection Agency power to regulate carbon dioxide emissions was particularly prognostic of opposition to nuclear technologies, but responses to almost all environmental questions predicted such opposition. This correlation is revealing of one of the main debates on nuclear power in the United States. Although nuclear energy has been proposed by pro-nuclear groups as a solution to global climate change, environmentalists usually oppose its expansion due to accident risks and waste storage concerns. Environmentalism also correlates strongly with opposition to nuclear weapons, again suggesting these attitudes result from a single latent belief system.
One important limitation of Study 1 is its focus on nuclear activities in the United States. Study 1 offers insights into US public concerns about the connection at home between nuclear power and weapons use by an established nuclear-armed state. However, Study 1 does not provide an adequate means of assessing Americans' fears about the spread of nuclear weapons to new states or various activities in nucleararmed states that may have adversarial relations with the United States.

Methods and data
Study 1 revealed that correlations between attitudes were stronger when subjects viewed the nuclear weapons battery first. Still, there was no direct effect of battery order on attitudes. To assess whether it is possible to shift attitudes regarding these technologies in principle, we fielded Study 2 on January 24, 2020. Study 2 involved a survey experiment ( = n 1, 003) conducted with a convenience sample of respondents recruited through the Amazon Mechanical Turk online labor market. 11 Experimental research designs such as these have become a common tool of causal inference in the social sciences. The rationale is that their randomized, controlled nature allows researchers to cleanly identify causal relationships without concerns about confounding [109,110]. Survey experiments combine these benefits with the use of polling methods to measure causal effects on public opinion and attitudes [111]. 12 Mechanical Turk is widely used in social science research to sample respondents for survey experiments and other studies. Scholars have shown that its demographically diverse samples perform more similarly to nationally representative US samples than do laboratory convenience samples [112][113][114]. Other work suggests similarities between treatment effect estimates in convenience samples and nationally representative samples [115]. Nonetheless, concerns of external validity may still persist because the individuals recruited in a convenience sample cannot generally be expected to be nationally representative. Accordingly, while we believe these results highlight broader associations within the US public, our findings remain exploratory.
The survey followed a 2 × 2 factorial design, whereby subjects were randomly assigned to read either a pro-or an anti-nuclear (first factor) vignette on nuclear weapons or nuclear power (second factor). After reading the assigned vignette, subjects answered both batteries from Study 1 in random order. We measured treatment effects using their responses. To simplify analysis, we averaged each subject's responses to the questions in each battery. This produced attitude indices, which we refer to as the "power approval index" (Cronbach's = 0.89) and the "weapons approval index" (Cronbach's = 0.86).
Respondents also viewed a free association question on both correlation matrices, conditional on domain order. We then computed the difference in these means to assess effects, along with 95% bootstrapped confidence intervals computed with 100,000 replicates (hereafter CI). 10 This CI does not intersect with 0.000, but it appears due to rounding. 11 This survey represented a confirmatory trial following two exploratory surveys (combined = n 455) conducted in 2017. The confirmatory trial, preregistered as Evidence in Governance and Politics pre-analysis plan 20200124AB, corroborated the substantive findings of these earlier studies. 12 Gaines et al. [111] note some survey experimental practices that could yield problematic inferences like not measuring effect duration and only measuring effects of "one-shot" treatments. These limitations are somewhat relevant to our study, which due to budgetary constraints, does not re-sample respondents or assess effects of multiple treatments over time. However, the results presented in Study 2 are from a confirmatory trial based on earlier pilot studies. While the confirmatory trial and pilots were conducted over two years apart, the substantive findings successfully replicated.
surveys, asking them to "Describe the first image that comes to mind when you see the word 'nuclear.'" We used responses to this question to produce three dummy variables indicating destructive, neutral, and constructive associations with the word "nuclear." Neither "destructive" nor "constructive" codings imply normative value judgments. For instance, a subject responding with the word "bomb," without further qualification, could associate nuclear bombs with harmful radioactive fallout, or with safety from nuclear deterrence. However, bombs themselves are clearly a destructive technology. 13 All ambiguous responses were coded as neutral.
In both experiments, pro-nuclear vignettes described the virtues of technologies as put forth by advocates. The pro-nuclear weapons treatment argued that nuclear deterrence reduces the chances of war between nuclear-armed adversaries, whereas the pro-nuclear power vignette contended that nuclear power is more environmentally friendly than fossil fuels. The anti-nuclear power treatment followed a similar format and discussed high economic and environmental costs associated with nuclear energy generation. The anti-nuclear weapons treatment used the translated eyewitness account of a Nagasaki bombing survivor followed by a statement that the most powerful nuclear weapon ever built was approximately 1,500 times more powerful than "Fat Man," the bomb dropped on Nagasaki. This treatment asymmetry was motivated by the desire to field the strongest possible interventions for exploratory purposes. Study 2's design did not prime subjects on potential relations between civilian and military nuclear technologies. In other words, nuclear energy vignettes did not mention nuclear weapons and vice versa. Thus, our survey experiment enabled us to generate inferential leverage on any underlying psychological linkage that respondents observe-either overtly or subconsciously-between these technologies. Table 6 presents the results of Study 2 using bivariate and multiple ordinary least squares regressions to estimate attitudinal effects of assigning a respondent to read an anti-nuclear vignette (measured on a seven-point Likert scale, as in Study 1). The key predictor of interest is an indicator (dummy) variable of whether a respondent was randomly assigned to read an anti-nuclear vignette. Controls include age, female gender identification, and three-point partisan identification (-1 for Democrats, 0 for Independents, and 1 for Republicans). Because each survey featured only two treatment arms (pro-and anti-nuclear vignettes), Table 6 shows only results of anti-nuclear treatments on attitudes. Pro-nuclear treatments yielded similar effects that were positive and highly significant. Model specifications (hereafter "specifications") (1)-(4) show effects of the anti-nuclear power treatment on attitudes, whereas specifications (5)-(8) indicate effects of the anti-nuclear weapons treatment. Anti-nuclear treatments in both surveys had strong and significant negative effects. Table 6 also shows the relationship between relevant demographic covariates and attitudes carried over from Study 1. Three-point party identification was significantly and positively correlated with outcomes in all specifications except specification (4). That is, conservative respondents tend to be more positive toward both nuclear power and nuclear weapons. Although less robust, we also saw that female respondents were typically more likely to view nuclear technologies negatively. This result reflects a persistent finding in studies of nuclear energy [15,[116][117][118][119].

Results
These findings suggest that persuasive informational treatments can have strong effects on nuclear attitudes-including across technologies. In conjunction with Study 1's results, the observed effects of nuclear weapons vignettes on attitudes toward nuclear energy, and vice versa, provide an empirical basis for studying cross-domain nuclear attitudes.
Responses to the free association question were also striking, indicating that subjects overwhelmingly associated the word "nuclear" with destruction. This result provides further evidence of interdependence and reveals that public associations of nuclear technologies with destruction exist even absent priming about nuclear weapons. As Table 5 Relationship between environmental attitudes and nuclear attitudes. Weapons questions and energy questions are respectively subscripted "W" and "E." The question in column 1 of Table 5 asked respondents to rank the policy importance of the environment on a five-point Likert scale. All other questions asked about binary support or opposition to a specific policy proposal: "Give Environmental Protection Agency power to regulate Carbon Dioxide emissions"; "Raise required fuel efficiency for the average automobile from 25 miles per gallon to 35 miles per gallon"; "Require a minimum amount of renewable fuels (wind, solar, and hydroelectric) in the generation of electricity even if electricity prices increase somewhat"; "Strengthen enforcement of the Clean Air Act and Clean Water Act even if it costs US jobs." Table 4 Correlation matrix of weapons and energy questions by order.
Pairwise correlations produced when weapons questions (subscripted "W") were shown first are highlighted in orange. Pairwise correlations produced when energy questions (subscripted "E") were shown first are highlighted in blue. Cell saturation reflects the magnitude of the correlation. Black dotted lines circumscribe 4 × 4 correlation matrices used to compute average correlations. 13 References to nuclear weapons including specific commentary on the benefits of deterrence, however, would be coded "constructive." expected, 80.4% of respondents who viewed a weapons vignette noted destructive connotations. However, 68% of respondents who saw a nuclear power message also noted destructive associations. Fig. 2 displays the distributions of responses and highlights the extent to which respondents associate destructive imagery with the mere mention of "nuclear," signifying a belief that most, if not all, nuclear technology is inherently violent or dangerous. Relatively few responses referenced nuclear energy or suggested a familiarity with either technology. This trend was more extreme among respondents to the weapons survey. Just under 10.5% of these free association responses noted a constructive association with "nuclear," whereas approximately 21.5% of responses among those treated with power vignettes indicated constructive associations-almost all describing nuclear power plants.
All remaining responses were neutral.
Beyond showing similar connotations with the word "nuclear," we observed considerable crossover in the words respondents used to describe associations in each survey. Table 7 shows the ten most commonly used one-and two-word terms for each survey, measured as total uses divided by total responses. Notably, the specific terms most commonly used among all respondents were strongly evocative of nuclear weapons and nuclear war: "bomb," "explosion," "cloud"-often paired with "mushroom," as in the two-word term, "mushroom cloud." Less common were terms referencing peaceful uses of nuclear technology: "power," "plant," their co-occurrence "power plant," and "energy." Table 6 Ordinary least squares analysis of the effects of anti-nuclear treatments on attitude indices.  These terms associated with peaceful uses appear in only a small minority of responses. 14 Whereas Study 1 did not allow us to see the relation between nuclear energy and nuclear proliferation in the public eye, Study 2's free association question offered some relevant insights. The vignettes used in Study 2 did not have the limitation of specifically talking about nuclear technologies in a US context. That is, subjects could discuss nuclear issues in any domain that came to mind. A few respondents mentioned proliferation concerns like North Korea, and one even directly referenced diversion from civilian to military nuclear programs. However, such responses were unusual. Many respondents instead discussed nuclear weapons use (e.g., "mushroom cloud") or offered ambiguous responses that may have been motivated by use just as well as proliferation (e.g., "bomb").
Overall, our analyses reveal the strong confluence of associations with nuclear power and nuclear weapons. The reaction of one respondent in our confirmatory trial especially lucid (the present survey revealed similar sentiments): "Power, strength, bomb, clean energy. It's not a single image, but a collage of explosions and clean energy and power stations." While this was an atypically sophisticated response, the logic underlying the statement vividly portrays the critical policy significance of examining attitudes toward both technologies together. If perspectives on these technologies are indeed interdependent as we observe, then public opinion toward one should not be considered in isolation from the other.

Discussion and conclusion
Despite strong claims by some advocates that opposition to nuclear energy is irrational [121,122], civilian nuclear technology carries risks. A robust literature addresses perceptions of these risks, justifying attitudes and beliefs on cost/benefit and psychological grounds [90,116,[123][124][125][126]. Even normal nuclear power plant operations present potential health risks from ionizing radiation and high-level nuclear waste [127]. The shadow of nuclear accidents, including Fukushima, has contributed to doubts among the US public over the safety of nuclear power. More than three decades later, Chernobyl's legacy looms especially large, with publications [128,129] and television treatments of the disaster [130] continuing to receive mainstream attention. Deeper psychological fears may also play a role, with prior studies showing that nuclear technologies are associated with feelings of dread not found in response to other risky technologies, including sources of electric power like coal-fired power plants and other ionizing radiation sources like medical X-rays [90]. Slovic et al. [90] note that the risks associated with nuclear power are thought to be uncontrollable, involuntary, catastrophic, and relatively poorly understood by science. Erikson [131] even describes nuclear energy as evoking imagery of "contamination." It is revealing that this association between nuclear technology and contamination was common before the first US commercial nuclear reactor began operating in 1957 [132]. Weart [35, pg. 99] describes the entrenchment among the US public in the 1950s of the belief that nuclear weapons "violated the order of nature." According to Weart [35, pg. 99], such an "idea was bound up with one of the strongest of primitive themes: contamination. In most human cultures, the violation of nature, and indeed all forbidden acts or things, are identified with contamination." Boyer [33, pp. 142, 267] provides evidence that these motifs arose in US discourse almost immediately following the bombings of Hiroshima and Nagasaki. Indeed, nuclear fears appear to stem from deeper associations than those provided by relatively recent nuclear disasters alone. The attention given to accidents like Chernobyl may account for some of the graphic imagery of contamination conjured in discussions of civilian nuclear energy. These events, however horrific, cannot fully explain the projection of images first seen in relation to nuclear weapons onto civilian nuclear energy.
Our research supports this claim with systematic evidence from public opinion data showing that respondents continue to strongly associate civilian nuclear energy with military nuclear imagery, such as mushroom clouds and bombs. That is, our studies mark a new addition to the literature by focusing on the microfoundations of well-studied risk perceptions. They experimentally validate the theorized position that US nuclear attitudes draw from a fundamental interdependence between nuclear energy and nuclear weapons.
We accordingly show that another consideration should factor into debates over nuclear energy in the United States, namely, the connection between US public attitudes toward nuclear power and nuclear weapons. The existing literature hints at the possibility that many fears related to nuclear energy stem from the technology's military origins. Yet, we find that the overwhelming majority of US-based public opinion polls and studies ignore this potential connection. Our analyses provide an important step in examining this relationship. Studies 1 and 2 showed strong correlations between individuals' perceptions of nuclear energy and nuclear weapons. More importantly, both studies provided initial evidence of a causal link between Americans' attitudes toward nuclear technologies. In Study 1's nationally representative US survey, subjects showed a stronger association between both technologies when first asked about their attitudes toward nuclear weapons.
In fact, our finding suggests that perceptions of nuclear weapons may drive this connection. Study 2 demonstrated that Americans tend to relate even peaceful nuclear energy with destructive imagery, a linkage that appears to stem from associations with nuclear weapons. Given the historical influence of political action in changing public opinion on nuclear power [12,25,27,133], and ongoing persuasive efforts by advocates on both sides of the debate, policymakers would do well to be aware of such attitudes.
However, our results may be limited in their generalizability beyond the United States. While the United States is the world's largest producer and consumer of nuclear energy, it is just one of 30 countries with civilian nuclear power infrastructure. Additional countries may also invest in civilian nuclear power in the future. Countries-from Britain to China, France, Japan, Russia, and beyond-all have their own unique histories with nuclear power and nuclear weapons. Thus, it would be Table 7 Ten most frequent terms used in response to free association question: "Describe the first image that comes to mind when you see the word 'nuclear.'" (total uses / total responses). Hyphens indicate two-word terms.
14 It should be noted that asking respondents about an "image" may yield slightly different results than a simple word association (see, e.g., [120]). "Explosions," for example, are easier to visualize than "decreased carbon dioxide emissions." In another poll-on the 2018 Cooperative Congressional Election Survey-we posed the task as a word association. The results were substantively similar to the present study, although we observed a higher frequency of abstract terms such as "scary" and a lower frequency of graphic terms such as the two-word term "mushroom cloud." valuable for scholars to undertake further work in varied international contexts to assess cross-national similarities and differences with our findings. This could yield important policy implications for the regional and global trajectories of nuclear energy. Moreover, our findings present another potential new direction for future studies. To more comprehensively understand nuclear attitudes, scholars could emphasize inference aimed at pinpointing sources of existing nuclear attitudes. Such research would complement extant work offering descriptive analyses of nuclear risk perceptions. The existing literature tends to "black box" the sources of public risk perceptions of nuclear technologies, but abstracting away from their pivotal foundations could result in misunderstanding public concerns. Further examination of these foundations is required, and we have shown that modern social scientific experimental techniques can allow researchers to probe them effectively.

Funding
Study 1 was carried out as part of the 2016 Cooperative Congressional Election Survey with research funds from the Institution for Social and Policy Studies, Yale University. Study 2 received funding from the Department of Political Science, Yale University. Stephen Herzog completed this research with support from the Stanton Foundation.

Declaration of Competing Interests
The authors declare that they have no conflicts of interest to report.