Soldiers' Dilemma: Foreign Military Training and Liberal Norm Conflict

On April 14, 1979, the “rice riots” broke out in Monrovia, Liberia. Several thousand protestors took to the streets in frustration and desperation over a government proposal to increase the price of rice, a food staple that most Liberians relied on for subsistence. As demonstrators marched toward the president's mansion, President William Tolbert ordered soldiers to fire into the crowd.¹ Over forty protestors were killed in the ensuing chaos. Although the soldiers obeyed Tolbert's order, the command to shoot their fellow Liberians sparked deep animosity and contributed to a bloody coup the following year.²

These rice riots illustrate the conflict that can occur between the two liberal norms of respect for human rights and civilian control of the military. When political leaders order militaries to harm the population, these two norms enter into conflict. Militaries are expected to obey political leaders, and they are also expected to protect populations. When the rice riots broke out in 1979, the United States had been training the Liberian military for eighteen years. The goal was to create a disciplined, democratic force, and this meant training the military to both respect human rights and obey civilian authority.³ Tolbert's order pitted these norms against each other, presenting the military with a dilemma. This article examines how norm conflict drives militaries to prioritize cohesion instead of either liberal norm.

Liberal great powers fail to acknowledge this dilemma as they seek to spread liberal norms globally through foreign military training. Foreign military training is a flexible form of security assistance that seeks to modify recipient militaries' behavior by increasing warfighting capacity and transmitting a set of professional norms or ideas about standards of appropriate behavior.⁴ Norm content and emphasis varies across providers, time, and space. For liberal democratic providers such as the United States, two of the most salient norms are civilian control of the military and respect for human rights.⁵ The theory and empirics that follow focus on the United States as a key liberal power. While the argument applies to all liberal providers, and all great powers use training as part of their security assistance repertoire, the United States is the largest such provider.⁶ In 2019, for example, the United States spent nearly $905 million to train over 71,000 military students from 157 states.⁷

Today, the United States ostensibly promotes these two liberal norms as a primary foreign policy objective to nearly half of all states that receive its military training.⁸ A mix of idealistic and strategic motives explains this emphasis. Imparting liberal norms is partially bound up in Western states' efforts to promote global democracy after the Cold War, which included providing security assistance to liberalize militaries.⁹ But imparting liberal norms is also strategic. Shaping how partner militaries think is an inexpensive way to gain voluntary policy compliance.¹⁰ Training also helps to inculcate a shared military identity, common skills, and communication, indirectly strengthening a third norm of cohesion. Training can thus reduce the costs of security management by creating more competent, cohesive, disciplined, and loyal partners.¹¹

The empirical record, however, suggests that training often fails to deliver on its promise of liberal norm transmission. Security assistance frequently produces noncompliant, norm-violating militaries that conduct coups and abuse human rights.¹² Policymakers and scholars offer divergent explanations for these shortcomings. Some suggest that norm violations result from insufficient investment in training or inadequate emphasis on norm socialization.¹³ Most scholars, conversely, favor rationalist arguments that point to interest misalignment between providers and recipients.¹⁴ While these arguments ascribe norm violations to different mechanisms, they share a common assumption that socialization never occurred in the first place. This assumption, however, overlooks the dilemma that arises when these two liberal norms conflict.

In this article, I argue that norm conflict weakens military support for both liberal norms, thus creating the conditions under which perverse behavioral outcomes can occur despite socialization. The implicit expectation in U.S. policy, should norm conflict arise, is that militaries will privilege human rights over civilian authority, temporarily defying civilian leaders to protect the higher order of rights “rooted in natural law.”¹⁵ This expectation rests on the assumptions that trained militaries share this rank-ordering, will easily choose human rights over civilian control, and that norm conflict will not damage either norm.¹⁶ If anything, conflict might clarify norms and promote socialization.¹⁷

I contend that these assumptions rest on shaky foundations. Rather than strengthening liberal norms, I argue that conflict between norms makes them less salient and undermines their power to restrain decision-making.¹⁸ Norm conflict invites cost-benefit calculations and creates openings to pursue self-interest.¹⁹ Conflict thus produces incentives for militaries to strategically choose among norms in an effort to satisfy both norms and interests. When this happens, they are more likely to fall back on the third norm of cohesion, a norm that is tied to interests. Fraught domestic crises that pit liberal norms against each other may further increase the salience of cohesion. Rank-ordering happens, but not in the direction that the United States expects.

If a military prioritizes cohesion, then it will choose the path that best serves its organization, which may entail violating human rights, civilian control, or both. I thus seek to explain shifts in military support for liberal norms rather than to predict specific instances of disobedience or abuse in any given political context. I examine the conditions under which norm violations become possible and shed light on the links between socialization efforts, attitudes, and ultimate behavior.²⁰ Norm conflict, I argue, can weaken liberal norms and frustrate socialization over the long run.²¹ My study explores the determinants of decision-making to explain how cohesion comes to occupy a central place in military concerns, over and above liberal normative considerations.²²

As a plausibility probe of these competing claims, I use experimental evidence from a survey of the Armed Forces of Liberia (AFL). The United States rebuilt the AFL after Liberia's civil war ended in 2003. U.S. trainers heavily emphasized liberal norms, making Liberia a most-likely case for norm socialization. The survey experiment presents soldiers with a scenario in which a political leader orders the military to put down protests with force, pitting the norms of civilian control and respect for human rights against each other. The scenario thus provides an experimental “stress test” of competing norms.²³ Additionally, I conducted over fifty elite-level interviews with U.S. officials, trainers, and Liberian military officers.

My research indicates that when soldiers hear this scenario, their willingness to choose between conflicting liberal norms decreases, and they express more concern for maintaining cohesion. Liberal training conditions this response but in unexpected ways: Soldiers with more U.S. training express less willingness to prioritize human rights and are less supportive of democratic norms.²⁴ Importantly, the survey evidence also shows that soldiers with more U.S. training express the strongest support for liberal norms in the absence of norm conflict, undermining the alternative argument that socialization never happened.

These findings shed light on a pathway by which norm violations can happen even in the presence of socialization. By highlighting a previously unrecognized dilemma in the model of civil-military relations that liberal powers seek to export, I contribute to studies that explore contradictions and complexity in military norms,²⁵ as well as studies that examine how norms and interests interact and shape behavior.²⁶ Liberal great powers typically consider the norms of civilian control of the military and respect for human rights to be mutually reinforcing, even though they often are not, particularly in weak democracies. My findings suggest that norm conflicts can have a corrosive effect on military support for liberal norms by incentivizing militaries to privilege a third norm of cohesion. This outcome indicates that the assumptions underlying a major component of U.S. foreign policy may be flawed, and it suggests the need to fundamentally reevaluate the approach to norm transmission.

The remainder of this article is organized as follows. First, I describe liberal powers' efforts to impart ideas about civilian control of the military and respect for human rights to the militaries that they train. Second, I lay out my argument about the implications of conflict between these two liberal norms—situating it within the literatures on norm contestation, psychology, and sociology—and I generate testable hypotheses. Third, I discuss research design, including case selection. The fourth section presents the experimental results and explores mechanisms that aid in interpreting the results. I conclude with a discussion of implications for theory and policy.

Since the end of the Cold War, foreign military training by liberal democratic powers has increasingly focused on promoting norms.²⁷ British policy, for example, has stated that “support to build the capacity of security forces must be matched with efforts to build accountability, legitimacy and respect for human rights.”²⁸ Similarly, Canada's Military Training and Cooperation Program seeks to “promote Canadian democratic principles, the rule of law and the protection of human rights.”²⁹ The United States likewise emphasizes human rights and civilian control of the military. The Barack Obama administration articulated a goal of U.S. security sector assistance as promoting “universal values, such as good governance, transparent and accountable oversight of security forces, rule of law … and respect for human rights.”³⁰ As of 2017, U.S. law requires that all efforts to build foreign military capacity include training on “observance of and respect for the law of armed conflict, human rights and fundamental freedoms, the rule of law, and civilian control of the military.”³¹

Importantly, the United States emphasizes norms training in different regions. I find that this training tends to occur in places where policymakers believe that more liberal and professional militaries are key to long-term political stability and democracy, or where they need to couch security assistance in liberal values to sell U.S. involvement.

Adherence to the liberal norms of human rights and civilian control of the military were primary foreign policy objectives of U.S. training in sixty-two states, as depicted in figure 1.³² Almost 90 percent of states in Africa receive training on these norms each year. This prominence suggests that the United States uses norms training in those places where it would prefer to delegate security management and avoid intervention.³³ Because military training also imparts human capital and technical skills, the United States hopes to simultaneously create more competent and more liberal armed forces, capable of providing security while avoiding repressive and corrupt behavior that can jeopardize stability.

Foreign military training is a preferred tool to impart norms.³⁴ Whereas other forms of security assistance such as arms transfers may come with conditions to modify behavior or extract policy concessions, training seeks to change military preferences primarily through socialization. Training offers opportunities for teaching and persuasion, which can be powerful in shaping professional worldviews.³⁵ “Professional training,” according to Martha Finnemore and Kathryn Sikkink, “actively socializes people to value certain things above others.”³⁶ Trainers use their expertise, information, and resources to promote certain norms while discouraging others.³⁷ Consistent messaging plays a role; as a military instructor tasked with training the new AFL put it, “You tell somebody something long enough, they'll believe it. You have to capture their mind for their body to follow.”³⁸ Additionally, training offers opportunities for interpersonal interaction and relationship-building, which may help foster shared preferences.³⁹

Training takes diverse forms. Military officers attending regional seminars or courses in U.S. military schools are exposed to liberal norms through programs such as the International Military Education and Training program.⁴⁰ Tactical training can also include a normative component. For example, when the U.S. military trains African counterterrorism forces, they spend up to a week on human rights education in the context of targeting practices.⁴¹ Finally, trainers try to tailor their application of norms to the local context. In Liberia, for example, Michigan National Guard advisers explained human rights in terms such as “it's wrong to shoot individuals with 50-caliber bullets.”⁴²

But U.S. training imparts more than liberal norms—it also helps to cultivate cohesion, which refers to the bonds that enable military forces to operate in a unified, group- and mission-oriented way.⁴³ The imperative to preserve these bonds functions as a norm that motivates soldiers to prioritize loyalty and commitment to their unit, the military, and shared goals. It is almost always in the individual's best interest to pursue cohesion, even though doing so can sometimes require self-sacrifice for either the group or certain strategic goals.⁴⁴ Indeed, at the military institutional level, cohesion represents a core interest. “Cohesion, discipline, and morale within the corps” are key to the military's self-preservation as well as its ability to secure other organizational interests.⁴⁵ As David Pion-Berlin et al. note, “preservation of institutional unity has always been the centerpiece of military interests.”⁴⁶

Cohesion functions as both a military norm (particularly at the individual level) and an interest (particularly at the institutional level). Because this article's theory and tests focus on individual decision-making, I treat cohesion as a norm, albeit one that is aligned to interests in ways that are difficult to fully separate.⁴⁷

Cohesion reflects societal and political factors as well as military factors such as operational practices and training.⁴⁸ This article focuses on training, which fosters a common military identity, skills, and modes of communication that help to create cohesion over time.⁴⁹ Cohesion is not taught as a principle to follow (e.g., “do not abuse civilians”); rather, it is instilled through training and practices that emphasize solidarity, shared identity, and teamwork to accomplish goals. Over time, this emphasis on solidarity, shared identity, and teamwork contributes to a sensemaking process through which cohesion emerges as a norm. In Liberia, for example, trainers frequently used competition between small groups to both foster team-building and hold squad members accountable to one another.⁵⁰ AFL officers described basic training as a “culture shock”⁵¹ and that training ideas ranged from “intriguing” to “goofy,” such as doing push-ups as punishment for not shining boots. Eventually, however, the need for discipline and teamwork began to make sense to the trainees.⁵² Although U.S. training does not explicitly seek to teach cohesion in the same way that it teaches liberal norms, it increases capacity for unified action, strengthens organizational interests, and it indirectly imparts a third norm of cohesion.

The return on investment for imparting liberal norms often appears meager. Evidence suggests that U.S.-trained militaries launch coups, subvert the rule of law, and oppress the people they are supposed to protect.⁵³ Infamously, the U.S. Army's School of the Americas—created in 1946 to educate Latin American military officers on “the virtues of democratic civilian control over the armed forces”—produced a generation of coup-makers instead.⁵⁴

The international relations literature offers contrasting explanations for norm violations. One set of arguments suggests that the problem is insufficient training or attention to norms.⁵⁵ As the head of U.S. Africa Command said after U.S.-trained soldiers were implicated in the 2012 coup d'état in Mali: “We didn't spend, probably, the requisite time focusing on values, ethics, and military ethos.”⁵⁶ In this view, norm violations reflect a failure to successfully socialize militaries to new norms in the first place. A second set of arguments, the rationalist perspective, uses a principal-agent framework to explain recipient behavior.⁵⁷ Providers function as principals that arm and equip weaker states to act as agents on their behalf, but divergent priorities and asymmetric information lead to agent-shirking and noncompliance.⁵⁸ These studies suggest that norm transmission, if attempted at all, will be anemic compared with the powerful interests that motivate behavior.

Both explanations oversimplify decision-making in different but problematic ways. The first perspective, which emphasizes socialization, discounts the power of interests in motivating behavior. It also fails to consider which norms will exercise the most influence over decision-making. Conversely, the rationalist perspective discounts the role of norms in shaping preferences and constraining behavior. In contrast, I treat military decision-making as the joint product of norms and interests, contributing to scholarship that examines the conditions under which utilitarian or normative considerations govern decision-making.⁵⁹ Next, I look at how conflict between two liberal norms weakens support for those norms and opens a pathway for other, more utilitarian norms to dominate.⁶⁰

I argue that in moments of norm conflict, militaries will prioritize cohesion over the two liberal norms of respect for human rights and civilian control of the military. My argument begins with the premise that norms mediate decision-making because they serve as social constraints that people follow for psychological and sociological reasons. Nonconformity with group norms can lead to social sanctions and shaming;⁶¹ violating personal beliefs can affect self-esteem.⁶² People thus forgo pursuit of naked self-interest to maintain their positive self-image and social standing. Similarly, collective actors comply with norms to legitimize their behavior and minimize resistance from other actors.⁶³

Determining which norms to follow, however, is not always clear. There are many different norms with potentially countervailing implications for behavior—what Paul Kowert and Jeffrey Legro call the “ubiquity” of norms problem.⁶⁴ The problem of conflicting norms is well-noted in the literatures on law, sociology, and international relations.⁶⁵ For example, the international relations literature has explored the tension between the norm of state sovereignty and humanitarian norms such as the “responsibility to protect.”⁶⁶ Moreover, following certain norms may come at a cost to other values, even if the decision is not framed as a choice between them.⁶⁷ As Jon Elster puts it, “In some contexts, following the lodestar of outcome-oriented rationality is easy compared with finding one's way in a jungle of social norms.”⁶⁸

Surprisingly, scholars and policymakers have neglected the problem of norm conflict that is at the heart of liberal civil-military relations.⁶⁹ In the liberal formulation, the norms of human rights and civilian authority are mutually reinforcing. The empirical record underscores this assumption: where the United States provides norms training, it typically bundles both norms as a package deal. Almost every country in sub-Saharan Africa, for example, receives military training on both norms every year (see figure 1).

Liberalism's inherent support for civilian control of the military assumes that this civilian control represents the best way to protect both human rights and democracy.⁷⁰ Although these two norms usually are mutually reinforcing in strong democracies, this assumption may fail in precisely the places where norms training is most prevalent: weak democracies and fragile states. Such states often lack strong legal institutions to rein in executive leaders, and their militaries often lack robust military justice systems to establish the legality of orders. Despite Eric Nordlinger's warning that “liberalism's abiding and indiscriminate preference for civilian control is a debatable issue,” the problem of norm conflict in liberal civil-military relations has avoided scrutiny.⁷¹

When situations arise with conflicting implications for behavior, militaries face consequential choices without clear answers. The following section examines competing predictions about the effects of norm conflict on military decision-making.

How will militaries respond to the conflicting norms of respect for human rights and civilian control? Although U.S. policy discourse rarely broaches the problem of norm conflict, there is an implicit expectation, should contradictions arise, that the military will temporarily prioritize human rights rather than defer to civilian authority. This expectation derives from liberal understandings of natural law; as the Defense Department notes, “the English philosopher John Locke believed that human rights, not governments, came first in the natural order of things.”⁷² Given that orders to abuse human rights violate natural law, the military can justifiably disobey such presumably illegal orders. This understanding reflects a long tradition in U.S. military norms, which holds that military personnel must disobey orders that are “manifestly illegal.”⁷³ Less clear is whether officers should obey orders that they perceive to be immoral but not necessarily illegal.⁷⁴ Still, the preference ordering is clear: people first, governments second. If norms conflict, the military should choose human rights over civilian control.

There is some evidence that the United States tries to impart this rank-ordering of norms. In testimony before U.S. Congress, the commander of U.S. Africa Command said, “We recognize building legitimate defense institutions is critical for African governments that prioritize the security of their citizens over that of the state” (emphasis added).⁷⁵ Evidence from my interviews suggests that U.S. officers at least occasionally socialize other militaries to follow only legal orders. For example, the AFL's top officers, intensively trained by the United States, said that the United States prepared the AFL to be “very bold” in telling political leaders when they were asking the military to do something they could not or should not do. In contrast, one officer said, other actors in Liberia lacked “courage” to defy illegal orders.⁷⁶

The U.S. policy expectation that militaries will choose human rights rests upon two assumptions. First, it assumes that choosing between conflicting norms is easy. At the individual level, people are presumed to choose between values with relative ease. Although people may resist substituting deeply held values for less important ones, when the choice is between two important values, they can frame the problem as a “tragic tradeoff” and avoid damage to self-esteem or social standing.⁷⁷ Moreover, the choice of which norm to privilege should be facilitated by the presence of a blueprint for prioritization (“people first”).

The second assumption is that norm conflict will not damage either norm. Rank-ordering norms in a crisis should not make the violated norm any less salient; rather, it is temporarily downgraded to accommodate the more important principle in that moment.⁷⁸ If anything, the process of adjudicating between the norms might strengthen them both. Research on norm contestation suggests that norm conflicts can drive norm violations, but these violations ultimately strengthen socialization in the long term by clarifying ambiguity and resolving future norm conflict.⁷⁹

Insomuch as liberal powers think about other norms and interests beyond these two central liberal norms, they treat them as unproblematic for crisis decision-making. Norm conflicts are not viewed as problematic for cohesion; if anything, cohesion should make militaries more unified in doing the right thing. Properly socialized military personnel, according to the liberal training logic, should be willing to make costly choices even as liberal normative considerations should continue to guide decisions.

Counter to U.S. policy expectations, I argue that these assumptions rest on shaky foundations. Instead, when liberal norms clash, military members fall back on a third norm of cohesion that is consistent with group interests. Rank-ordering of norms does happen but not in the direction that the United States expects. Rather than choosing human rights over civilian control, military decision-making is less likely to be constrained by either liberal norm in favor of cohesion.

Norm conflicts make it more likely for soldiers to fall back on cohesion in at least two ways. First, conflict makes the contested norms less salient during the crisis. The assumption that norm conflict does not adversely affect norms misses the alternative possibility that it makes them easier to ignore instead.⁸⁰ Scholarship has shown that norm conflicts can weaken support for prohibition norms in warfare, such as norms against the use of chemical weapons.⁸¹ Contestation eliminates the “taken for granted” quality of norms; once challenged, norms may no longer “operate as either a focal point for mutual expectations or as a naturalized guide for behavior.”⁸¹ Conflict between liberal norms invites cost-benefit calculations because the norms are no longer unquestionable. It can be costly to choose between either disobeying leaders and protecting civilians or, conversely, obeying leaders and brutalizing civilians. Weighing costs and benefits thus creates incentives to instrumentally follow other norms that best serve soldiers' interests. Militaries have another such norm as recourse: cohesion, which is neither inherently liberal nor conflictive with their self-interest.

Second, liberal norm conflicts are often situations that also threaten cohesion. Crises that pit political leaders against the population, with the military in the middle, are fraught situations that can threaten cohesion. Disagreement over whether to obey orders may create fears of fissures in the ranks.⁸³ Individuals may be reluctant to make choices that they perceive as potentially divisive and harmful to cohesion. Indeed, insights from the literature on military obedience show that militaries often prioritize cohesion when deciding whether to obey or defect during crises.⁸⁴ I argue that psychological dynamics explain why cohesion dominates decision-making. Structurally, cohesion becomes more salient even as the clashing liberal norms become less salient.

Prioritizing cohesion offers military members a way to behave normatively while protecting their interests and preserving the organization. Contrary to a purely instrumental logic under which norms are cast aside as soon as they clash with interests, people still try to justify their actions in normative terms.⁸⁵ But people can also choose strategically among norms. Because cohesion satisfies both interests and norms, I argue that it is more likely to be selected in a crisis. This dynamic explains both the ascendance of cohesion and interests to the forefront of decision-making and the descendance, even if temporary, of regard for liberal norms. This dynamic does not mean that cohesion is fundamentally at odds with liberal norms, nor does it automatically predict norm violations. Rather, conflict between liberal norms creates the conditions that make violations possible. By vitiating liberal normative restraints, military crisis decision-making will be driven by whatever best serves cohesion, which may involve violating human rights, civilian control, or both norms. While this argument seeks to explain changes in support for norms rather than specific behavioral outcomes, I consider implications for behavior below.

Finally, several factors may exacerbate norm conflict. First, the liberal preference ordering of “people first, governments second” is not always communicated clearly in practice. U.S.-trained militaries are not taught to choose human rights nearly as systematically as the rhetoric suggests. One reason for the disconnect between theory and practice is that policymakers may deny the possibility of norm tradeoffs between human rights and civilian control because of the tendency to engage in wishful thinking.⁸⁶ In private conversations, U.S. military officers have described this potential for norm conflict as a “conundrum” or “third rail” in security assistance, admitting that they try to avoid addressing it altogether. When civilians order the military to do things that are illegal, U.S. advice tends to be inchoate, offering “fifty-four different answers” depending on the situation.⁸⁷

Another reason for the disconnect may be divergent preferences between the U.S. military and its own civilian leaders; the Defense Department's public Lockean position has been undercut in unofficial presidential rhetoric.⁸⁸ Bureaucratic silos may also play a role: Different entities conduct U.S. training on civil-military relations and on human rights.

Second, militaries with histories of factionalism or civil war may be more prone to prioritizing cohesion. Not only might these militaries be more sensitive to the risk of internal rifts, but they often lack robust military justice systems and doctrines to evaluate the legality of orders and provide guidance for action. Illustrative evidence points to this kind of sensitivity in the Liberia case. When asked in the survey what the Liberian military's motto (“A Force for Good”) meant to them, many soldiers' responses emphasized cohesion: “a force that can't ever be factionalized,” “united,” and “here to stay.”

I predict that norm conflict will weaken support for liberal norms, undermining their robustness by making them less salient in crises.⁸⁹ Rather than prompting soldiers to choose human rights over civilian control, as the United States expects, moments of norm conflict are more likely to undermine support for both norms and prompt militaries to select a third norm, cohesion, which is consistent with their interests. This argument produces the following hypotheses:

H1: Moments of norm conflict reduce soldiers' willingness to prioritize human rights over civilian control.

H2: Moments of norm conflict reduce soldiers' support for democratic norms.

H3: Moments of norm conflict increase soldiers' prioritization of cohesion.

In contrast, U.S. policy expectations predict that soldiers experiencing norm conflict will prioritize human rights. Although this choice temporarily rankorders human rights over civilian control, norm conflicts may enhance—or at least not undermine—support for democratic norms. This outcome suggests the following hypotheses:

H4: Moments of norm conflict increase soldiers' willingness to prioritize human rights over civilian control.

H5: Moments of norm conflict increase soldiers' support for democratic norms.

Finally, I consider how liberal foreign military training might condition how militaries respond to norm conflict. On the one hand, insomuch as training promotes liberal norms and actively socializes a norm hierarchy of “people first,” then U.S. training should strengthen the propensity to prioritize human rights. On the other hand, training cultivates a common identity and solidarity at the individual level, increases military autonomy, and indirectly bolsters organizational interests. Should norm conflict arise, training may strengthen cohesion, at the expense of liberal norms. These alternatives suggest two final hypotheses:

H6a: Soldiers with U.S. training will be less likely to prioritize human rights or support democratic norms in moments of norm conflict.

H6b: Soldiers with U.S. training will be more likely to prioritize human rights and support democratic norms in moments of norm conflict.

Before turning to the empirical analyses, I consider the conditions under which norm conflicts are more likely to occur and where their effects may be more severe. The first two scope conditions are regime type and institutional strength—the likelihood and severity of norm conflict may increase in weak democracies. First, democratic or pseudo-democratic⁹⁰ states typically at least attempt (or feign) to adhere to both norms. These are also the states that are most likely to receive norms training, which means that the seeds of the dilemma are present. Other mechanisms can drive norm violations, but for norm conflict to occur, both liberal norms must be present—otherwise there is no contradiction. Further, sometimes the preferences of the government and the population diverge in new or weak democracies, and these divergences may prompt elites to use the military against the public.⁹¹

Moreover, such states typically have weak institutions and fragile rule of law. Strong courts, legislatures, and military justice systems both constrain leaders from making illegal orders and hold militaries accountable. These institutional restraints are one reason why norm conflicts surface less frequently in established democracies and, if they do, why their effects tend to be less severe. For example, when U.S. President Donald Trump suggested in 2018 that soldiers would shoot any migrants who threw rocks along the Mexican border, military leaders and analysts quickly emphasized that such a disproportionate response would be illegal.⁹² Even if norm conflicts vitiate normative considerations, institutions restrain behavior. The U.S. Uniform Code of Military Justice, for example, provides legal guidance for military decision-making. Robust legal institutions capable of adjudicating between such conflicts, however, are almost always missing in weak states where most norms training occurs.

The military's orientation toward external defense or internal security may also affect the likelihood of norm conflict and the severity of its effects. Militaries tasked with law enforcement or regime protection are far more likely to face crisis scenarios where norm conflicts can surface.⁹³ Conversely, established democracies generally avoid using the military for domestic law enforcement. Tension between respect for human rights and civilian control of the military would not seem as remote a possibility in the United States if the military were used for law enforcement, a problem avoided by the 1878 Posse Comitatus Act, which banned use of the military in policing functions. Even still, the United States has had historic flashpoints during which states have used their national guards to repress protests or to resist desegregation.⁹⁴ President Trump's deployment of active-duty troops to Washington, D.C., in June 2020 to contain protests against police brutality and systemic racism prompted one analyst to write: “Such action weakens the fundamental contract between the military and the American people.”⁹⁵

The theoretical argument presented here focuses on military decision-making rather than military behavior. The argument does not predict which norms may be violated in a given crisis, only that conflict weakens liberal norms, which increases the likelihood that norm violations will occur. To explore the empirical implications, the analyses that follow look at the pathways by which individual shifts in attitudes occur.

The decision to focus on attitudinal shifts, rather than behavioral outcomes, has theoretical justification. Studying how beliefs and attitudes change is valuable in its own right. We cannot assume that beliefs change just because behavior changes (or vice versa). Robert Jervis noted that while beliefs often drive behavior, “such a correspondence is not automatic.”⁹⁶ Studying each stage separately allows scholars to assess whether and how changes in attitudes affect behavior.⁹⁷ This separation is particularly important when dealing with complex social processes such as military behavior in domestic crises. This separation also has precedent in the psychology literature, which tends to study either attitude changes as the dependent variable or the effects of attitudes on behavior as independent or intervening variables.⁹⁸

While this article focuses on how norm conflict affects military attitudes toward norms (rather than behavior), a large psychology literature shows that attitudes are strong predictors of intentions, which in turn shape behavior.⁹⁹ These studies show that when people have competing attitudes, they act according to their strongest attitude.¹⁰⁰

These insights allow us to consider the predicted behavioral implications, although direct tests remain an important step for future research. On the one hand, the U.S. policy expectation is that militaries will continue to prioritize liberal norms, which should in turn continue to shape behavior in benign ways. Even if prioritizing human rights means defying civilian orders, we would expect to see fewer abuses and less intervention if the policy expectation is correct that temporarily choosing human rights does not erode the norm of civilian control. On the other hand, if norm conflict reduces support for liberal norms and soldiers prioritize cohesion instead, the potential for norm-violating behavior increases. Cohesion does not automatically predict liberal norm violations, but it does predict that soldiers will do whatever best serves cohesion, whether that means violating one or both liberal norms.

A broad comparative literature picks up the question of behavioral outcomes, investigating what happens when militaries repress revolutions or defect from the regime instead.¹⁰¹ These works examine how the regime, and those in opposition to the regime, can activate different military interests and identities.¹⁰² Many studies suggest that the decision to defect hinges on military cohesion and whether militaries think that repression will undermine cohesion.¹⁰³ In short, this literature explains the paths that behavior can take once cohesion is a dominant factor in military decision-making.

Cases in Tunisia and Egypt tentatively illustrate these divergent behavioral outcomes. In Tunisia, mass protests in 2011 jeopardized Zine El Abidine Ben Ali's grasp on power. Rather than repress the protests, the military withdrew support from the regime, allowing Ben Ali to fall—but the decision to prioritize human rights did not correspond with reduced support for civilian control.¹⁰⁴ In short, the behavioral outcomes in this case aligned to U.S. expectations. Egypt's case initially followed a similar trajectory. When protests erupted against Hosni Mubarak in 2011, the chief of staff of the armed forces assured the U.S. military “that the armed forces would defend Egyptian institutions, not individuals, and would not open fire on civilians.”¹⁰⁵ Yet after Mubarak's fall and democratic elections in 2012, the military intervened again a year later, seizing power. Under General Abdel Fattah al-Sisi, elected president in 2014, human rights abuses soared, pointing to an erosion of both liberal norms. The argument advanced in this article presents a causal pathway by which liberal norm conflict may have created the conditions that allowed interests to dominate the Egyptian military's decision-making. The Egyptian case is noteworthy because it serves as a behavioral model for other militaries. In an interview with two AFL officers, I asked what they might do if the norms of human rights and civilian control came into conflict. Without pausing, they answered: “Egypt [the Egyptian military in 2011].”¹⁰⁶

To explore the seven hypotheses, I designed an experiment embedded in a survey of active-duty AFL personnel across military bases in Liberia. The experiment primes soldiers in the treatment group to think about liberal norm conflict by asking them to consider a scenario in which the president orders the military to put down protests with force. To avoid potentially biased responses, the soldiers were asked to evaluate the commander's response rather than provide their own preferences directly.¹⁰⁷

In Liberia, liberal norm conflict might not only occur but also lead to pernicious outcomes. Dysfunctional civil-military relations contributed to fourteen years of civil war that ravaged the country between 1989 and 2003.¹⁰⁸ Today, Liberia is a transitional democracy; elections in 2017 marked the first peaceful transfer of power in over seventy years.¹⁰⁹ Rule of law has yet to fully mature; as of 2017 (when the survey was conducted), the AFL still lacked a functioning military justice system.¹¹⁰ Although Liberia is a typical case for liberal norm conflict, it is not an easy case for my theory. After the war, the United States rebuilt the AFL.¹¹¹ The U.S. training program heavily emphasized liberal norms, socializing the Liberian military to respect human rights and civilian authority. The program architects believed that these norms were key to preventing a repeat of the brutal civil war; by trainers' accounts, the message was well received.¹¹² The war had shattered old norms and AFL recruits were eager for change, making Liberia a most-likely case in which liberal norms could take root. Interviews with the AFL leadership further reveal that the United States tried to instill a preference ordering among the norms, teaching the AFL to obey only legal orders.

Liberia is also a valuable case for assessing the effects of U.S. training because of the high level of U.S. involvement, which helps to shed light on causal pathways. Cases with “extreme” values on independent variables are useful for illustrating mechanisms.¹¹³ Furthermore, as exposure to postwar U.S. training varied across recruits, who joined the military in eight waves, there is built-in variation in the army-building program, which tapered off partially by design and partially because it ran out of money before completion.¹¹⁴ In the end, roughly one-third of the rebuilt AFL received the full U.S. training sequence, one-third received partial U.S. training, and one-third received no U.S. training. I leverage this variation to explore differences in experimental treatment effects across levels of training.

I partnered with a local firm, Q&A, to administer the survey in December 2017 to 270 AFL respondents.¹¹⁵ A multistage sampling process drew a random sample of respondents from every unit roster on bases across Liberia.¹¹⁶ AFL rosters indicated that it had fewer than 2,000 soldiers, and thus the sample represents around 15 percent of the force. Because of the small sample size, I treat the results that follow as a plausibility probe of the argument. The survey sample is representative of the force, with AFL recruitment group, rank, and gender distributions tracking closely to the population. Ranks in the sample ranged from private to colonel; the median rank of the survey respondents was specialist, and the median age was thirty-seven years old.¹¹⁷

I designed the experiment to prime respondents to recognize a conflict between the liberal norms of human rights and civilian control. Respondents who were randomly assigned to the treatment group heard the following scenario:

After a big tariff increase, local business owners go on strike and there are protests in the streets. The Liberian National Police are managing it, but the president wants to send a forceful message and calls on the military to intervene to stop the protests. The military commander refuses to send soldiers into the streets to stop the protests.

Immediately afterward, as part of the treatment condition, respondents selected the statement that was the closest match to their own opinion about the case: “The military should not have intervened, it was an illegal order”; “The military should have intervened, it was a legal order”; or “It does not matter, as long as the military follows the commander's orders and sticks together.”

I worked with AFL leadership to identify and design a treatment scenario that was a salient example of norm conflict (i.e., an illegal order that soldiers should refuse to obey). The AFL's constitutional role is modeled on that of the United States; it similarly restricts military involvement in domestic law enforcement.¹¹⁸ The 1979 rice riots, a civil-military flashpoint in Liberia, also give this scenario historical precedence.¹¹⁹

Respondents who were randomly assigned to the control group did not hear any scenario. Because the treatment condition included a follow-up question that I used to construct a measure of prioritizing cohesion (described below), respondents in the control group received a similarly structured question about an unrelated topic. They were asked to choose the statement closest to their own opinion about AFL participation in international peacekeeping:¹²⁰ “International peacekeeping should be a primary mission for the AFL”; “It is okay to contribute to international peacekeeping occasionally, but not as a primary mission”; or “It does not matter, as long as the military executes the mission successfully.”¹²¹

After hearing the scenario (or control prompt), respondents were asked a series of outcome measure questions. First, I tested respondents' willingness to prioritize human rights. Respondents were asked whether they agreed with the following statement on a scale from 1 (“not at all”) to 4 (“a lot”): “The military should follow an approach that prioritizes the security of the people over the security of the government.”¹²² Second, I tested respondents' support for democracy and its alternatives, such as military rule, which represents a rejection of liberal civilian authority. These questions used standardized language from Afrobarometer public opinion surveys.¹²³ Respondents were asked whether they would disapprove or approve of the following alternatives: “Only one political party is allowed to stand for election and hold office”; “The army comes in to govern the country”; and “Elections and the House of Representatives are abolished so that the president can decide everything.” For each alternative, respondents expressed disapproval or approval on a scale from 1 (“strongly disapprove”) to 5 (“strongly approve”).

Respondents were then asked to choose the statement closest to their own opinion from three options: “Democracy is preferable to any other kind of government”; “In some circumstances, a nondemocratic government can be preferable”; and “For someone like me, it doesn't matter what kind of government we have.” Because these response options are categorical, the results that follow used a dummy variable that I coded as 1 if the respondent chose the first option and 0 otherwise.

To evaluate whether respondents prioritized cohesion after hearing the scenario, I coded a dummy variable that took a value of 1 if the respondent chose the third response option to the treatment question (“It does not matter, as long as the military follows the commander's orders and sticks together”) or control question (“It does not matter, as long as the military executes the mission successfully”). These options were designed to capture a rough measure of prioritizing cohesion in each case. This measure cannot provide conclusive evidence of causality because the questions were not worded identically. But it offers suggestive evidence for how norm conflict may affect preferences over cohesion. I also used additional tests to further probe the relationship between norm conflict and cohesion.

To evaluate how foreign military training affects soldiers' responses to norm conflict, I leveraged individual-level data on exposure to U.S. training. The U.S. program to rebuild the AFL began in 2006, with recruits entering the new force in eight groups called “batches” from 2006 through 2015.¹²⁴ The first batch of recruits were, as the instructors called them, the “guinea pigs.”¹²⁵ In this “proof of concept” class, 105 soldiers entered basic training in July 2006, and 102 graduated five months later.¹²⁶ The entire batch moved together through eleven weeks of Initial Entry Training, four weeks of Advanced Individual Training, and a four-week Basic Non-commissioned Officer Course (BNOC).¹²⁷ Eleven candidates with bachelor's degrees entered an additional six-week Officer Candidate School (OCS), forming the fledgling AFL officer corps. The second batch entered basic training in July 2007.

The period of U.S. training ran from 2006 to 2009 and was primarily executed by two firms, DynCorp International and Pacific Architects & Engineers.¹²⁸ The early training heavily emphasized liberal norms because the program designers believed “that after 14 years of civil war, most Liberians knew how to fire an AK-47 but did not know when or at what.”¹²⁹ To inculcate liberal norms, DynCorp invited civics trainers, Liberian academics, and other experts to conduct three weeks of intensive training on the laws of war, civilmilitary relations, and human rights that “dwarfed all other training.”¹³⁰ This three-week curriculum was the first casualty of funding shortfalls,¹³¹ but norms training continued alongside technical training from basic training through OCS.¹³² As the AFL soldiers went through basic training, their U.S. instructors “kept beating it into [their] heads that the military's role was to serve the population,” subject to civilian control.¹³³ This emphasis on liberal norms was “surprising”; there was “a lot of new doctrine being preached.”¹³⁴

Importantly, exposure to U.S. training varied across the recruitment batches, which allowed me to test for heterogeneous treatment effects. Batches 1–3 received the full U.S. training sequence, the middle batches (4–5) received varying amounts of training,¹³⁵ and the final three batches received no U.S. training.¹³⁶ By 2009, all training responsibilities had transitioned to a Liberian lead; DynCorp only observed the training for batch 6.¹³⁷ Four years later, the AFL managed its own recruitment and basic training for batch 7, followed by the final batch in 2015.

The amount and quality of training that the Liberian-led batches received declined after batch 5, owing to the lack of Liberian resources or capacity to conduct training on their own. Because the Liberians were not prepared to field their own BNOC or OCS classes, the training sequence was truncated and batches 6–8 failed to produce more than a handful of noncommissioned officers and officers. In 2017, a U.S. military training team went to Liberia to help design new OCS and BNOC courses. They found only one working computer on the training base and no sign of prior curricula.¹³⁸ The emphasis on norms, which had been rooted in the trainers' beliefs about the causes and effects of Liberia's war, also faded as the U.S. instructors cycled out of the training sequence.

For the tests that follow, I operationalized U.S. training by collapsing the “batch” variable into a dummy “training” variable that I coded as 1 if the respondent joined the AFL in batches 1–5 (any U.S. training) and 0 otherwise.¹³⁹

This section presents results from empirical tests of my argument about the effects of liberal norm conflict on military decision-making, along with alternative arguments about socialization and the mechanisms that connect norm conflict with prioritization of cohesion. First, I conduct difference-in-means tests across the experimental groups. These tests compare average responses across the treatment and control groups, which allows me to evaluate the effects of respondents' hearing the norm conflict scenario. Because my argument and U.S. policy expectations make opposing predictions, the hypothesis tests are two-tailed. Table 1 reports the results.¹⁴⁰

The results show that respondents who were exposed to norm conflict—the treatment—are less likely to support prioritizing human rights. The scenario about the government ordering the army to repress protests corresponds with a nearly 0.25 unit decrease in a four-point scale for prioritizing human rights. This finding is consistent with H1, which predicts that moments of norm conflict will reduce soldiers' willingness to prioritize human rights over civilian control. This finding contradicts the U.S. policy expectation laid out in H4, which is that soldiers ought to be more willing to prioritize human rights. Figure 2 graphs the distribution of support for prioritizing human rights across experimental conditions.¹⁴¹

The results in table 1 also suggest that exposure to norm conflict drives soldiers to prioritize cohesion. The likelihood that respondents prioritize cohesion doubles from no norm conflict (control) to norm conflict (treatment). This finding provides additional support for my argument, consistent with H3, which predicts that moments of norm conflict will lead soldiers to prioritize cohesion. Because of how this measure is constructed, however, these results must be interpreted with caution. Additional tests, described below, provide further evidence that a connection between liberal norm conflict and prioritization of cohesion exists.

The other variables, which measure support for democratic norms, are not significantly different across treatment and control groups. I argue that norm conflict will weaken support for democratic norms (H2), while U.S. policy expectations predict that conflict will sharpen support for them (H5). The signs of the baseline results are generally in the expected direction—respondents exposed to norm conflict are slightly more likely to support army rule and one-person rule, and slightly less likely to always prefer democracy—but none of these effects are statistically significant. In short, the results do not clearly support either hypothesis.

One interpretation of these null results is that while moments of norm conflict produce immediate decisions over which norms to prioritize, norm support erodes more slowly over time.¹⁴² Another possibility is that subsequent acts of norm violation weaken norm support, as soldiers attempt to minimize cognitive dissonance caused by divergence between beliefs and behavior. If prioritizing cohesion leads soldiers to defy civilian orders, for example, then we might expect to see a sharper drop in support for democratic norms. Because the survey only measures shifts in support for norms in the immediate moment of norm conflict, assessing the effects of subsequent behavior on attitudes remains an important question for future investigation.

Overall, these initial results provide mixed support for my argument. These baseline results suggest that norm conflict reduces soldiers' willingness to prioritize human rights and increases their inclination to prioritize cohesion, although the effects on support for democratic norms are muted. Importantly, these shifts in support for norms undercut two common policy assumptions. First, liberal training policy hinges on the expectation that soldiers will prioritize human rights when faced with norm conflict; if anything, the opposite is true. Second, liberal policy ignores the possibility that soldiers will prioritize cohesion, but the results suggest that soldiers may choose this norm instead— creating the conditions under which liberal norm violations can occur. These initial comparisons, however, mask the variation across subgroups that I explore next.

To examine how U.S. training affects responses to norm conflict, I compare differences in means across treatment and control groups conditioned on level of prior training. I use the “training” dummy variable that takes a value of 1 if the respondent was in an AFL recruitment group that had U.S. training, and 0 otherwise. Figure 3 presents the results.

Overall, respondents with U.S. training were significantly less likely to express willingness to prioritize human rights after hearing the norm conflict scenario. They were somewhat less likely to express absolute support for democracy and somewhat more likely to express support for army rule, although these results are not statistically distinguishable from zero. In contrast, respondents without U.S. training were significantly less likely to express support for one-party rule and somewhat less likely to support one-person rule.¹⁴³ Finally, U.S.-trained respondents were significantly more likely to emphasize cohesion when exposed to norm conflict. Even though not all of the differences are statistically significant, the pattern of coefficients is consistent across measures and provides support for H6a, which predicts that U.S. training will make respondents less willing to prioritize liberal norms.

It is possible that these results are an artifact of small sample size because there were only forty-nine respondents in the “no U.S. training” category.¹⁴⁴ More likely, however, is that these results reflect growing identification with the military institution associated with training and time in service. I hypothesize that people who have been in the military longer, receiving more training and socialization, perceive a higher need to protect the organization and become more likely to prioritize cohesion. If this is the case, then, the apparent willingness to prioritize human rights among the least-trained soldiers might reflect less loyalty to the military institution—a quality that would, paradoxically, make them worse soldiers. It is also worth considering that militaries are hierarchical institutions: leaders give orders and the rank and file obey. In the Liberian case, respondents with more U.S. training are also more likely to have command roles. In a crisis, these respondents would be the decision-makers issuing orders. Their unwillingness to prioritize human rights or support liberal norms when exposed to the norm conflict scenario has troubling implications for behavioral outcomes in a real-world crisis.

Next, I conduct tests to probe the mechanisms and rule out alternative explanations, including that socialization never happened and that time in service alone explains attitudes toward norms. First, respondents in the survey sample with the most U.S. training are also the soldiers with the most time in service; in other words, U.S. training is collinear to tenure in the military. Rather than U.S. training conditioning the effects of norm conflict, it is possible that tenure in the military conditions attitudes toward norms. If that is the case, then the findings may reflect the failure of training to impart norms in the first place, in line with prevailing arguments about the causes of norm violations among U.S.-trained forces.¹⁴⁵

To test this alternative explanation, I evaluate the relationship between U.S. training and support for liberal norms in the absence of norm conflict. I limited the survey sample to respondents who were randomly assigned to the control group and did not hear the treatment scenario (n = 144). I then estimated models that regress the outcome measures on U.S. training, using the “batch” variable that ranges from batch 1 to batch 8. For ease of interpreting the results, I flipped the values of the variable so that higher values indicate earlier batches that received more training.

Each model includes three control variables. The first variable, “education,” is included because it could affect both support for liberal norms and exposure to training. For example, soldiers must meet certain English language proficiency standards to qualify for training in the United States. Education also determines eligibility for the officer corps, in which soldiers receive additional training. The second variable, “wealth,” is an index variable of personal assets commonly used to proxy for wealth.¹⁴⁶ As with education, wealth could affect attitudes toward norms as well as the types of education and access to employment opportunities. The third variable is “rank,” a dummy variable that takes a value of 1 if the respondent is in the officer corps. Officers receive different training opportunities and often come from different social backgrounds than the rank and file, which could similarly affect attitudes.

Table 2 presents the results of ordinary least squares regressions; for the democracy model, because of the binary outcome measure, I use a linear probability model with robust standard errors.¹⁴⁷ The results show that soldiers with more U.S. training were significantly more likely to express support for liberal norms than soldiers with less U.S. training. Soldiers who received full training (batch 1) were much more likely to express willingness to prioritize human rights over regime security than those who received no training (batch 8). Yet these are the same soldiers who expressed less support for norms in the presence of norm conflict. These results lend support for rejecting two alternative explanations: that socialization never happened in the first place, and that tenure in the military alone shapes attitudes toward norms (at least not in ways that are inherently antithetical to liberal norms).

One issue that complicates the ability to interpret these results is that assignments to U.S. training were not random.¹⁴⁸ Consequently, age or time effects could also explain variation in attitudes across batches. For example, respon-dents who joined in batch 1 would have been closer to the war chronologically and thus more affected by it. Age could have a similar effect, with older respondents' preferences shaped by longer exposure to conflict. Two factors help to mitigate these concerns. First, because of the timeline of the U.S. program, batches 1–6 were recruited and trained within a narrow window (2006–2009). The largest gap between recruitment and training was between batch 6 and batch 7 (2009–2013). Second, the ages of AFL soldiers are relatively homogenous. Although 60 percent of Liberians are less than twenty-five years old, the median AFL soldier is thirty-seven years old. The AFL has struggled to recruit younger people who can meet its literacy requirements, given the mass disruption of education during the war.¹⁴⁹ As a result, 87 percent of the survey sample was over the age of thirty, reducing concerns over age effects.¹⁵⁰

Another potential inferential issue is that people who joined in earlier batches may have already left the military, leaving behind those most committed to military service. This selection bias is unlikely for two reasons. First, the AFL lacks formal processes for separating from the force; soldiers' initial contracts were automatically renewed, which kept people in who otherwise might have left.¹⁵¹ Second, the distribution of batches across the survey sample tracks closely to the initial distribution of batches at the force level based on batch graduation figures, which suggests that attrition did not disproportionately affect earlier batches.¹⁵²

Finally, I consider other alternative arguments. In particular, it could be that U.S. training emphasized civilian control, command and control, or cohesion over human rights, which would explain the decreased support for prioritizing human rights among earlier cohorts. Although the survey experiment does not allow for testing these alternatives directly, qualitative evidence suggests that U.S. training did not emphasize other norms over human rights.

The training program designers regarded the two liberal norms of civilian control and respect for human rights to be mutually reinforcing, but in practice they emphasized human rights because they perceived it as the greater problem in the aftermath of war.¹⁵³ Interviews with AFL officers likewise revealed an independently minded military that was not unduly subordinate to civilian control. Liberian officers expressed some contempt for what they viewed to be inappropriate political efforts to control the military. As one officer put it, civilians “need to go to school” to get a full understanding of what civilian control means.¹⁵⁴ “They don't want the AFL to be too powerful,” he said, “but we control the weapons.” The trainers also did not emphasize blind obedience or command and control in the training. In the words of one of the drill instructors, “If jumping off that cliff was going to kill you, but somebody gave you an order to jump off that cliff, what are you going to do?”¹⁵⁵

Interviews with program designers and trainers did not reveal an emphasis on cohesion over liberal norms. The program designers assumed that cohesion would develop organically if they recruited from an ethnically diverse cross section of the population. Their main strategy for fostering cohesion was to conduct statewide recruitment to ensure a representative force.¹⁵⁶ For their part, the U.S. military trainers did not teach cohesion in the same way as liberal norms. Instead, they fostered teamwork and solidarity, and they emphasized that a shared military identity should take priority over ethnic or tribal identities.¹⁵⁷ In the aftermath of a civil war fought along tribal lines, cohesion was initially weak in the nascent AFL. Together, the U.S. approach and Liberia's history of civil war help explain why AFL soldiers would be motivated to prioritize cohesion under pressure despite intensive liberal norms training.

Finally, I conduct two additional tests to explore the relationship between norm conflict and prioritization of cohesion. While each test individually represents a plausibility probe of the argument, their cumulative effect helps to increase confidence in the overall findings.

First, I examine the conditional effects of treatment, given soldiers' prior perceptions of unit-level cohesion. Before exposure to treatment, all respondents were asked a series of questions to probe their perceptions of cohesion and belonging in their units. Drawing on the military sociology literature on cohesion,¹⁵⁸ respondents were asked whether they agreed with three statements: “Everyone in this unit works together to achieve our missions”; “If this unit were in combat, any soldier would be willing to risk his life to help another”; and “Soldiers in this unit treat each other equally regardless of their tribe or religion.” The first two statements differentiate between trust in shared competencies and common goals (task cohesion) and interpersonal bonds (social cohesion),¹⁵⁹ and the third statement captures perceptions of ethnic divisions or exclusionary practices at the unit level (equality),¹⁶⁰ which is also known to affect the quality of cohesion. Answers ranged from 1 (“strongly agree”) to 5 (“strongly disagree”).

Figure 4 displays the conditional treatment effects of unit-level cohesion.¹⁶¹ The solid lines graph the effects of assignment to treatment or control conditioned on prior perceptions of cohesion; the shaded areas indicate the 95 percent confidence intervals. The results show a striking pattern: Soldiers who perceived weak task or social cohesion in their units (i.e., who strongly disagreed that their units were cohesive) were significantly more likely to select the third response option (“It does not matter, as long as the military follows the commander's orders and sticks together”) when exposed to the norm conflict scenario. This finding suggests that soldiers who are worried about cohesion in the first place are more likely to interpret norm conflicts as posing a threat to cohesion. Surprisingly, however, the effect inverts when the question concerns equality. Soldiers who perceived inequitable treatment in their units were significantly less likely to select the third response option; soldiers who felt excluded at the unit level may become alienated from the military organization and thus less vested in its survival.

As a final test, I examine how exposure to the norm conflict scenario shaped respondents' thinking about military priorities. After the outcome measures, all respondents were asked the following open-ended question: “What, if anything, does the AFL motto ‘A force for good’ mean to you?” I coded these open-ended responses to create a dummy variable that takes a value of 1 if the response used key words or phrases that invoked cohesion.¹⁶² Respondents in the treatment group were more likely to give answers that highlighted cohesion. Table A13 in the online appendix shows that the difference is significant at p < 0.05. This finding provides further evidence that liberal norm conflict heightens soldiers' prioritization of cohesion.

This article has demonstrated that conflicts can arise between the liberal norms of respect for human rights and civilian control of the military. Liberal powers use military training around the world to inculcate these two norms, yet policymakers have neither confronted the reality of this dilemma nor seriously considered its effects. The implicit U.S. policy expectation is that if norm conflict arises, then well-trained, liberally oriented militaries will prioritize human rights over civilian control. Such a choice should be easy to make and benign in its effect on the norms in question. Using experimental data from the Liberian armed forces, I have presented preliminary evidence that challenges this sanguine perspective. First, in contrast to U.S. policy expectations, norm conflict tends to reduce soldiersy willingness to prioritize human rights. Second, norm conflict seems to increase soldiers' prioritization of cohesion. Third, insomuch as U.S. training shapes responses to norm conflict, it makes these effects stronger. These results shed light on a pathway by which liberal norms lose their influence on decision-making relative to other norms aligned to soldiers' interests (e.g., cohesion), creating opportunities for normviolating behavior.

Using survey experiments solves some inferential problems, such as internal validity, but it also introduces others, such as external validity.¹⁶³ Random assignment to treatment or control conditions is the main way to address internal validity because it helps to establish that other unobserved variables are not driving results. External validity, on the other hand, concerns the generalizability of results. One key concern about external validity is whether the sample accurately represents the population of interest.¹⁶⁴ Although random selection ensures that the sample accurately reflects the AFL population, it is important to note that the AFL is unique in two ways. First, it was rebuilt from scratch, with few members having previous military experience. Second, the timing of norm transmission occurred in tandem with basic training. Foreign military training likely affects individuals differently when they come from intact militaries with stronger preexisting norms, and when norms training occurs after early, formative socialization experiences.¹⁶⁵ These factors do not necessarily affect the dynamics of norm conflict, but they could make it harder for norms to stick in the first place, diluting the effects of training.

Questions of external validity point to a limitation of survey experiments, which can only be conducted in one context at a time. The solution to this limitation is replication and extension of the experiment across time and space.¹⁶⁶ To establish generalizability, future research should explore how different structural and environmental conditions affect how soldiers respond to norm conflict.¹⁶⁷ Although Liberia is characterized by weak rule of law, it nonetheless aspires to keep the military out of domestic law enforcement. Future work could look at the problem of norm conflict in states where the military has an internal security mandate; it is also worth replicating in contexts with strong rule of law.¹⁶⁸ Additional research could also examine other factors besides foreign training that affect cohesion, such as comparing the effects of norm conflict in states without civil war legacies, or where regime policies promote or undercut cohesion.

How long do the effects of liberal norm conflict persist? Do norms bounce back after the crisis passes, or does norm conflict contribute to permanent erosion of these liberal norms? Connecting attitudinal shifts to behavioral outcomes is an important next step for research. The evidence presented here sheds light on the micro-level dynamics that explain shifts in military support for norms; this experimental work can complement and inform other approaches that link attitudes to behavior.

When the United States, like other liberal powers, builds military capacity in weak states, it uses training to promote liberal norms to solve the dilemma of how to increase capacity while maintaining civilian control and protecting human rights. The policy implication of my findings is that training militaries might encourage norm-violating behavior. This does not mean that U.S. training is worse than other training—military training may generally increase soldiers' prioritization of cohesion under pressure and drive norm violations via other mechanisms.¹⁶⁹ But these findings do suggest that U.S. training can backfire in ways that are unique to liberal security assistance; they also suggest that the United States puts false confidence in the power of norms to restrain the militaries that it trains. More training is not the solution. Instead, liberal providers should emphasize building institutions that help to regulate military behavior rather than prioritizing individual or unit-level training with a normative component tacked on.

Liberal providers could also do more to clarify norm hierarchies. By promulgating clear guidelines for behavior and blueprints for decision-making (e.g., “people first, governments second”), some of the problems associated with norm conflict may be mitigated. But doing so requires two policy changes. First, policymakers must admit that this rank-ordering exists. Second, they must modify training curricula to address this hierarchy of norms. The United States might not always want foreign militaries to side with the population, particularly if it means defying a friendly regime. This is an inescapable tradeoff.

Finally, it is worth considering whether the problem of liberal norm conflict gives some providers an advantage over others when it comes to shaping military behavior. China, for example, is a major training provider in Africa. China does not emphasize human rights in its training; instead, it puts a strong and singular value on the norm of civilian (party) control.¹⁷⁰ Future research should explore whether there is an “authoritarian advantage” when it comes to foreign military training. The United States prides itself on promoting values in and through security cooperation, a feature that it views as a comparative global strength. Protecting those values means acknowledging their limits and dilemmas.

The author is grateful to Brian Blankenship, Risa Brooks, Jonathan Caverley, Michael Desch, Virginia Page Fortna, Robert Jervis, Richard McAlexander, Barry Posen, Robert Ralston, Laura Resnick Samotin, Jack Snyder, Monica Duffy Toft, and the anonymous reviewers for helpful comments. She also thanks participants in seminars at Columbia University, Yale University, George Washington University, University of Notre Dame, Massachusetts Institute of Technology, Ohio State University, and Cato Institute, as well as participants at the 2018 American Political Science Association and 2019 International Studies Association conferences. An online appendix is available at https://doi.org/10.7910/DVN/OA2MTX.