International Studies Review (2017) 19, 646-666 ANALYTICAL ESSAY An Experimental Agenda for Securitization Theory StephaneJ. Baele and Catarina P. Thomson University of Exeter Securitization theory has developed into a fruitful research program on the construction of security threats. The theory has experienced growing sophistication, and empirical studies have produced stimulating insights on issues as varied as the politics of immigration, health, climate change, or cybersecurity. Understanding how social issues become perceived as threats seems timelier than ever given the rise in securitizing narratives in recent political elections across the globe. We propose that this research agenda would benefit from broadening its methodological diversity. In particular, the use of experiments could complement existing methods in securitization theory, mitigate some of the program's methodological weaknesses, and help explain when securitizing moves are likely to succeed or fail. Keywords: securitization, experiments, methods Introduction: Diversifying Securitization Research Methods with Experiments In the context of a rise in securitizing narratives in national and international politics—as visible, for example, in the depiction of foreigners and immigrants as threats in the recent Brexit campaign, US presidential election, or referendum in Hungary—understanding the determinants of foreigner acceptance/rejection by the public seems more crucial than ever. This article argues that experimental designs could strengthen our understanding of the factors that play a role in the failure or success of such narratives. Experiments are a useful tool to examine public reactions to elite discourses ,and as such, they could help understand the processes through which issues as varied as immigration, climate change, or epidemics come to be perceived by the public as security threats and thereby moved outside the realm of regular politics in order to be dealt with through extraordinary measures. First developed by Wsever (1995) and Buzan, Wsever, and De Wilde (1998), securitization theory explicitly deals with these questions and has developed into a rich research agenda on the dynamics through which political actors transform issues into security threats and the consequences of that type of transformation. As summarized by Williams, "in securitization theory ... the social construction of Acknowledgements: The authors wish to thank the journal editors and the anonymous reviewers for their extensive, detailed, and constructive comments. They are also grateful to P. Van Arsdale for her helpful comments on earlier drafts and to A. Mory for his support. Catarina P. Thomson gratefully acknowledges financial support from the UK Economic and Social Research Council (grant ES/L010879/1). Baele, Stephane J. and Thomson, Catarina P. (2017) An Experimental Agenda for Securitization Theory. International Studies Review, doi: 10.1093/isr/vix014 © The Author (2017). Published by Oxford University Press on behalf of the International Studies Association. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons. org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. An Experimental Agenda for Securitization Theory 647 security issues (who or what is being secured, and from what) is analyzed by examining the 'securitizing speech acts' through which threats become represented and recognized. Issues become 'securitized,' treated as security issues, through these speech acts which do not simply describe an existing security situation, but bring it into being as a security situation by successfully representing it as such" (2003, 513).1 Since its inception, a wealth of empirical contributions examining issues as varied as immigration, global health, or cyber-security have sought to test and refine an increasingly multifaceted theoretical framework. The theory now considers non-linguistic aspects ranging from the impact of images on threat perceptions (e.g., Hansen 2011), to the role of administrative/bureaucratic processes, routines, and practices in securitizing processes (e.g., Bigo 2002; Huysmans 2011), to the emotional dynamics associated with threat dissemination (Van Rythoven 2015). Securitization theory now also connects with adjacent disciplines like anthropology (Holbraad and Pedersen 2012) and psychology (Theiler 2003).2 Yet, in spite of its empirical development and increased theoretical refinement, the securitization research agenda has traditionally relied on a limited range of research methods (see, e.g., Balzacq 2011b, 31; Balzacq and Leonard 2011, 3). Baele and Sterck (2015, 1122) go as far as to claim that "methods are the Achilles' heel of securitization studies, casting doubt on their conclusions." While we do not share such a radical opinion, we nonetheless believe that securitization research would benefit from diversifying its methodological toolbox. In particular, we argue that a careful use of experiments3 could not only help mitigate general methodological shortcomings that tend to characterize securitization research, but also more importantly contribute to answer one of securitization's central questions—that of the determinants of the success or failure of attempts to securitize particular issues. Our objective is not to make a bold claim in favor of experiments, but rather to identify how these method can contribute to the development of securitization theory—so far, only Vultee (2010, 2011) has used an experimental approach to this end. We argue that the complementarity experiments and qualitative work achieved in other fields can also be fruitful in the securitization arena. To do so, we proceed in two steps. First, we explain why and how experiments could enrich securitization theory by mitigating its most apparent methodological shortcomings. Building on this reflection, we then focus specifically on what types of experimental designs can help identify the conditions under which securitizing moves succeed or fail. Experiments and Securitization Theory: Strengthening and Generalizing Causal Pathways, Reducing Biases Experimental methods have experienced an exponential growth in political science and more recently made their way into international relations 1 Securitization theory's original emphasis on speech acts comes from the use of John Austin's philosophy of ordinary language (1962), which claims that in some instances language does not simply describe things or states of affairs, but also does something through its very utterance—hence the concept of a "speech act'. The classical example of a speech act is that of a priest declaring two individuals to be married and through this very utterance creating the very reality of marriage. The same logic is claimed by securitization theorists to be at play in the realm of security, where seemingly straightforward descriptions of particular issues as threats do in fact contribute to these issues being perceived and acted upon as threats. 2Balzacq's definition of securitization (2011a; 2011b, 3) exemplifies this diversity; securitization, he writes, is "an articulated assemblage of practices whereby heuristic artifacts (metaphors, policy tools, image repertoires, analogies, stereotypes, emotions, etc.) are contextually mobilized by a securitizing actor, who works to prompt an audience to build a coherent network of implications (feelings, sensations, thoughts, and intuitions) about the critical vulnerability of a referent object, that concurs with the securitizing actor's reasons for choices and actions, by investing the referent subject with such an aura of unprecedented threatening complexion that a customized policy must be immediately undertaken to block it." 3See Baele 2017 for suggestions on how other quantitative methods and techniques such as automated language use analysis, mass content analysis, or network analysis could also contribute to the development of securitization theory. 648 Stephane J. Baele and Catarina P. Thomson (see Hudson and Butler 2010). Experimental research has cast a new light on some of the most classic questions of the discipline, many of which are directly related to security (e.g., Andreoni, Castillo, and Petrie 2003; Druckman and Broome 1991; or Roberts and Sherratt 2002 on cooperation; Geva, Mayhar, and Skorick 2000 on foreign policy decision-making; McDermott and Cowden 2001 or McDermott et al. 2007 on crisis behavior; Boettcher and Cobb 2006 or Thomson 2016 on public support for war and economic coercion; Nyhan and Reifler 2010 on misperceptions). As explained by McDermott (2002a, 2002b, 2011a) or Mintz, Yang, and McDermott (2011), experiments are undoubtedly destined to become an important component of evidence-gathering and theorization in international relations.4 Experiments are generally associated with quantitative approaches, which are not favored by securitization scholars and critical security studies in general.5 Balzacq's cornerstone methodological contribution (2011b), for instance, only lists qualitative methods as available to securitization scholars. Inspired by valuable contributions offered by recent and rare quantitative contributions to the securitization debate (Vultee 2010, 2011; Karyotis and Patrikios 2010; Lahav and Courtemanche 2011; Baele and Sterck 2015), we propose to move forward in a constructive way following recent calls for more methodological pluralism and quantitative-qualitative dialogue in constructivist international relations and critical security studies (e.g., Lake 2013; Barkin and Sjoberg 2017; Balzacq 2014). More specifically, we claim that experiments offer a distinct methodological advantage that can contribute to securitization theory: they naturally stand as a useful second step that directly compares and tests case-specific findings previously obtained through other methods. This advantage can help securitization scholars mitigate two interrelated weaknesses that are common across securitization research—general causal claims based on case-specific findings and selection/analysis biases—and thereby to answer the crucial question of the determinants of success or failure of securitizing attempts. First, experiments can help clarify the causal pathways identified by previous securitization research and in doing so assess how generalizable previous findings are. Securitization research has gone a long way in identifying key factors of theoretical interest. Based on these important findings, experimentation can now systematically allow a direct comparison of the effects of these factors. Evidence supporting securitization theory is so far made up of a wealth of case studies whose overlapping local findings aggregate to the point of producing a convincing broad picture —but not really to the point of constructing a robust and integrated body of knowledge.7 Experimentation can help transcend context- and case-specific results (obtained through methods usually employed to investigate securitization case studies, such as process tracing, content analysis, or discourse analysis) into robust general conclusions about the success/failure of securitizing moves,8 thus finding more generally valid causal mechanisms. When a researcher 4The same trend is even more noticeable in economics, where a methods handbook presents experiments as "the most credible and influential research design" (Angrist and Pischke 2008). 5None of the most recent and comprehensive methods handbooks in critical security studies (Salter and Mutlu, 2013; Aradau et al. 2015) contain chapters on quantitative methods. 6This clearly appears in Balzacq, Leonard, and Ruzicka (2016)'s presentation of securitization theory's main empirical achievements. See also Balzacq (2011b, 33) on the predominance of case studies in securitization research. It is in this way that Druckman and colleagues recently called to shape political science into "a discipline whose contributions are deepened and strengthened by experimental research" (Druckman et al. 2006, 634, our emphasis)." experiments in the social sciences are indeed particularly useful when combined with other methods, such as case studies, comparative analyses, or large data quantitative statistical work. This complementary role, to which we will come back later in the paper, is also emphasised by Falk and Heckman, who see experiments as "productive in complementing the information obtained from other empirical methods" (2009, 537); by Angrist and Pischke (2008), who assign them the role of generating general theorization on the basis of fragmented insights; or by An Experimental Agenda for Securitization Theory 649 controls the manipulation of the independent variables (and randomization of participants effectively takes place), we can say that experiments have the highest internal validity relative to other research methods; hence, the experimental method constitutes the prime method to conceptualize causal inference and guide theoretical development on the basis of already existing results gathered through other methods (Druckman et al. 2006). McDermott sharply states that, "no other methodology can offer the strong support for the causal inferences that experiments allow" (McDermott 2002a, 38). Even more crucially, experiments can also test the very assumptions and axioms on which theories are based—in the case of securitization, the existence of "facilitating" conditions of securitizing moves (such as the legitimacy of the securitizing actor; see Buzan, Wsever, and De Wilde 1998). Specific areas in which experiments could help clarify securitization causal pathways and transcend findings arrived through case study work include asking the following: when do the facilitating conditions of securitization effectively facilitate securitizing moves? Or, more specifically, is it consistently the case that scientific language enhances securitization attempts, as Berling (2011), for example, argues using only the case of climate change? If not, when does it matter, and how so? Alternatively, is it consistently the case that collective emotions determine the success of securitizing moves, as Van Rythoven (2015) claims based on studying the case of US foreign policy? This role of experiments to clarify, strengthen, and generalize causal pathways is not necessarily incompatible with securitization theory. While a series of securitization scholars are reluctant to endorse the idea of a causal securitization pathway, others are more affirmative. Chiefly, Balzacq unambiguously introduced causation in securitization processes (2005, 188), eventually depicting securitization processes as structured by a "network of causality" (2011a, 18). More recently, he stressed that "the task of a sociological theory of securitization is not just to grasp what it means to say that a phenomenon is a threat, however. It wants to decipher the sequences of cause-and-effect in securitization," an enterprise that should stand "at the center of its explanatory architecture" (2015, 110).9 Guzzini lamented that securitization theory has "left 'explanation' and/or all versions of causality to the positivist other" (2011, 329). Second, due to the shift from cases to causal processes, experiments are uniquely positioned to help securitization theory mitigate against biases, both in case selection (when cases are chosen given particular values of the dependent variable; see, e.g., Geddes 1990) and case analysis (when the researchers' normative views permeate the analysis). These concerns can be mutually reinforcing when a researcher studies unique courses of events such as those commonplace in securitization theory. Selection biases can be particularly salient in securitization theory, where many scholars focus on rather clear cases of securitization and/or cases where desecuritization is normatively desirable10 (typically immigration). Studying cases where securitization has clearly occurred or where such an outcome is morally undesirable tends to beget biases in case analysis. Observing that "the methodological choices and tools that bring about the conclusions seldom appear in empirical securitization papers," Baele and Sterck (2015, 1122), for example, explain that "securitization scholars are often tempted (not) to McDermott, who reminds that "archival work, case studies, field work, surveys, quantitative analysis, and formal modeling ... need not compete with experimentation" (2002a, 31). 9Appraising the way forward for securitization theory, Wasver more recently explained that his "version of securitization theory claims it can integrate causal explanations, social mechanisms, hypothesis testing and political theory" (2015, 123) and called elsewhere to specify more clearly "the causal mechanisms in different parts of the analysis," castigating the "proponents of the theory [who] were wrong to resist causality tout court" (2011, 475-476). 10On the normative dimension of securitization and desecuritization, see, for example, Aradau (2004), Floyd (2011), Roe (2012). 650 Stephane J. Baele and Catarina P. Thomson detect securitization where they want, and not where it actually is (not) - an inclination which might be curved down by a real commitment to transparent and robust methods." Such biases in case analysis reinforce static accounts of securitization that excessively focuses on the move itself and neglects its outcome (Wsever 2011, 467) or more crucially ignores what links the move with this outcome, leading scholars to infer that securitizing moves are almost automatically successful (a point well argued in Van Rythoven 2015). Selecting cases on the basis of securitization having occurred is often unavoidable and does not constitute a problem in itself. However, it is important that "the analyst should understand what can and cannot be accomplished with cases selected for this reason" (Geddes 2007, 90). This should include identifying boundary conditions and clarifying what lessons inferred from examining the selected cases can tell us about a wider universe of potential cases. Experiments can help in this regard, by complementing the study of cases that risk being studied in a biased way and offering a direct comparison of the effects of causal processes among more varied scenarios or members of the public. Focusing on causal processes could contribute to achieving greater methodological clarity and transparency, thereby potentially helping move securitization theory away from case selection/analysis biases. This is not to say that experiments are immune to biases. Clearly researchers chose what experiments to conduct out of the many possible interactions between theoretically relevant variables that exist. Choice of specific experimental designs can also affect measured outcomes; for instance there is evidence to suggest that participants don't make the same choices in multiple or single-choice procedures (Morton and Williams 2010). However, as part of a multimethod approach, experimentation can contribute to achieving a more systematic analysis of the conditions under which particular types of securitizing moves succeed, fail, produce unexpected outcomes, or lead to mixed results. Evidently, securitization theory emphasizes many aspects that can simply not be explored or explained by experiments; experimental methods are neither tailored to investigate the complexities of specific cases of securitization nor pertinent for the study of the long-term political processes, group practices, or administrative dynamics that support or diffuse some securitizing moves. We contend that the role of experiments would be limited to that of a second-step research agenda seeking to strengthen particular causal pathways.11 Experiments are principally useful tools that artificially create securitizing moves understood as relatively short-term limited interactions.12 In the following section we delineate their potential benefits for the theory around the sole question of securitization failure/success. We do not therefore argue in favor of a generalization of a cognitive approach to threat perception, which would lead researchers to ignore long-term processes and contribute to securitization theory becoming a mere "subfield of framing" (Watson 2012, 280). Experimental Contributions to Securitization Theory: When and How Does Securitization Succeed or Fail? We argue that experiments should be seen as an important tool in the effort to answer what is arguably one of the most central yet also one of the less clearly answered questions of securitization research: what makes securitizing moves nIn fact, securitization theory's notable strength in these areas is something that could be used to enrich experimental analyses in terms of contextualizing their empirical findings, which are often wrongly generalized to different contexts. In this sense, an experimental agenda for securitization should only be conducted in constant dialogue with other methods, in an effort to build strong results backed by several methodological endeavors, in the spirit of Balzacq's call for triangulation in critical security studies (2014). 12Although some experiments have longer term follow-ups after the initial experiment has been conducted, such designs are rare due to participant mortality and increased costs (McDermott 2011b). An Experimental Agenda for Securitization Theory 651 succeed or fail?13 This might seem like a limited role, yet the question is crucial. If securitization theory truly aims at "specifying how, and under which conditions, the security-ness of an issue is fixed" (Balzacq, Leonard, and Ruzicka 2016, 24), then it is surprising to observe that the question of the determinants of success or failure of securitizing moves is one that most securitization scholars agree is underdeveloped (e.g., Balzacq, Leonard, and Ruzicka 2016, 6). Securitization theory originally did mention conditions that determine the success or failure of speech-based securitizing moves, pointing to "internal, linguistic-grammatical" factors, as well as and "external, contextual and social" ones (Buzan, Wsever, and De Wilde 1998, 32). However, these conditions were never fully developed and remained undertheorized. Almost fifteen years later, Watson (2012, 298), for example, still observed that "audience acceptance is a core condition of successful securitization, but few studies of securitization have attempted to demonstrate the effects that securitizing discourse has on the targeted audience." Guzzini (2011, 331) has gone as far as saying that a certain formulation of securitization theory literally "begs the question of under which (political, cultural, economic, etc.) conditions, with which audience, while using which discourse, etc., (de)securitization attempts (can) succeed."14 Yet as Williams argued, in principle the securitizing "claims that are likely to be effective, the forms in which they can be made, the objects to which they refer, and the social positions from which they can effectively be spoken are usually deeply 'sedimented' (rhetorically and discursively, culturally, and institutionally) and structured in ways that make securitizations somewhat predictable and thus subject to probabilistic analysis (2003, 514). Experiments can move securitization theory toward a systematic analysis of the conditions under which particular types of securitizing moves succeed or fail in shaping particular audiences' acceptance of an issue as being a security one, thereby directly addressing Balzacq's recognition that tracing the causal link between a securitizing move and its effects is "definitely one area where much research is still required" (2011b, 30). We disaggregate the question into three subcomponents, each of which corresponds to one of Balzacq's (2011b, 35-36) levels of securitization analysis: acts (the securitizing move itself), agents (both the securitizing actor and the audience and more crucially the interaction between them), and context (within which the entire process takes place). These subcomponents are therefore: (1) Which types of securitizing moves are more likely to succeed or fail? (2) What types of interactions between securitizing actors and audience increase or decrease the likelihood of successful securitization? and (3) Which types of contexts make securitization more likely to succeed or fail? 15 We must note that while these subcomponents help expose specific experimental contributions more clearly, to some extent they are artificial, and there will inevitably be some degree of overlap between them. Experiments provide us with the opportunity to both zoom into the effects of particular variables in each of these areas, as well as to examine interactions that take place between these three different levels. 13As stated above, experiments can only shed light on the success or failure of language-based securitizing moves, not on the effects of more long-term securitizing practices. 14We agree with Van Ryfhoven (2015) when he argues that this question has not yet been successfully investigated because securitization research is biased toward the idea that securitizing moves are almost automatically successful. 15This choice also corresponds to Williams' identification of factors theoretically limiting securitization dynamics, which are "in practice structured by the differential capacity of actors to make socially effective claims about threats, by the forms in which these claims can be made in order to be recognized and accepted as convincing by the relevant audience, and by the empirical factors or situations to which these actors can make reference. Not all claims are socially effective, and not all actors are in equally powerful positions to make them" (2003, 514). 652 Stephane J. Baele and Catarina P. Thomson Which Types of Securitizing Moves Are More Likely to Succeed or Fail? Balzacq has been a vocal advocate of the idea that securitization includes much more than language strictly understood as speech act, comprising instead "various artifacts (metaphors, emotions, stereotypes, gestures, silence, and even lies) to reach its goals" (2005, 172; 2011a). While language still stands at the center of securitization theory, Balzacq's view is now widely shared, prompting scholars to look at nonlinguistic components of the securitizing move—either at nonlinguis-tic aspects of a speech act (for example, a display of emotions) or at non-language-based securitizing moves (e.g., Campbell 2008 or Hansen 2011 on images). Watson, for instance, explains how the framing literature, which analyses the effects of "a variety of framing devices other than just speech acts," could reinforce securitization research (2012, 297). However, claims on how exactly these various aspects impact on securitization dynamics are still rather speculative. Experiments would be pertinent tools to test general claims on how linguistic and nonlinguistic dimensions of securitizing moves work together, thereby offering key distinctions at two levels. First, experimental designs are useful tools to compare the effects of different types of securitization support. By "type of support," we refer to the form taken by the securitizing move—does it proceed through spoken language (speech/discourse), written language (text), or images (both still and video)? Between-group experimental designs are an optimal choice for such direct comparisons. In these studies participants make choices in only one state of the world and responses made by those in different groups are compared.16 For example, Nyhan et al. (2014) showed that textual information about the dangers of the diseases prevented by the MMR vaccine shaped public attitudes in a different way than pictures or graphs conveying the same message. This finding, among others, shows that there is no reason to think that securitizing moves relying, say, on written text have the same effects on public attitudes than those relying on images. The main advantage of this kind of design is that it avoids oversensitizing respondents to the experimental factors, as exposure to different scenarios risks participants identifying the experimental variables and can trigger a comparison between them that might artificially increase differences in the dependent variable measures. The risk commonly associated with between-groups designs is that subject-specific observable and unobservable characteristics might affect participants' responses (Morton and Williams 2010), but random assignment of participants to scenarios can assure that the internal validity of the study is not threatened by using a between-groups design. Although both laboratory and online survey experiments can be used for this purpose, a laboratory setup is ideal as it safeguards the internal validity of studies making these comparisons. 17 Although it is usually argued that external validity is the greatest pitfall of experimental designs, many experimentalists in international relations are more concerned with potential internal validity flaws (McDermott 2002a, 2002b). After all, internal validity can only be the greatest strength of experiments when they are correctly executed (that is, when the researcher effectively controls the manipulation of the independent variables).18 Privileging "mundane realism" over internal validity (which can be particularly risky when conducting field experiments) can 16As opposed to a within-groups experiment that could, for instance, expose each individual participant first to an audio manipulation, then a video one, and then a text one. 1 Being with the participants in a lab means the researcher can ensure they are exposed to the entirety of the independent variable manipulation, which is particularly important when a video or audio clip is introduced (McDermott 2002a; Clifford andjerit 2014; Thomson, Geva, and Tkach 2017). In contrast, participants of an online survey experiment might not see the entire clip or might have audio issues. 18A straightforward way to check if participants understood the introduction of the independent variables as intended by the researcher is to include manipulation check questions after measuring the relevant dependent variables. An Experimental Agenda for Securitization Theory 653 be a mistake, for as noted by McDermott, "without internal validity there can be no external validity" (2002b, 334). Once the internal validity of an experiment is safeguarded, however, there are certain steps that can be taken to help external validity. Researchers should strive to work with samples that are as representative as possible of the broader general population of interest and conduct demographic checks to assess whether or not key demographic variables are distributed similarly across the sample of experimental participants and the broader population encompassed by the theory. Since one of the nuclear issues of theoretical concern for securitization theory is examining the effects of elite discourse on the mass public,19 internally valid laboratory experiments, where serious consideration has been given external validity, can help gauge how and when particular types of support are most effective in securitizing issues. More precisely, the standard experimental design would rest on the principle that the same securitizing message can be delivered in several forms, through different types of support, sometimes combined, and would assess the relative effects of each form or possible combinations thereof. This would build on a growing field of research on the different effects produced by different types of political communication. For example, using a vignette-based2 lab experiment, Rashotte (2003) has shown sharp differences in participants' affective reactions when facing events reported in either written or videotape form. In an experiment designed to identify under what circumstances domestic audiences support the use of coercive interrogation during times of armed conflict, Thomson, Geva, and Tkach (2017) exposed participants to almost identical information transmitted in either text or audio form. Tellingly, the effects of other variables of interest such as whether or not the target of coercive interrogation was an in-group member, or whether or not it secured actionable intelligence, were significantly affected by how participants learned about the specific coercive interrogation scenario. These examples illustrate that more research is needed, not only on which types of support impact the outcome of securitizing moves, but also on how, to whom, and in which circumstances they do so. Laboratory settings also provide the opportunity to answer research questions employing a wide array of dependent variable measures. This is particularly important when investigating what type of securitizing moves are successful as individuals might not consciously recognize that an issue has been securitized or openly admit that they accept a securitizing move, even though subtle corporal reactions would suggest otherwise. For instance, an individual might not want to admit feeling threatened by immigrants, but raised Cortisol levels when exposed to images of immigrants might suggest otherwise. A first level of measurement is largely non-intrusive and can also be readily employed in survey experimental designs (participants might, for instance, be asked about their perceptions or support for a given policy or to report levels of fear or other emotions). A second level of measurement can be accessed in the lab and is typical of experiments used in the psychological tradition, as the effects produced by the experimental manipulations can be observed via physiological processes. Evidence that a treatment does indeed produce a specific effect can include, for example, high stress revealed by blood pressure and Cortisol levels, the experience of fear registered using functional magnetic resonance imaging scan, or the aggression as attested by testosterone levels. 19There are exceptions, for example, Roe (2008), who considers various strands of the audience and thereby includes elite-to-elite securitizing moves. 20Vignette experiments "use short descriptions of situations or persons (vignettes) that are usually shown to respondents in order to elicit their judgments about these scenarios" (Atzmiiller and Steiner 2015)—see also Collett and Childs (2011) or Hainmueller, Hangartner, and Yamamoto (2015) on the internal and external validity of vignette experiments. 21For a comprehensive review of politics experiments using this second level of measurement, see McDermott (2011a, 515-17). 654 Stephane J. Baele and Catarina P. Thomson Second, experiments could also be fruitfully used to assess the impact of variations in communication style within a single type (or combination) of securitization support. For example, does exposure to different types of pictures (for instance, shocking images versus symbolic ones) provoke different emotional reactions that in turn lead to differing levels of support for an underlying securi-tizing claim? Or do different styles of securitizing speeches (for instance, highly emotional versus highly rational) elicit different levels of support? Employing a factorial experimental design, researchers can compare responses between participants exposed to higher or lower values of the specific theoretical factor of relevance.22 For example, Brader (2005) showed that including uplifting music or positive images (e.g., children playing) in political video ads significantly impacts viewers' attitudes. Exploring the role of still images, Caverley and Krupnikov (2015) have, for example, recently identified the specific elements that, if present in a picture together with the US president, significantly enhance viewers' support for hardline military policies. They showed participants pictures of the US president with different backgrounds—soldiers, students, children, or "ordinary" people—and found out that the image of soldiers significantly shifts participants' preferences toward strengthening defense instead of education. These findings are directly relevant to securitization theory, especially its recent theorization regarding the role of images and "visual icons" in securitization (e.g., Hansen 2011, 2014). Exploring the effects of variations in communication style within a single type of securitization support is also useful to assess the impact of factors that have not been fully integrated into securitization theory but that have been shown in adjacent fields to play an important role in threat communication and perception. A prominent example of such a factor are emotions, which have been highlighted in neuroscience as significantly affecting policy preferences and how decisions are made but remain undertheorized in securitization theory (see Van Rythoven 2015 for a comprehensive analysis). Existing work on the emotional impact of pictures and videos, especially anger-eliciting images (e.g., Cheung-Blunden and Blunden 2008) or shocking ones (e.g., Jones and Henriksen 1987) could serve as a basis for further experimentation on securitization. For instance, an experiment could be conducted in which the same securitizing message delivered by the same securitizing actor in a short video clip (identical medium of securitization exposure) is conveyed with either graphic content (in one experimental condition) or only nonshocking content (in the other condition). In order to safeguard the internal validity of such manipulations, it is vital to measure participants' evaluation of the overall emotional salience of these different conditions, though, through manipulation, check items to be included after the dependent variable measures in the postexperimental questionnaire.23 Importantly, emotions in securitization processes can be examined as independent variables (as in the example just described), but also as dependent variables or even as variables that mediate the relationship between two other variables of theoretical interest. As for the former, an experiment could, for example, study the effects of certain securitizing moves on the emotional polarization of key constituent groups. Regarding the latter, the coercive interrogation experiment mentioned above found that emotional contexts affect the impact of other variables on participants' willingness to support of the use of coercive interrogation. Although some experiments compare one (or more) treatment condition (s) to a control condition (where no value of the independent variable is introduced), these designs must be used ensuring that all experimental scenarios are comparable in terms of length, for instance. 23Although manipulation checks should ideally be used for every independent variable being manipulated in an experiment, this becomes especially important when varying emotional exposure. Manipulation checks should be introduced at the end of the postexperimental questionnaire to avoid oversensitizing participants to the key variables of interest. An Experimental Agenda for Securitization Theory 655 Another area in which experiments might prove a useful tool through which to study variations in communication style within a single type of securitizing communication support would be to assess the influence of specific securitizing words. Since the original securitization framework (Buzan, Wsever, and De Wilde 1998), it is assumed that security words are performative and can trigger perlocutionary effects broadly speaking—that is, they "produce certain consequential effects upon the feelings, thoughts or actions of the audience" (Austin 1962, 101). But is this actually the case? Which words have a particular power to "do something" as Austin would say—that is, to exert an effect on the chances of success of a securitizing move? How strong exactly is this power? Is this power enhanced if these words are repeated? To the best of our knowledge, these crucial questions have so far received no clear and direct answer in the securitization research program, and experiments excel in isolating these factors. Watson (2012) is right in directing securitization scholars willing to investigate this kind of question toward framing studies, which have a good record in experimentally studying the performative dimension of language, with findings sometimes directly related to securitization dynamics. For example, Baele, Sterck, and Coan (2016) examine the impact of using numbers in security communication by conducting both a field and an online survey experiment in which participants were exposed to a text securitizing tuberculosis (naming it a security threat and calling for exceptional measures like tight police control, the criminalization of immigrants and low-revenue "target populations," and the possibility of taking urgent executive action on the issue). One scenario used a text full of quantitative claims whereas the other one used the exact same text but replaced each quantitative statement by its best qualitative equivalent.24 Very specific differences were found between the treatments, demonstrating the need for direct experimental evidence on the impact of this particular style of communication. Another example is that of Baele et al. (2015). Use of a field vignette-based experiment allows them to evaluate the respective performative power of the "terrorist" and "Islamist" labels, which were found to elicit greater levels of fear and greater acceptance of exceptional and intrusive mass-surveillance policies than more "neutral" labels like "gunman." Studies like these, although not explicitly grounded in securitization theory, are nonetheless highly pertinent for the study of securitizing moves and would merit being integrated into a forward-looking experimental agenda for securitization theory. What Types of Interactions Between Securitizing Actors and Audience Increase or Decrease the Likelihood of Securitization ? Experiments can help securitization theory identify, map, and assess the strength of the numerous variables associated with the actors involved in the securitizing move, usually conceptualized as the securitizing actor on the one hand and the targeted actor or audience on the other hand. More crucially, experiments can help study the interaction between these two sets of actors, backing the kind of relational approach that has recently been favored by securitization theory. As evidenced by research on framing, individuals vary in their responsiveness to attempts to shape their perceptions and attitudes; motivation, level of education, time constrains, and previously held beliefs all matter (Chong and Druckman 2007; Sniderman and Theriault 2004). First, it is well known that the characteristics of the securitizing actor matter in the success or failure of securitizing moves (e.g., Williams 2003). Bourdieu's For example, "every year 25,000 individuals die in the United States from multiresistant bacterial infections" in scenario one, and "every year many individuals die in the United States from multiresistant bacterial infections" in scenario two. 656 Stephane J. Baele and Catarina P. Thomson legacy on "authorized language" and symbolic power (e.g., 1991) has been influential in shaping securitization theory's recognition that not everyone has an equal probability of success in securitizing an issue. However, which of the securi-tizing actor's characteristics exactly matter and when is still unclear. Do women, for example, have a smaller chance of being successful securitizing actors than men? Do older individuals tend to be more successful than young individuals? Does occupying a prestigious social position or possessing a higher level of education really matter? For example, in the abovementioned experiment of Baele, Sterck, and Coan (2016), both the quantitative and qualitative scenarios were delivered by either an individual evidently imbued with attributes enhancing high scientific legitimacy25 or an individual obviously devoid of these attributes. 6 The experiment disclosed unexpected results, debunking the intuitive idea of an unambiguously positive effect of legitimacy or "authorized language" in a security communication akin to a securitizing move. Similarly, the design followed by Coan et al. (2016) contrasts the type and amount of information individuals seeks after being confronted with terrorist threat messages (group 1) or government's reports of increased risks of terrorist attacks (group 2), highlighting different effects depending on the status of the source. An advantage of the experimental method is that not only does it allow researchers to assess the effects of variables that might not naturally occur often, but it also allows them to evaluate if a combination of these unlikely factors might have a confounding effect.27 This can become relevant when studying the effects of securitizing actors' gender, ethnicity, or age on securitization for instance, as we know that women, members of ethnic minorities, and younger individuals are underrepresented in the security community. Ali, for instance, clearly shows that the effects of securitizing news on an audience depend on the ethnicity of the presenter (2014). Similarly, are younger female securitization actors less likely to be successful in securitizing an issue than older female actors? Employing experiments also offers the opportunity to assess whether or not varying national threat levels might increase or decrease these effects. Going back to our example: perhaps in times of eminent threat, women, members of ethnic minorities, and younger individuals are perceived as being even less credible interlocutors (and perhaps younger women are the less credible group)? Second, experiments can also be useful in evaluating the role of the audience. This remains a relatively unchartered domain in securitization theory. As noted by Balzacq, "scholars are wont to claim that audience is a crucial element in sanctioning the intersubjective nature of securitization ... However, the literature seldom provides clear examples of cases that meet the overriding assumption that an (observable) audience has agreed with the securitizing claims" (2015, 107). Building on Balzacq's (2005) and Vuori's (2008) calls to better problematize the audience in securitization theory, a series of scholars have tried to offer a more sophisticated, less one-sided conceptualization of the audience. Leonard and Kaunert (2011) or Roe (2008) have highlighted the multiplicity of audiences and hence the complexity of their reception of securitizing moves, which is neither direct nor self-evident across various audiences. Yet as Herborth and Kessler (2010) rightly observe, in spite of these efforts the account of the audience in securitization theory remains quite simplistic, resting on homogenous conceptions of large groups who remain passive. We do not know, for example, whether 25a wise-looking older white man with a white lab coat, presented as the "centennial professor of infectious diseases at mit" and director of an "independent congressional committee" on public health. 26a young man with untidy hairs and a casual jumper, only labelled as a "blogger." 27Since all these variables and others are linked to the overarching question of legitimacy, the following examples illustrate the impact of legitimacy taken in its most obvious sense, but other experiments can manipulate variables not directly or not as clearly linked to legitimacy—for instance how much the intentionality or undecidedness of the actor affects the success of the move. An Experimental Agenda for Securitization Theory 657 or not men tend to be convinced more easily by securitizing narratives than women, whether or not education level matters in this process, whether or not the audience's expertise, political affiliation, social class, or level of religiosity play a role, and, if so, exactly what this role would be. The effect of such variables is routinely controlled for in most rigorous studies in economics, psychology, medicine, and elsewhere, which means that they cannot be dismissed in securitization research, where it is reasonable to expect that these variables would moderate or shape resistance to securitizing moves. In this context, experimental designs can help us learn more about whether or not specific demographic or other individual factors mediate the effect of experimental manipulations (such as certain securitizing moves or actors) on the likelihood of securitization occurring. That each participant may end up in any of the experimental groups through random assignment, being exposed to any given constellation of independent variable manipulation means that they will not self-select into certain groupings as occurs in the natural world. This can play a very important role in debunking a homogeneous and passive conceptualization of audience as it allows distinguishing between, on the one hand, the effects caused by theoretical variables of interest in a specific study and, on the other hand, demographic variables that might affect selection into groups more generally. Building on prior knowledge from adjacent fields, securitization researchers can suspect that certain variables might play a mediating effect. We know, for instance, that individuals on either side of the ideological spectrum, or experiencing different emotions or mood,28 are inclined to react differently to political stimuli, and postexperimental analyses can focus on political affiliation as well as other theoretically plausible moderating variables such as gender, education, or ethnicity.29 For example, Baele et al. (2015)'s experiment described above revealed significant differences in how women and Muslims react to the "terrorist" and "Islamist" labels, thereby contesting the idea that these labels have unambiguous effects on a homogeneous audience. Exciting new methodological developments are now available to help uncover important mediating factors even when existing theory provides little guidance. These include fitting finite mixture models (Schlattman 2009; Lanza and Rhoades 2013) that accommodate observed and unobserved heterogeneity in responsiveness to treatments. This can be particularly useful when it is difficult to identify a priori what individual-level variables might play key mediating roles. For example, given the relative novelty of the recent Brexit referendum and that supporters of both the leave and remaining camps did not follow clear party lines, Thomson et al. (2017) fitted a finite mixture model to examine which segments of the UK electorate would have supported the prime minister had he backed down from the campaign promise to hold the referendum. Given that new areas can be prone to securitization (even if they had not in the past), it is not difficult to imagine how employing these models could help advance the field by allowing researchers to closely examine individual-level variables that affect the probability of specific securitizing moves' success. Third, experiments present a unique opportunity to identify interaction patterns between a specific securitizing actor and a particular audience, which can help us closely examine the interactional dimension that lies at the core of securitization theory. Although previous research might have identified certain variables as increasing or decreasing the likelihood that an actor will succeed in securitizing In framing experiments, even the audience's emotional state have been shown to "either amplify or depress a frame's impact" (Druckman and McDermott 2008, 297). 29This can be done very straightforwardly by interacting these factors with the independent variables (similarly as one does in natural variance quantitative studies). 30The impact of the "terrorist" label was weaker among women, and the impact of the "Islamist" label was stronger on Muslims. 658 Stephane J. Baele and Catarina P. Thomson an issue, this effect might not carry to different audiences (and might actually have the opposite effect). It is indeed known that exposure to certain information or variables can provoke contradictory effects in specific subgroups (Nyhan and Reifler 2010, 2015), and there is no reason to doubt this when it comes to security communication. Baele, Sterck, and Coan's experiment on quantitative versus qualitative threat narratives (2016), for example, confirmed the idea that legitimacy is not an objective feature attached to the speaker; rather, it results from an intersubjective, context- and group- specific norm on what is recognized and accepted as legitimacy attributes. Indeed a securitizing actor might enjoy high legitimacy within a specific group but a very low one within another group, as Kahan and colleagues' extensive series of pioneering experiments on risk communication (e.g., Kahan et al. 2009 on nanotechnologies; Kahan et al. 2012 on climate change) and security communication (e.g., Kahan and Braman 2003 on gun control) have shown. Ali's experiments on securitizing news (2014) provide emerging evidence on this interdependence of the characteristics of the audience on the one hand and those of the securitizing actor on the other hand, showing that people's acceptance or refusal of a securitizing framing in the news depends on the trust and credibility they assign to the presenter, which heavily depends on his/her characteristics (e.g., member of an ethnic minority). In terms of studying interaction patterns between a specific securitizing actor and a particular audience, securitization theory could particularly benefit from experimental efforts already done by audience-cost/democratic accountability scholars, who claim that executives who act in ways that do not represent popular preferences will see their approval decrease and therefore strive to act accordingly. As Schultz notes regarding the specific, securitization-related case of military crises, "to the extent that leaders value holding office, they are unlikely to make choices that lead to outcomes with high domestic political costs" (2001, 33). Methodologically, this means that assessing public opinion only after the executive has advanced national policy in, say, terms of economic and military coercion might omit precisely the cases in which the public should theoretically be more likely to hold their leaders accountable. Thus, experiments have become quite popular among scholars of democratic accountability (e.g., Tomz 2007; Weeks 2008; Trager and Vavreck 2011), in ways that could inspire securitization scholars who wish to study the potential costs/benefits of failed/successful or abandoned securitizing moves. Similarly, survey experiments can be conducted simultaneously on two different samples: security elites and the general public. For example, Thomson's work (2017b) is currently examining whether or not security attitudes match across these samples, and most importantly, whether or not they can correctly predict what the other sample's preferences are in the context of the British Strategic Defence and Security Review. Admittedly, access to a sufficiently large elite sample in order to allow conducting an experiment is difficult to come by, so experimental contributions to securitization theory in this aspect are limited.31 Diverse experimental technologies mean that there are countless options for innovations in research design that can assist in understanding the intersubjective dynamics that link both types of actors involved in the securitizing move. The DecTracer, for instance, is an innovative experimental platform that can be especially valuable for securitization scholars interested in precisely tracing the intersubjective intricacies of the securitizing move.32 It is a multilingual web-based process tracing system that allows experimenters not only to see what decision a participant made, but can also inform how the actual decision-making process occurred. Experimental setups can include, for example, access to additional Notable exceptions include Chin, Bond, and Geva 2000 or Nyhan and Reifler 2015b. 'Proprietary experimental platform developed for Nehemia Geva by Infinity. An Experimental Agenda for Securitization Theory 659 information items individual participants can select to aid their decision-making process. Participants might be able to see news items from different sources on their screens and click on one or more to read/listen/view (sources to be introduced might include news items from different broadcasting agencies, governmental intelligence reports, or academic papers). It might very well be the case that individuals with a given set of demographic characteristics seek more/less/ different information relative to baseline population rates or that seeking additional input is contingent on personal or institutional traits of the securitizing actor. Group decision-making experiments can also be useful. Although they can be costly to conduct (and the researcher runs the risk of losing some degree of control over how the independent variable manipulations are introduced), they offer experimenters the chance to observe individual and group processes simultaneously (Morton and Williams 2010). In the context of securitization theory, use of group experiments could help pinpoint how group dynamics affect relationships between securitizing actors and the targeted audience. It may very well be that, when it comes to securitization dynamics, the influence of group pressures plays a greater role than the actor or that the presence or absence of shared traits between the securitizing actor and the group significantly affects the chance of success of the securitizing move (regardless of whether or not an individual respondent shares this given trait). Designs like these could potentially help explain the dynamics involved in recent prominent securitizing speeches, like those on foreigners and immigrants voiced by pro-Brexit and pro-Trump campaigners. Which Types of Contexts Make Securitization More Likely to Succeed or Fail? Particular contexts affect particular securitization outcomes. Put differently, a securitizing move may prove to be successful in one context but its iteration in another one may fail. Balzacq (2005) distinguished two levels of context: (1) broad macrolevel sociocultural environment within which the securitizing move happens (for instance the "war on terror" period in Western states) and (2) the more immediate microlevel setting in which the securitizing move takes place (a presidential address, for instance) .3 Other scholars understand context in a more Foucauldian way as a "local regime of truth" (Salter 2008) or dominant system of beliefs and discourse that constrains which securitizing moves could potentially be meaningful and accepted by the population. Securitization scholars have already documented the link between securitization context and outcomes, using case-studies or limited comparative analyses. For example, Mnatsakanyan (2015) has explained the failure of the British government to get a vote authorizing an intervention in Syria after the 2013 gassing of children by highlighting the presence of a broad, post-2003 Iraq war context of defiance against any government justifying foreign intervention on the basis of weapons of mass destruction. For the study of how particular contexts shape or constrain the outcome of securitizing moves, a methodological focus on single case studies makes a lot of sense, and here experimental designs are certainly less useful. However, securitization theory's account of the context still needs further refinements, and experiments can help answer "whether differences in the outcomes of securitizing moves can be derived from and explained by differences in contextual features" (Balzacq, Leonard, and Ruzicka 2016, 9). To ask this question is to adhere to an "externalist" view of the context (Balzacq 2005), where the context preexists and is not solely directly coconstituted by the securitizing move—in other words, is This distinction is reiterated in Balzacq 2011b, where he distinguishes between "proximal" and "distal" contexts. 660 Stéphane J. Baele and Catarina P. Thomson external to it and therefore "has an independent status, which allows it to influence security articulations in a distinctive way" (Balzacq, Leonard, and Ruzicka 2016, 10). Not all securitization scholars may endorse this conceptualization of the context, however. Paradoxically, experiments are often presented as the perfect method to generate findings that are not influenced by context. In other words, most experiments have sought to gain internal validity by making sure that there is no contaminating impact on the experimental design and therefore results of external elements. Yet we argue that experiments can help in further conceptualizing and assessing the role of the context in securitization, in three (albeit limited) ways. First, experiments can in principle carefully integrate proximal situational variables within their design, in a way that highlights the effects of microlevel contextual settings. One clear example of an experiment varying the microlevel context is Darley and Batson's famous "good Samaritan" study (1973). In this experiment, all participants were told to go to another building in order to prepare a task, but some of them were told that they had little time. This design revealed the impact of a context of time pressure on helping behavior; those who were told they had little time tended not to help a suffering person lying down in between the two buildings. One can easily imagine a similar experimental design aimed at evaluating the effect of specific contextual settings within which the securitizing move occurs. For example, a laboratory experiment testing participants' reactions to a securitizing move could be conducted in which half of the participants wait "before" the experiment in a room where a TV displays security-related news and the other half wait in the same room but with the TV showing more trivial programming or switched off. Such a design would provide information on the hypothesis that people with more recent exposure to security-related news are more likely to back a securitizing move when facing a specific issue. Second, the role of a broader macrolevel context can also (and perhaps chiefly) be assessed, this time through replication—that is, running the same experiment in a series of different contexts broadly conceived, which allows to directly document the importance of and theorize the workings of the broad context of securitization. Contextual variation can be achieved through replication in two ways, either diachronically—running the exact same securitization experiment regularly across a long period of time to gain a longitudinal understanding of securitization and context—or geographically—running the exact same securitization experiment simultaneously in a series of different places to gain an understanding of how factors like culture, popular references, political history, or recent local events impact on securitizing moves effects. Such replications could therefore evaluate the "decisive effect [of the context] not only on the reception of the securitizing moves, but also on the perception of those who utter them" (Balzacq, Leonard, and Ruzicka 2016, 11). Another example of such an approach would seek to test whether or not security speech acts have the same effects across cultures broadly conceived, or if all things being equal, the success/failure of securitizing moves really depends on the contextual features, and if yes, which ones.34 Such an approach could enrich our understanding on the impact of political events on securitization dynamics—an experiment mimicking an attempt to securitize immigrants could, for instance, have had very different results if conducted both before and after the recent US presidential campaign. The third way in which experimental designs can potentially help our understanding of the effects contextual factors can have on securitization processes highlights the role of incentive structures. Experiments that study the role of incentive structures set up stylized situations (typically called "games") in which Here the line is thin between an assessment of the context and that of the audience, as it can be argued that the macro-level context eventually influences audience's perceptions. An Experimental Agenda for Securitization Theory 661 participants' reactions to expected outcomes ("if you do X, you get Y") are scrutinized. These types of experiments are typically used in behavioral economics, where participants' deviations from the rational choice are studied. Ironically, they are known as "context free" experiments, due to the highly stylized and abstract scenarios that are typically employed (Dickson 2011). Probably the most renowned example of such an experimental design is Axelrod's study of iterated prisoner's dilemmas (1984), which revolutionized the way anarchy is conceptualized in international relations. More recently, Bragg, Geva, and Thomson (2016) designed a stylized cooperation game in order to better understand why the local population in Afghanistan was not cooperating with security and development projects designed by the coalition forces. Conducting field work in Afghanistan was not an option at the time, so the team opted to develop a rational choice framework to formalize costs and benefits and the respective probabilities that underlined the choices individuals faced in the particular context in Afghanistan. Admittedly, this take on context differs from the traditional ways it is perceived in securitization theory, yet we believe securitization scholars could gain in taking it into account in order to determine which incentive structures play a role in the acceptance of securitizing moves. As originally envisioned by the Copenhagen School, securitizing moves contain a performative element of promise—chiefly the promise that the threat under consideration is going to be addressed strongly and urgently with extraordinary policies. Experimental designs could study how participants, facing a particular threat situation in a stylized game, respond to various types of incentives. For instance, a promise to extend the autonomy of intelligence services' autonomy or putting specific amounts of army personnel in city streets can be presented in different ways: positively as a reinforcement of citizen security or as a precondition for their freedom or negatively as a necessary temporary limitation on citizen rights. This echoes Kahneman and Tversky's famous findings on the impact of presenting logically identical information either as loss risks or gain probabilities (1984; Tverskey and Kahneman 1981). Designs like these would therefore closely intertwine the context—the very structure of the game situation and incentives—with the form of the securitizing move—how the incentives are presented—with which words, according to which logic. Conclusions Our objective was to propose ways in which experiments can be applied in securitization research. We believe this would have two important benefits. First, it would help mitigate two longstanding methodological shortcomings of securitization research: difficulties in producing general conclusions that transcend particularistic results and biases in case selection and analysis. Second, it would contribute to establishing a more precise understanding of the determinants of the success or failure of securitizing moves, by examining how the characteristics of the securitizing move itself, its actors, and context impact audiences' attitudes. The article contributes to the broader recent debate on how quantitative methods can contribute to critical security studies in general. It should, also, be understood as an effort to decompartmentalize securitization theory by bringing it close to neighboring theories similarly focusing on the use and effects of security language and behavior—framing theory, social psychology of intergroup relations, and cognitive approaches to political perception. We leave open to further research the task to precisely map these potential links. The overall ambition of the present article is to help unlock securitization theory's full potential. Finally, we would insist that the political and ethical implications of utilizing experiments to securitization theory should not be underestimated. At the indirect level of what is commonly called the "normative dimension" or "politics" of securitization theory, the risk is to see experiments providing almost turnkey 662 Stephane J. Baele and Catarina P. Thomson solutions to those who seek to securitize an issue. Wsever rightly reminds us that securitization theory highlights "the inherently political nature of any designation of security issues and thus it puts an ethical question at the feet of analysts, decision-makers and activists alike" (2011, 334). Building a strong and cumulative empirical research agenda whose ambition is to identify, measure, and map as precisely as possible the various dynamics of securitization might simply not be politically (or ethically) prudent. At the second level of the direct ethical issues related to experiments, experimenting on security in general and securitization more specifically significantly enhances the classic ethical problems associated with experiments, chiefly the problem of the participants' reactions following the experiment itself.35 For example, if not carefully debriefed, participants in a securitization experiment could be shocked or even can be led to believe and endorse the securitizing narratives put forward in the scenario they were exposed to, since these should precisely have been tailored by the experimenter to trigger a change in appraisals, beliefs, perceptions, or policy preferences among participants. Generating emotional distress is a common risk that researchers should be wary of when debriefing participants postexperiment. As securitization theory's main axiom indeed precisely teaches us, people's reactions and fears should not be taken lightly. References Ali, Sadaf R, "Trust, Credibility and Authenticity: Race and its Effect on Audience Perceptions of News Information from Traditional and Alternative Sources" (PhD diss., Wayne State University, 2014). Andreoni, James, Marco Castillo, and Ragan Petrie. 2003. "What Do Bargainers' Preferences Look Like? Experiments with a Convex Ultimatum Game." American Economic Review 93 (3): 672-85. Angrist, Joshua, and Jorn-Steffen Pischke. 2008. Mostly Harmless Econometrics. An Empiricist's Companion. Princeton, NJ: Princeton University Press. Aradau, Claudia. 2004. "Security and the Democratic Scene." Journal of International Relations and Development 7 (4): 388-413. Aradau, Claudia, Jef Huysmans, Andrew Neal, and Nadine Voelkner, eds. 2015. Critical Security Methods: New Frameworks for Analysis. New York: Routledge. Atzmüller, Christiane, and Peter Steiner. 2015. "Experimental Vignette Studies in Survey Research." Methodology 6 (3): 128-38. Austin, John. 1962. How to Do Things with Words. Oxford, UK: Clarendon Press. Axelrod, Robert. 1984. The Evolution of Cooperation. New York: Basic Books. Baele, Stephane J. 2017. "Stagnation ou Perte d'ldentite? Resoudre le Dilemme du Programme de Recherche en Securitisation." Etudes Internationales. Baele, Stephane J., and Olivier C. Sterck. 2015. "Diagnosing the Securitisation of Immigration at the EU Level: A New Method for Stronger Empirical Claims." Political Studies 63 (5): 1120-39. Baele, Stephane, Olivier C. Sterck, and Travis Coan. "Security through Numbers? Assessing Quantitative Arguments in Securitization." Working paper, University of Exeter, UK, 2016. Baele, Stephane, Olivier Sterck, Gregoire Lits, and Thibaut Slingeneyer. "What Does the 'Terrorist' Label Really Do? Measuring and Explaining the Effects of the 'Terrorist' Category." Working paper, University of Exeter, UK 2015. Balzacq, Thierry. 2005. "The Three Faces of Securitization: Political Agency, Audience and Context." European Journal of International Relations 11 (2): 171-201. -. 2011a. "A Theory of Securitization: Origins, Core Assumptions, and Variants." In Securitization Theory: How Security Problems Emerge and Dissolve, edited by Thierry Balzacq, 1-30. London: Routledge. -. 2011b. "Enquiries into Methods. A New Framework for Securitization Analysis." In Securitization Theory: How Security Problems Emerge and Dissolve, edited by Thierry Balzacq, 31-53. London: Routledge. This question of what happens to participants after the experiments has been called by Stanley Milgram "subjects' reactions" in an article ironically calling for more attention on this problem (1977). An Experimental Agenda for Securitization Theory 663 -. 2014. "The Significance of Triangulation to Critical Security Studies." Critical Studies on Security 2 (3): 377-81. -. 2015. "The 'Essence' of Securitization: Theory, Ideal Type, and a Sociological Science of Security•." International Relations 29 (1): 103-13. Balzacq, Thierry, and Sarah Leonard. "Methodology and Progress in Securitization Studies." Working paper, University of Namur, Belgium, 2011. Balzacq, Thierry, Sarah Leonard, and Jan Ruzicka. 2016. "Securitization' Revisited: Theory and Cases." International Relations 30 (4): 494—531. Barkin Samuel, and Laura Sjoberg, eds. 2017. Interpretive Quantification: Methodological Explorations for Critical and Constructivist IR. Ann Arbor: University of Michigan Press. Berling, Trine Villumsen. 2011. "Science and Securitization: Objectivation, the Authority of the Speaker and Mobilization of Scientific Facts." Security Dialogue 42 (4): 385-97. Bigo, Didier. 2002. "Security and Immigration: Toward a Critique of the Governmentality of Unease." Alternatives 27: 63-92. Boettcher, William A., and Michael D. Cobb. 2006. "Echoes of Vietnam? Casualty Framing and Public Perceptions of Success and Failure in Iraq." Journal of Conflict Resolution 50 (6): 831-54. Bourdieu, Pierre. 1991. language and Symbolic Power. Cambridge: Polity Press. Brader, Ted. 2005. "Striking a Responsive Chord: How Political Ads Motivate and Persuade Voters by Appealing to Emotions." American Journal ofPolitical Science 49 (2): 388-405. Bragg, Belinda, Nehemia Geva, and Catarina Thomson. "Cooperation Under the Shadow of the Future: Experimental Analysis in the Context of Applied Issues." Working paper, Texas A&M University, University of College Station, Texas, 2016. Brendan, Nyhan, and Jason, Reifler (2015b) "The Effect of Fact-Checking on Elites: A Field Experiment on U.S. State Legislators", Americal Journal of Political Science 59 (3): 628-640. Buzan, Barry, Ole W^ever, and Jaap De Wilde. 1998. Security: A New Framework for Analysis. Boulder, CO: Lynne Rienner. Campbell, David. "The Visual Economy of HIV/AIDS." Last modified 2008, Accessed June 13, 2017. https: / / www.david-campbell. org/visual-hivaids/. Caverley, Jonathan, and Yanna Krupnikov. 2015. "Aiming at Doves: Experimental Evidence of Military Images' Political Effects." Journal of Conflict Resolution, doi: 10.1177/0022002715605634. Cheung-Blunden, Violet, and Bill Blunden. 2008. "Paving the Road to War with Group Membership, Appraisal Antecedents, and Anger." Aggressive Behavior 34 (2): 175-89. Chin, Michelle L., Jon R. Bond, and Nehemia Geva. 2000. "A Foot in the Door: An Experimental Study of PAC and Constituency Effects on Access." Journal of Politics 62 (2): 534—49. Chong, Dennis, and James N. Druckman. 2007. "Framing Theory." Annual Review of Political Science 10: 103-26. Clifford, Scott, and Jennifer Jerit. 2014. "Is There a Cost to Convenience? An Experimental Comparison of Data Quality in Laboratory and Online Studies." Journal of Experimental Political Science 1 (2): 120-31. Coan, Travis, Jennifer Merolla, Elizabeth Zechmeister, and Daniel Zizumbo-Colunga. 2016. "Distinct Emotions under Threat Lead to Distinct Types of Information Seeking." Working paper, University of Exeter, UK, 2016. Collett, Jessica L., and Ellen Childs. 2011. "Minding the Gap: Meaning, Affect, and the Potential Shortcomings of Vignettes." Social Science Research 40 (2): 513-22. Dan Kahan, Ellen Peters, Maggie Wittlin, Paul Slovic, Lisa Larrimore Ouellette, Donald Braman Gregory Mandel (2012) "The polarizing impact of science literacy and numeracy on perceived climate change risks", Nature Climate Change 2, pp. 732-735. DOI: 10.1038/NCLIMATE1547 Darley, John M., and C. Daniel Batson. 1973. "From Jerusalem to Jericho: A Study of Situational and Dispositional Variables in Helping Behavior." Journal of Personality and Social Psychology 27 (1): 100-8. Dickson, Eric. 2011. "Economics versus Psychology Experiments." In Handbook of Experimental Political Science, edited by James Druckman, Donald P. Green, James H. Kuklinski, Arthur Lupia. Cambridge: Cambridge University Press. Druckman, Daniel, and Benjamin J. Broome. 1991. "Value Differences and Conflict Resolution: Familiarity or Liking?" Journal of Conflict Resolution 35 (4): 571-93. Druckman, James N, and Rose McDermott. 2008. "Emotion and the Framing of Risky Choice." Political Behavior 30 (3): 297-321. Druckman, James N, Donald P. Green, James H. Kuklinski, and Arthur Lupia. 2006. "The Growth and Development of Experimental Research in Political Science." American Political Science Review 100 (4): 627-35. 664 Stephane J. Baele and Catarina P. Thomson Falk, Armin, and James J. Heckman. 2009. "Lab Experiments Are a Major Source of Knowledge in the Social Sciences." Science 326 (5952): 535-8. Floyd, Rita. 2011. "Can Securitization Theory be used in Normative Analysis? Towards a Just Securitization Theory." Security Dialogue 42 (4-5): 427-39. Geddes, Barbara. 1990. "How the Cases you Choose Affect the Answers You Get: Selection Bias in Comparative Politics." Political Analysis 2: 131-50. -. 2007. Paradigms and Sand Castles: Theory Building and Research Design in Comparative Politics. Ann Arbor: University of Michigan Press. Geva, Nehemia, James Mayhar, and Mark Skorick. 2000. "The Cognitive Calculus of Foreign Policy Decision Making: An Experimental Assessment." Journal of Peace Research 44 (4): 447-71. Guzzini, Stefano. 2011. "Securitization as a Causal Mechanism." Security Dialogue 42 (4—5): 329-41. Hainmueller, Jens, Dominik Hangartner, and Teppei Yamamoto. 2015. "Validating Vignette and Conjoint Survey Experiments Against Real-World Behavior." Proceedings of the National Academy of Sciences 112 (8): 2395-400. Hansen, Lene. 2011. "Theorizing the Image for Security Studies. Visual Securitization and the Muhammad Cartoon Crisis." European Journal of International Relations 17 (1): 51-74. -. 2014. "How Images Make World Politics: International Icons and the Case of Abu Ghraib." Review of International Studies 41 (2): 263-88. Herborth, Benjamin, and Oliver Kessler. 2010. "The Public Sphere." In International Studies Encyclopedia, edited by Robert Denemark. http://www.oxfordreference.com/view/10.1093/acref/ 9780191842665.001.0001/acref-9780191842665-e-0316?rskey«BdCvMc&result«316. elSBN: 9780191842665 HoLBRAAD, Martin, and Morten Axel Pedersen. 2012. "Revolutionary Securitization: An Anthropological Extension of Securitization Theory." International Theory 4 (2): 165-97. Hudson, Natalie Florea, and Michael J. Butler. 2010. "The State of Experimental Research in IR: An Analytical Survey." International Studies Review 12 (2): 165-92. Huysmans, Jef. 2011. "What's in an Act? On Security Speech Acts and Little Security Nothings." Security Dialogue 42 (4-5): 371-83. Jones, H, and L. Henriksen. 1987. "The Impact of the 'Scream': To Use or Not to Use." Health Education 18 (1): 9-11. Kahan, Dan M., and Donald Braman. 2003. "More Statistics, Less Persuasion: A Cultural Theory of Gun-Risk Perceptions." University of Pennsylvania Law Review 151 (4): 1291-327. Kahan, Dan M., Donald Braman, Paul Slovic, John Gastil, and Geoffrey L. Cohen. 2009. "Cultural Cognition of the Risks and Benefits of Nanotechnology." Nature Nanotechnology 4 (2): 87-91. Kahneman, Daniel, and Amos Tversky. 1984. "Choices, Values, and Frames." American Psychologist 39 (4): 341-50. Karyotis, Georgios, and Stratos Patrikios. 2010. "Religion, Securitization and Anti-immigration Attitudes: The Case of Greece." Journal of Peace Research 47 (1): 43-57. Lahav, Gallya, and Marie Courtemanche. 2011. "The Ideological Effects of Framing Threat on Immigration and Civil Liberties." Political Behavior 34 (3): 477-505. Lake, David A. 2013. "Theory is Dead, Long Live Theory: The End of the Great Debates and the Rise of Eclecticism in International Relations." European Journal of International Relations 19 (3): 567-87. Lanza, Stephanie T., and Brittany L. Rhoades. 2013. "Latent Class Analysis: An Alternative Perspective on Subgroup Analysis in Prevention and Treatment." Prevention Science 14 (2): 157-68. Leonard, Sarah, and Christian Kaunert. 2011. "Reconceptualizing the Audience in Securitization Theory." In Securitization Theory: How Security Problems Emerge and Dissolve, edited by Thierry Balzacq, 57-76. London: Routledge. McDermott, Rose. 2002a. "Experimental Methods in Political Science." Annual Review of Political Science 5: 31-61. -. 2002b. "Experimental Methodology in Political Science." Political Analysis 10 (4): 325-42. -. 2011a. "New Directions for Experimental Work in International Relations." International Studies Quarterly 55 (2): 503-20. -. 2011b. "Internal and External Validity." In Handbook of Experimental Political Science, edited by James M. Druckman, Donald P. Green, James Kuklinski, and Arthur Lupia, 27-40. Cambridge, MA: Cambridge University Press. McDermott, Rose, and Jonathan A. Cowden. 2001. "The Effects of Uncertainty and Sex in a Crisis Simulation Game." International Interactions 27 (4): 353-80. McDermott, Rose, Dominic Johnson, Jonathan Cowden, and Stephen Rosen. 2007. "Testosterone and Aggression in a Simulated Crisis Game." Annals of the American Academy of Political and Social Science 614 (1): 15-33. An Experimental Agenda for Securitization Theory 665 Milgram, Stanley. 1977. "Subject Reaction: The Neglected Factor in the Ethics of Experimentation." Hastings Center Report 7 (5): 19-23. Mintz, Alex, Yi Yang, and Rose McDermott. 2011. "Experimental Approaches to International Relations." International Studies Quarterly 55 (2): 493-501. Mnatsakanyan, Tatevik "Reconceiving Performativity in Securitization: 'The Shadow of Iraq' and a Dialogical-Relational Analysis of the Failed Syria Securitization in the British Parliament in August 2013." Working paper, University of Loughborough, England, 2015. Morton, Rebecca, and Kenneth Williams. 2010. Experimental Political Science and the Study of Causality: From Nature to the Lab. Cambridge, MA: Cambridge University Press. Nyhan, Brendan, and Jason Reifler. 2010. "When Corrections Fail: The Persistence of Political Misperceptions." Political Behavior 32 (2): 303-30. -. 2014. "The Effect of Fact-Checking on Elites: A Field Experiment on U.S. State Legislators." American fournal of Political Science 59 (3): 628-40. -. 2015b. "Does Correcting Myths about the Flu Vaccine Work? An Experimental Evaluation of the Effects of Corrective Information." Vaccine 33 (3): 459-64. Nyhan, Brendan, Jason Reifler, Sean Richey, and Gary L. Freed. 2014. "Effective Messages in Vaccine Promotion: A Randomized Trial." Pediatrics 133 (4): e835-42. Rashotte, Lisa Slattery. 2003. "Written Versus Visual Stimuli in the Study of Impression Formation." Social Science Research 32 (2): 278-93. Roberts, Gilbert, and Thomas Sherratt. 2002. "Behavioral Evolution: Does Similarity Breed Cooperation?" Nature 418 (6897): 499-500. Roe, Paul. 2008. "Actor, Audience (s) and Emergency Measures: Securitization and the UK's Decision to Invade Iraq." Security Dialogue 39 (6): 615-35. -. 2012. "Is Securitization a 'Negative' Concept? Revisiting the Normative Debate over Normal versus Extraordinary Politics." Security Dialogue 43 (3): 249-66. Salter, Mark B. 2008. "Securitization and Desecuritization: A Dramaturgical Analysis of the Canadian Air Transport Security Authority." fournal of International Relations and Development 11 (4): 321-49. Salter, Mark B., and Can Mutlu, eds. 2013. Research Methods in Critical Security Studies: An Introduction. New York: Routledge. Schlattman, Peter. 2009. Medical Applications of Finite Mixture Models. Berlin: Springer. Schultz, Kenneth A. 2001. "Looking for Audience Costs." fournal of Conflict Resolution 45(1): 32-60. Sniderman, Paul, and Sean Theriault. 2004. "The Structure of Political Argument and the Logic of Issue Framing." In Studies in Public Opinion, edited by Willem Saris, and Paul Sniderman, 133-65. Princeton, NJ: Princeton University Press. Theiler, Tobias. 2003. "Societal Security and Social Psychology." Review of International Studies 29 (2): 249-68. Thomson, Catarina P. 2017a. "Public Support for Economic and Military Coercion and Audience Costs." British fournal of Politics and International Relations 18 (2): 407-21. -. "Constraints on the Design of Security Policy: Audience Costs and the Strategic Defence and Security Review." Working paper, University of Exeter, England, 2017b. Thomson, Catarina P., Nehemia Geva, and Benjamin Tkach. "In-Group Bias in Public Support for Coercive Interrogation in Wartime." Working paper, University of Exeter, England, 2017. Thomson, Catarina P., Gabriel Katz, Daniel Stevens, Susan Banducci, and Travis Coan. "Audience Costs and the EU Referendum." Working paper, University of Exeter, England, 2017. Tomz, Michael. 2007. "Domestic Audience Costs in International Relations: An Experimental Approach." International Organization 61 (4): 821-40. Tracer, Robert F., and Lynn Vavreck. 2011. "The Political Costs of Crisis Bargaining: Presidential Rhetoric and the Role of Party." American fournal of Political Science 55 (3): 526-45. Tversky, A., and D. Kahneman. 1981. "The Framing of Decisions and the Psychology of Choice." Science 211 (4481): 453-8. Van Rythoven, Eric. 2015. "Learning to Feel, Learning to Fear? Emotions, Imaginaries, and Limits in the Politics of Securitization." Security Dialogue 46 (5): 458-75. Vultee, Fred. 2010. "Securitization: A New Approach to the Framing of the War on Terror.'" fournalism Practice 4 (1): 33-47. -. 2011. "Securitization as a Media Frame." In Securitization Theory: How Security Problems Emerge and Dissolve, edited by Thierry Balzacq, 77-93. London: Routledge. Vuori, Juha A. 2008. "Elocutionary Logic and Strands of Securitization: Applying the Theory of Securitization to the Study of Non-Democratic Political Orders." European fournal of International Relations 14 (1): 65-99. 666 Stephane J. Baele and Catarina P. Thomson W^ever, Ole. 1995. "Securitization and Desecuritization." In On Security, edited by Ronnie Lipschutz, 46-86. New York: Columbia University Press. -. 2011. "Politics, Security, Theory." Security Dialogue 42 (4—5): 465-80. -. 2015. "The Theory Act: Responsibility and Exactitude As Seen From Securitization." International Relations 29 (1): 121-7. Watson, Scott D. 2012. "Framing the Copenhagen School: Integrating the Literature on Threat Construction." Millennium: Journal of International Studies 40 (2): 279-301. Weeks, Jessica L. 2008. "Autocratic Audience Costs: Regime Type and Signaling Resolve." International Organization 62 (1): 35-64. Williams, Michael C. 2003. "Words, Images, Enemies: Securitization in International Politics." International Studies Quarterly 47 (4): 511-31.