Rebels against Mines? Legitimacy and Restraint on Landmine Use in the Philippines

Abstract Rebels have become the most prolific users of landmines but still display significant variation in how they employ and restrict the weapon’s use. This article argues that how rebels exercise restraint on landmine use depends on which audiences they rely on most. In a comparative case study of three Philippine rebel groups—the Moro Islamic Liberation Front, the Abu Sayyaf Group, and the New People’s Army—this article highlights three main findings. First, rebels reliant on voluntary compliance from local communities are more likely to limit the effects of landmines on their perceived constituency. Second, when rebels have conciliatory relations with the government, they are more likely to comply with national law, reciprocate government behavior, and limit the effects of landmines on the government’s constituents. Finally, rebels seeking legitimacy from human-rights-conscious foreign sponsors are more likely to comply with international law related to landmine use.

one-sided violence, but both used landmines extensively. 3 The Lord's Resistance Army in Uganda, a heterodox Christian armed group, displayed very high levels of one-sided violence but used landmines only sporadically. 4 Although many communist and Islamic groups have used landmines, these ideologies have also provided the moral grounds for restraint. The Taliban, for instance, officially considered the use of landmines "an un-Islamic and anti-human act," which "would be punished in accordance with Islamic Law. " 5 Likewise, the Revolutionary Workers' Party of the Philippines renounced the use of landmines based on socialist principles. 6 Why do some nonstate armed groups use landmines while others do not? The few studies on the topic focus on codified restraint norms, namely the antipersonnel mine ban. 7 Put simply, this work poses that rebels commit to, and to some extent comply with, these norms to gain legitimacy and avoid reputation costs from larger domestic and international audiences. 8 Furthermore, these studies employ a cross-national quantitative approach, favoring larger datasets over fined-grained qualitative data. This approach has several limitations. First, focusing on the ban does not capture the use of anti-vehicle mines and command-detonated explosive devices, which are often used in conjunction with or as alternatives to antipersonnel mines. Second, it ignores the restraint behavior of groups that do not abide by the antipersonnel mine ban. Third, the large-N approach, though useful to identify broader trends in rebel behavior, risks missing the nuance of whether and how rebels actually exercise restraint on the ground. Fourth, it does not account for local communities' agency and role in influencing rebel restraint.
I address these shortcomings in three steps. First, I examine the drivers of restraint on landmine use instead of commitment and compliance to specific norms. For this, I build on key insights from the growing literature Sjöberg, Armed Non-state Actors and Landmines, 1:59-64. 5 Islamic Emirate of Afghanistan, "Statement of the Islamic Emirate of Afghanistan on the Problem of Landmines, " 6 October 1998, 1-2, http://theirwords.org/media/transfer/doc/ut_af_taliban_1998_01-d631e 9853b6909a692f10408438eda18.pdf. on restraint in civil wars and rebel governance. 9 In line with this scholarship, I argue that not only are legitimacy-seeking rebels more likely to exercise restraint but also that how restraint gets exercised depends on which audiences rebel groups rely on most. This theoretical framework considers how local, national, and international audiences influence rebel behavior.
Second, I conceptualize how restraint on landmine use varies across a spectrum, irrespective of commitment to or compliance with specific treaties. Instead, I draw on the practical application of broader humanitarian norms, namely the principles of distinction, proportionality, and precaution. This allows moving beyond the prevailing dichotomous operationalization and captures a broader variation in restraint behavior, including "in-between" cases. Further, my conceptualization is also valid for periods before the codification of norms regarding landmine use.
Third, I favor a qualitative research design to allow for a more nuancedthough still parsimonious-account of rebel behavior. Specifically, I document variation in legitimacy-seeking behavior and restraint on landmine use by three nonstate armed groups in the Philippines: the Moro Islamic Liberation Front (MILF), the Abu Sayyaf Group (ASG), and the New People's Army (NPA). This approach complements previous quantitative studies, allowing for a better understanding of variation in restraint behavior and a more precise and nuanced analysis of dynamics on the ground, across both cases and time.
This article's findings support its central theoretical argument. First, rebels reliant on voluntary compliance from local communities are more likely to limit the effects of landmines on their perceived constituency. Second, when rebel groups have conciliatory relations with the national government, they are more likely to comply with national law, reciprocate government behavior, and limit the effects of landmines on the government's constituents. Finally, rebels seeking legitimacy from human-rightsconscious foreign sponsors are more likely to comply with international law related to landmine use.
This article proceeds in six sections. I first define restraint and how it relates to the concepts of violence, commitment, compliance, and governance. I then briefly review previous research on violence and restraint in war and identify gaps in explaining restraint on landmine use. Next, I argue for legitimacy as a primary driver of restraint, specifying how relationships with local, national, and international audiences may shape rebel restraint on landmine use. Then, I present the chosen research design, case selection strategy, and data collection. The ensuing section presents and compares legitimacy-seeking behavior and restraint on landmine use between MILF, ASG, and NPA. The final section offers some concluding remarks and avenues for future research.

Defining Restraint
Previous literature has taken different approaches to understanding variation in nonstate armed groups' use of violence in general, or of specific means and methods of war in particular, and their behavior toward civilians. Perhaps most prominently, many civil war scholars have tried to explain violence through instrumental 10 and noninstrumental theories. 11 Some have grounded their research in codified humanitarian norms and investigated the drivers of commitment to these norms, such as the Geneva Conventions. 12 Following naturally from this vein of research, scholars have moved beyond asking why rebels "talk the talk" to why they "walk the walk"-that is, what their reasons are for complying with humanitarian norms. 13 Still, the literature on rebel governance has focused on rebelcivilian relations and their possible consequences, from punishment and coercion 14 to security provision. 15 Finally, a growing body of research explores restraint behavior, usually understood as instances of limited or absent violence. 16 A series of conceptual overlaps stems from these strands of literature. Violence and restraint can be seen as opposite sides of the same phenomenon only to the extent that they share a logic. For example, Stathis N. Kalyvas's theory of selective violence 17 can, simultaneously, explain why rebel groups may exercise indiscriminate violence, selective violence, and restraint, depending on their level of territorial control vis-à-vis the government. However, drivers of restraint may also follow an independent logic and often present a barrier to the exercise of violence. 18  Stanton, for instance, shows that violence coexists with restraint, as the latter moderates specific aspects of the former, such as the strategy and intensity of civilian targeting. 19 Further, though commitment can be understood as formally adhering to a specific set of rules and standards, compliance entails following through with this commitment. Thus, commitment to the laws of war, for example, may signal the intention to exercise increased restraint, the codification of already-existing informal restraint norms, or only rhetoric. Noncompliance, however, does not necessarily mean the commitment was spurious. For instance, a group lacking effective command and control structures might commit to an agreement but fail to ensure compliance from the lower rank and file. The concept of restraint overlaps with compliance to the extent that commitments call for limited violence. For example, if an armed group commits to not employing sexual violence, following through with this commitment entails both compliance and restraint. However, depending on what the commitment entails, compliance may go beyond restraint and even enter the sphere of governance, such as punishing perpetrators of sexual violence and ensuring medical and psychological assistance for survivors. Likewise, restraint may be exercised without compliance with any codified norm. For instance, if a group is physically attacked, it can respond nonviolently despite its right to selfdefense. Building on these different strands of literature, this article focuses on restraint behavior. I define restraint as the self-imposed limitation on the use of violence, despite the existence or the possibility of pursuing such a capability. This definition is agnostic to whether there is an original intention to use violence. In other words, restraint can be exercised as the product of a purely rational cost-benefit calculation, where an initial intention to use violence is disregarded due to its disproportional costs. Alternatively, organizational cultures may embed norms of restraint so that armed groups do not even consider using specific means or methods of war.

Drivers of Restraint in Civil Wars
Previous research suggests strategic and normative incentives for restraint in war. According to the strategic argument, restraint is understood as the product of a rational cost-benefit calculation, "a trade-off between military considerations and a motivation to comply [ benefits of compliance and noncompliance with international law. 21 On the one hand, compliance costs range from ensuring discipline and command and control 22 to potential military disadvantages of forgoing specific means or methods of war. The benefits of noncompliance include potential military advantages, 23 political leverage, 24 civilian control, 25 and access to information. 26 These benefits are particularly relevant in asymmetric conflicts. 27 On the other hand, noncompliance risks uncontrolled escalation, 28 third-party interventions, 29 legal liabilities, 30 and reputation loss. 31 Benefits of compliance include enhanced command and control, 32 reciprocity in treatment by opponents, 33 and reputation gains. 34 Other factors-such as democracy, conflict dynamics, and geographymay tip this cost-benefit balance. For example, Stanton suggests that rebel groups may be more likely to target civilians when fighting democracies, which arguably have less tolerance for civilian casualties. 35 Moreover, rebels may both victimize civilians and exercise restraint due to competition with rival factions. 36 Finally, the utility of specific means and methods of war may increase due to geographic factors, such as terrain ruggedness and vegetation cover, 37 or to ensure the survival of armed groups facing battlefield losses. 38 Normative arguments emphasize how beliefs, values, expectations, and practices define and rank individual and collective preferences. 39  addition to a rational decision following a cost-benefit calculation, restraint can be perceived as a benefit in itself or as the only reasonable alternative. In line with this argument, previous scholarship has focused on organizational factors, 40 military culture, 41 ideology, 42 socialization, 43 and sociopolitical dynamics 44 as drivers of violence and restraint. Existing legitimacy-and reputation-based theories integrate strategic and normative drivers of restraint. 45 Legitimacy, defined as the "support and recognition that a rebel group is a viable political authority," 46 can be considered both ends and means for a rebel group. Legitimacy is intrinsically connected to the group's local, national, and international goals, relationships with its constituencies, and several sources of support. It follows that rebel groups often change their behavior to conform with key audiences' expectations, 47 either due to a rational cost-benefit calculation or genuine alignment with those audiences' core interests. For instance, Stanton finds that restraint is more likely when leaders are responsive to public demands, such as in inclusive political systems or unstable regimes requiring domestic and international support. 48 Authors have written extensively on nonstate armed groups' use of landmines. However, this literature is often descriptive, context-specific, or normative, and it lacks coherent and verifiable theoretical frameworks, hardly enabling cross-case comparisons. 49 The few exceptions follow the legitimacy-based approach. Sven Richters suggests that legitimacy-seeking rebel groups are more likely to commit to the antipersonnel mine ban to  appeal to international audiences. 50 Fazal and Margarita Konaev advance this argument and integrate military-strategic considerations and rebel political aims. 51 In a recent study, Sarah Littisha Jansen highlights how nonstate armed groups may commit to the ban to convey good faith ahead of negotiated settlements. 52 Kristian S. Gleditsch et al. focus on how the interaction between nonstate armed groups and governments influences the former's commitment to and compliance with the ban. 53 Although these studies constitute the backbone of this article, they contain several limitations. First, because they focus on the antipersonnel mine ban only, by design, they do not capture the use of anti-vehicle mines and command-detonated explosive devices. 54 Such devices are often used in conjunction with, or as alternatives to, antipersonnel mines and may also pose significant harm to civilians during and after conflicts. 55 Second, they operationalize commitment and compliance as dichotomous variables, thus missing "in-between cases," such as groups that did not abide by the antipersonnel mine ban but still attempted to limit the effects of landmines. Third, and relatedly, large-N studies favor generalizability over nuance, which risks missing practical details of whether and how rebels exercise restraint on the ground. Finally, these studies focus on larger domestic and international audiences and thus pay scant attention to the relationship between rebels and local communities.

Legitimacy as a Driver of Restraint
As a starting point, I concur with the argument that the relationships rebels have or aspire to have with various audiences drive restraint. 56 My contribution outlines how restraint on landmine use varies depending on each type of audience, as rebels attempt to mirror the audience's preferences. Reliance on local audiences should lead to a greater focus on civilian protection; stronger relationships with governments emphasize compliance with national law, reciprocity, and protection of the government's constituency; and human-rights-conscious international audiences should encourage compliance with international law.
As with other forms of violence, restraint on landmine use carries different costs and benefits. I focus on those directly linked with rebels' pursuit of legitimacy from different audiences. Common to all audiences is the assumption that expected or actual civilian victimization caused by landmines should lead to legitimacy costs. Legitimacy costs lead to a perceived decrease in the utility of landmines, which, in turn, compels rebels to exercise restraint on landmine use. Not only should restraint be higher for legitimacy-seeking rebels, but it should also qualitatively vary in terms of choice of means and methods of war, the intended target, avoidance or acceptance of collateral damage, and whether precautionary measures are established. This variation, I argue, depends on the audiences from which rebels seek legitimacy. Sukanya Podder identifies various sources of rebel legitimacy, which she clusters at three levels: "civilian communities, the state or regime in power, and external actors, including regional and international sponsors. " 57 First, local support has long been considered a central determinant of civil war outcomes, as communities can provide or deny material and nonmaterial support to the warring parties. 58 Accordingly, previous studies have found that dependence on voluntary support has driven lower levels of civilian victimization 59 and led to active protection of civilians. 60 Further, Luis De la Calle argues that weaker rebels avoid victimizing civilians to gain civilian support. 61 Rebel groups may also be accountable to traditional or religious leaders from whom they seek guidance. 62 Thus, rebels seeking legitimacy from local communities, whether because the rebels are interested in material and nonmaterial support or due to ideological reasons, should avoid civilian victimization and exercise restraint on landmine use. However, not all civilians exert the same influence on rebel groups. I expect rebels to exercise restraint toward the communities from which they seek legitimacy-that is, their perceived constituents-but not necessarily toward perceived nonconstituents, such as those with a differing political affiliation, ethnicity, religion, or social class. This does not mean that rebels seeking legitimacy from one community are more likely to target other communities or that rebel groups that do not rely on local communities are necessarily more likely to target civilians in general.
To be clear, constituents may behave differently toward nonconstituent civilian victimization. On the one hand, civilians may support and promote violence against out-groups for several reasons, such as genuine support of a radical political agenda, fear of persecution, and expected material and nonmaterial benefits such as protection, goods, and services. 63 On the other hand, if landmine effects are perceived as indiscriminate or immoral, victimizing nonconstituent civilians via landmine use may broadly alienate local supporters. For instance, constituents may fear falling victim to landmines themselves, especially if they live close to nonconstituent targets. Alternatively, they may fear indiscriminate reprisals or simply object to the use of such tactics on moral grounds. 64 Nevertheless, several armed groups have claimed to use landmines discriminately and precisely to protect their constituents from government forces. 65 Thus, I anticipate that rebels avoid collateral damage to their constituency by prioritizing the use of command-detonated devices in unpopulated areas and against either enemy combatants or nonconstituent civilians. I also expect such rebel groups to mark, fence, and control minefields, and inform their constituents of the location of hazardous areas or issue warnings before attacks.
Hypothesis 1 (H1) Rebels seeking legitimacy from local communities are more likely to limit the effects of landmines on their perceived constituency than rebels who do not depend on local legitimacy among a civilian population.
Second, rebel groups often need support from national audiences to achieve their political goals. Relevant national stakeholders include the state government, political parties, civil society, and the private sector. Rebels' perceived legitimacy allows them to engage in political negotiations, establish political alliances, and enhance their public image. 66 Nationallevel legitimacy should also encourage reciprocal relations between the rebel group and the opposing government. For instance, Fazal finds that secessionist groups are more likely to target civilians when government forces do the same. 67 There may be at least two competing logics driving reciprocal restraint behavior in relation to civilian targeting. First, a logic of retaliation may encourage armed groups not to target the opponent's civilian constituency, fearing having their own constituency attacked as a consequence. Second, civilian targeting by one party may reduce the possible reputation costs for the other to do the same. 68 In sum, previous scholarship thus suggests that nonstate armed groups both rely on key national audiences and reciprocate opposing forces' behavior.
Among national-level audiences, this article focuses on rebels' relations with the state government. It may seem counterintuitive to think that rebels simultaneously challenge the state's sovereignty and seek legitimacy from the government. However, there are several reasons this can be true. Related to the reciprocity argument, rebel groups may seek recognition as a legitimate enemy worthy of respect and certain constraints in the use of violence. While some political goals rebels pursue, such as autonomy, are, in theory, obtainable by brute force, it is more realistic that achieving such goals requires some degree of recognition as legitimate political actors from the state government. Governments are more likely to negotiate with and concede to perceived legitimate actors than, for example, groups labeled as terrorists.
Further, Paul Staniland, in explaining state-nonstate armed group relationships, argues that "rather than straightforwardly worrying about the size, power, or organizational characteristics of armed groups, regimes assess how [nonstate armed] groups' politics align with the government's own goals." 69 How governments perceive rebel groups' political goals, the argument goes, influences whether states engage in strategies of confrontation, containment, limited cooperation, or alliance with nonstate armed groups. It makes sense, then, that rebel groups may want to convey the legitimacy of their political goals to the state government, at least in an attempt to influence the regime's strategies. These dynamics are also arguably more salient for weaker groups and immediately before and during political negotiations. 70 In particular, I expect nonstate armed groups that seek legitimacy from the state government to exercise restraint on landmine use along general lines of compliance with national legislation and reciprocity toward government forces. For example, in countries where antipersonnel landmines are prohibited and not used by the state's armed forces, legitimacy-seeking rebels should also avoid using these devices. Further, unarmed military troops, police personnel, civilian officials, and the government's core constituency should not be targeted. However, rebels might still target combatants and specific groups of civilians, such as divergent ethnic or religious groups, especially if the government also marginalizes those groups.
Hypothesis 2a (H2a) Rebels seeking legitimacy from the state government are more likely to comply with national law related to landmine use.
Hypothesis 2b (H2b) Rebels seeking legitimacy from the state government are more likely to reciprocate government behavior on landmine use.
Hypothesis 2c (H2c) Rebels seeking legitimacy from the state government are more likely to limit the effects of landmines on the government's constituents.
Finally, international audiences can be valuable material and nonmaterial support sources. Nonstate armed groups have strong incentives to seek legitimacy from foreign states, diasporas, international organizations, nongovernmental organizations (NGOs), and other nonstate armed groups. International legitimacy is particularly relevant for secessionist rebels seeking admission to the international community. 71 In this vein, Clifford Bob demonstrates how insurgencies tailor their strategies and tactics to meet foreign patrons' expectations. 72 Not all foreign supporters, however, value restraint in the same way. Global terrorist networks and foreign governments with records of human rights violations, for example, are less likely to value restraint in assessing their support and partnership with rebel groups. Conversely, democracies and states with strong human rights lobbies 73 and international NGOs 74 are more likely to support rebel groups that do not victimize civilians. It follows that groups seeking recognition from international audiences at large or specific human-rights-conscious foreign sponsors should also be responsive to civilian victimization and, thus, exercise restraint on landmine use. I expect landmine use to follow general aspects of international law, targeting only combatants and military objectives via anti-vehicle mines and command-detonated devices, not antipersonnel mines. I also expect rebel groups to avoid employing explosive devices in populated areas and to share information regarding landmine contamination following international standards.
Hypothesis 3 (H3) Rebels seeking legitimacy from human-rights-conscious foreign sponsors are more likely to comply with international law relevant to landmine use.
Rebels rarely seek legitimacy from only one audience, however, and it is more likely that rebel groups would appeal simultaneously to different audiences to maximize material and nonmaterial support. Though I assume that each hypothesis presented here should produce independent effects, their empirical implications may overlap. For example, the avoidance of using antipersonnel mines might arise from the pursuit of legitimacy from local communities, the state government (in case the government adheres to the Ottawa Convention), or human-rights-conscious foreign sponsors. Each type of audience's effect on restraint on landmine use should be more visible in how and to whom groups communicate their restraint behavior, and be more salient in cases when one audience is more relevant than others.
The suggested causal mechanism is inherently a normatively framed cost-benefit calculation. When rebels actively seek legitimacy from specific audiences, often for material and nonmaterial support, these audiences communicate their preferences. Rebels interpret these preferences and weigh them against other considerations, such as military or political circumstances. When the perceived costs of using landmines outweigh the benefits, rebels exercise restraint. Consequently, as preconditions for exercising restraint, nonstate armed groups should be able to lay landmines or pursue such a capability and be aware of their audiences' preferences. This mechanism is arguably more relevant for rebel leaders, who are more likely aware of the costs and benefits of restraint on landmine use and whose preferences are expected to be transmitted to the whole group.
Finally, the argument here assumes a linear logic of causation, in which legitimacy-seeking behavior is exogenous to the exercise of violence and restraint. This assumption is made for the sake of simplicity. In reality, the relationship between legitimacy-seeking behavior and the production of violence and restraint is inherently interactive. For example, rebels' appeal to specific audiences may change as rebel groups gain or lose access to material resources, adjust political goals, or suffer a severe military backlash. Mainly related to landmine use, rebels' assessment of the costs and benefits of employing landmines is likely an evolving process akin to a feedback loop of insurgents and their various audiences' actions and reactions.
In the following sections, I mitigate such endogeneity problems through a qualitative research design and careful analysis. Specifically, in the cases where there is a significant shift in legitimacy-seeking behavior, I present the empirical material as a series of snapshots linking a specific pattern of legitimacy-seeking behavior to a corresponding change in restraint on landmine use.

Methods
In contrast to previous, primarily quantitative, literature on this topic, I test my hypotheses through case studies. A qualitative approach is a better fit for investigating the relationship between rebel groups and different audiences, the details of landmine use, and the mechanisms linking legitimacy-seeking behavior and restraint.
I adopt a most similar system design 75 and test the suggested hypotheses on three nonstate armed groups from the Philippines: MILF (1984-), ASG (1986-), and NPA (1968-). While the case selection favors cross-case comparisons, it also allows for identifying and analyzing cross-time variation within the same group. At first glance, the cases seem to display significant variation in both legitimacy-seeking behavior and restraint on landmine use. MILF has tried to establish itself as an autonomous political authority within the Philippine state and has committed to forgoing antipersonnel landmines. NPA aims to overthrow the Philippine government and institute a people's republic; the group justifies using landmines against government forces. Lastly, ASG is a revisionist armed group with minimal domestic and international support that has repeatedly targeted civilians with various explosive devices. 76 Yet the cases are similar along several key dimensions that should influence how rebels perceive the utility of landmines. First, all three groups show low arms procurement capabilities, but they all used landmines and command-detonated devices to different degrees. Second, the groups have centralized authority and moderate levels of command and control, which makes the risk for inadvertent use of landmines and other explosive devices similar across groups. Third, they have low fighting capabilities vis-à-vis the Government of the Republic of the Philippines (GRP). 77 According to the "poor man's weapon" predication, they should have strong incentives to use landmines to overcome this power asymmetry. Fourth, the three groups rebel against GRP and operate in the same country and time, thus controlling for the opposing government and its level of democracy, geographical characteristics, the number of insurgent factions, and other country-level variables. Fifth, the Philippines has signed the Ottawa Convention and criminalized the use of antipersonnel landmines in its territory. Finally, humanitarian organizations, such as the Philippine Campaign to Ban Landmines, the International Committee of the Red Cross, and Geneva Call, have engaged all groups in promoting relevant humanitarian norms in the country. Thus, I expect the groups to be aware of applicable national and international norms related to landmine use and the liability costs stemming from noncompliance.
The rebel groups do differ in their political goals, ideology, level of territorial control, and size (and thus strength relative to each other), 78 which may influence both legitimacy-seeking behavior and the perceived utility of landmines.
Concerning legitimacy-seeking behavior, previous literature has shown how these variables may influence relationships between rebel groups and various audiences. For example, Fazal argues that rebels' political goals are directly linked to their relationship with local, domestic, and international audiences. Revolutionary groups seek to overthrow the government and thus often seek legitimacy from broader local-and national-level audiences. Whereas ethnonationalist groups, she argues, are more likely to appeal to local constituencies and co-ethnic diasporas, secessionist groups are more likely to appeal to international audiences for recognition. Revisionist rebel groups often reject the international community and major national-level audiences and instead appeal to a narrowly defined constituency. Finally, resource-seeking rebels have few incentives to seek legitimacy from major audiences but still depend on the cooperation of key civilians involved in resource extraction and trade. 79 Francisco Gutiérrez-Sanín and Elisabeth Jean Wood pose that ideology defines "the identification of a referent group (a class, ethnic, or other social group)" 80that is, an armed group's core constituents, political allies, and potential external supporters. Still, territorial control implies a continued interaction between rebels and civilians. Territorial control thus influences how rebel groups access information, maintain some degree of order, and even provide public goods and services. Alternatively, as De la Calle argues, weaker  Gutiérrez-Sanín and Wood, "Ideology in Civil War, " 215. rebels with a low level of territorial control are more likely to avoid "unqualified civilian killings" to maintain civilian support. 81 This article, however, does not attempt to explain why or how rebel groups interact with particular audiences but instead looks at the effects these interactions have on restraint. As such, rather than possible confounding variables, these factors are embedded in nonstate armed groups' ensuing legitimacy-seeking behavior.
Thus, I argue that the effects of political goals, ideology, territorial control, and group size on restraint on landmine use are mediated mainly by the group's legitimacy-seeking behavior. For example, if groups renounce using landmines on ideological grounds, they ultimately appeal to their constituency's ideals. However, groups controlling larger areas could favor using landmines for defensive purposes, and smaller groups might lack the resources to employ viable alternatives to landmines. Finally, while the three Philippine groups are significantly weaker than government forces, differences in relative strength and competition among the groups might lead some to prioritize the use of landmines. Therefore, I explore the influence of these factors on restraint on landmine use in greater detail in the analysis section.
The analysis is based on secondary sources, including armed group profiles, official government reports, NGO and think tank publications, scholarship, and journalistic accounts. Data on landmine use, however, is by default inexact. It is often uncertain who laid a landmine, when they laid it, what type of device detonated, and what remains underground. Therefore, in assessing intent, the analysis is complemented with primary data, namely datasets, unilateral declarations, ceasefires, and peace agreements. 82

Restraint on Landmine Use
Because restraint behavior implies the absence or limitation of violence, measuring it presents a challenge. 83 To overcome this counterfactual hurdle, scholars have measured restraint against different benchmarks, namely existing possibilities of inflicting violence, 84 comparable cases of unrestrained violence, 85 past unrestrained behavior, 86 different patterns of 81 De la Calle, "Compliance vs. Constraints, " 427. 82 Although this article primarily draws on English-language sources, literature in other languages, such as Spanish, French, and Portuguese, was also reviewed. The empirics mostly build on Filipino authors, who often publish in English. unrestrained violence, 87 and norms of acceptable behavior. 88 In this article, I follow this latter approach and measure restraint against humanitarian norms. This allows for a broader basis for cross-case comparison, independent from other internal and external contextual references, and befits the literature on compliance with and the promotion of humanitarian and human rights norms.
Humanitarian norms restrict the conduct of hostilities in both the means and methods armed groups employ. Some means are considered inherently inhumane and indiscriminate or cause such disproportionate humanitarian and environmental damage that their use should be restricted or completely banned. Even when lawful means of warfare are employed, combatants should exercise violence following the principles of distinction, proportionality, and precaution. In this sense, violence can only be directed at combatants and military objectives. Civilian victimization-"collateral damage"-should be proportional to the anticipated military advantage, and all possible precautionary measures should be taken to limit the effects of the conflict on civilians. It follows that restraint on landmine use can be exercised along four main dimensions: type of device, intended target, location, and precautionary measures (see Table 1).
The type of device can be classified by the device's effects: that is, antipersonnel or anti-vehicle, and whether it is activated by the victim or remotely command-detonated. Antipersonnel landmines are more disruptive to civilian life, as children and adults can easily and inadvertently activate them. Antivehicle mines' impact on civilians is more restricted since they are designed to detonate due to a vehicle's presence, proximity, or contact. Finally, command-detonated devices should allow for some degree of distinction between military and civilian targets and the avoidance of collateral damage. 89 Regarding the intended target, restraint varies in whether landmines target only Bell and Terry, "Combatant Rank and Socialization to Norms of Restraint, " 837-38. 89 Improvised explosive devices are included in the definitions of landmines or command-detonated devices as long as they conform to the intended effect (that is, antipersonnel or anti-vehicle) and the trigger mechanism (that is, victim-activated or command-detonated). Booby traps conform to the definition of victim-activated antipersonnel mines. combatants and military objectives or noncombatants and civilian objects. Armed groups might consider unarmed military personnel, police officers, and government supporters legitimate targets and differentiate them from other noncombatants, such as ordinary civilians. Civilian victimization also depends on the location of landmines, that is, whether laid in unpopulated or populated areas, or directly on or near civilian objects. Finally, restraint on landmine use can also be exercised through precautionary measures. Armed groups can reduce the risk of civilian victimization by issuing warnings before attacks and marking, fencing, and controlling access to minefields.

Legitimacy-Seeking Behavior
I build on Podder to measure rebel legitimacy-seeking behavior concerning local communities and the state government. 90 Local legitimacy is gauged on a scale of people's voluntary, quasi-voluntary and coercive compliance with insurgents. Voluntary compliance indicates higher levels of rebel legitimacy, and coercive compliance indicates lower legitimacy-seeking behavior. Quasi-voluntary compliance sits between the two. Legitimacy vis-à-vis the state can be analyzed along three main relationship types: collusive, conciliatory, and conflictual. Collusive relations with states entail cooperation with the government, conciliatory relations suggest some level of acceptance of compromise, and a conflictual relationship implies a higher level of hostility. Finally, I use Hyeran Jo's legitimacy-seeking indicator for measuring rebel relations with international audiences-that is, human-rights-conscious foreign sponsors. 91 In contrast with her operationalization, I measure foreign sponsorship qualitatively by how significant rebels' relationships with international actors are, ranging from insignificant to significant relations. Table 2 provides an overview. In short, I look into how rebels relate to different audiences as a proxy for their legitimacy-seeking behavior. These relationships' observable implications may vary significantly depending on the case. However, indicators that rebels seek legitimacy from local communities may include dependence on voluntary recruitment, funding, intelligence and sanctuary provided by locals, reliance on guidance from traditional or religious leaders, and the provision of governance in controlled territories. Concerning the state government, legitimacy-seeking rebels may engage in formal politics through their political wing or political allies, participate in political negotiations, and interact with national civil society organizations. Finally, support from human-rights-conscious state and nonstate actors, secessionist goals, and engagement with international NGOs may indicate legitimacy-seeking behavior toward international audiences. These relationships 90 Podder, "Understanding the Legitimacy of Armed Groups, " 687-98. 91 Jo, Compliant Rebels, 97-100. may vary in kind (for example, different audiences, purposes, and topics) and intensity (for example, quality and number).
For each case, I identify which behavior is more prevalent over specific periods for the dependent and independent variables and then arrive at aggregate-level values for each.

Restraint on Landmine Use in the Philippines
Though all three of the armed groups in the Philippines have used landmines and other explosive devices to some extent, they have displayed significant variation in restraint behavior. MILF has committed to, and complied with, forgoing antipersonnel mines; NPA has justified their use against government forces; and ASG has shown no effort to minimize civilian casualties. This section explores this variation vis-à-vis the armed groups' legitimacy-seeking behavior.

Moro Islamic Liberation Front: Banned Antipersonnel Mines
Concentrated in central Mindanao in the southern Philippines, MILF is the largest rebel group in the country, with a peak of 30,000-40,000 members in 2018. 92 It emerged in 1984 as a splinter faction of the Moro National Liberation Front (MNLF) and shared many similarities with its predecessor, especially regarding ethnic composition and self-determination goals. In contrast with the more secular MNLF, MILF's official ideology places Islam at the core of all social and political institutions. 93 Although MILF's long-term goal was to establish an independent Islamic state in its homeland, 94 it changed its claims from independence to autonomy in 2010, 95 98 MILF's local funding sources include the revenues from drug trafficking, 99 collection of obligatory alms and extortion, 100 local farmers' provision of food and other supplies to nearby units, and remittances from the Muslim Filipino diaspora. 101 Communities also support MILF's guerrilla operations through information gathering and sanctuary. 102 The group recognizes the value of community relations through community participation in decision making and the provision of services. MILF's provincial, municipal, and village committees, for example, include traditional, secular, and religious elites. 103 Further, MILF has established parallel government structures and provided various services, from marriage and birth certificates and adjudication to infrastructure development. 104 Over time, MILF's relationship with the Philippine government evolved from conflictual to conciliatory. Initially, MILF refused to participate in national-level politics, for instance, through elections and forming a political party. However, MILF repeatedly engaged with the government in formal negotiations-including ceasefires, framework agreements, and joint statements-even before officially changing its political goals from secession to autonomy, in 2010. 105 National civil society 106 and international organizations often facilitated these negotiations. The talks finally led to the signing of the Comprehensive Agreement on the Bangsamoro in 2014, paving the way to creating an autonomous region and the disarmament, demobilization, and reintegration of MILF combatants. 107 In MILF's struggle for self-determination, international support has varied over time. Sponsorship from human-rights-conscious international audiences is considered insignificant until the early 2000s. MILF had allegedly received weapons and other kinds of material support from several nondemocratic regimes, such as Libya and Malaysia. 108 These two countries also played a role in mediating MILF-GRP peace negotiations and ceasefire agreements, both directly and through the Organization of Islamic Cooperation. 109 Further, throughout its early years, MILF cooperated with several terrorist groups, such as al Qaeda, Jemaah Islamiyah, and ASG. 110 Especially after al Qaeda's September 2001 attacks and the ensuing US-led "War on Terror," MILF actively tried to distinguish itself from terrorist organizations and increased its cooperation with human-rightsconscious actors. 111 During this period, interaction with foreign audiences is assessed as significant. For example, an International Monitoring Team (IMT) was created after a ceasefire agreement in 2003 that included nationals of Malaysia, Brunei, Indonesia, Japan, Libya, and Norway, and personnel from the European Union and international NGOs. 112 Similarly, in 2009 MILF and the Philippine government established the International Contact Group for the Southern Philippines Peace Process (ICG-SPPP) to mobilize international support and leverage the peace process. The ICG-SPPP included the United Kingdom, Japan, Turkey, Saudi Arabia, and four international NGOs. 113 Several other international NGOs also engaged with MILF to advocate for the peace process and humanitarian and development issues. 114 Regarding landmine use, MILF has overall shown a high level of restraint. Already in 1976, MILF had allegedly banned the use of antipersonnel mines, arguing that they were contradictory to the teachings of Islam. Despite this claim, the rebel group at times publicly acknowledged the use of landmines. For instance, after government attacks in 1996, MILF stated that it "fought back with mortar, anti-tank weapons and mined fields. " 115 Under those circumstances, restricted landmine use was justified: If needed for defense (self-preservation), MILF … will use anti-personnel mines, but discriminately, only when the need of it arises or as the situation dictates, and upon order of the concerned MILF Commander on the ground during actual combat. MILF … APMs [antipersonnel mines] and ATMs [anti-tank mines], foreign and home-made, are command-detonated, tripwire, pressured triggered, or any of it. 116 Evidence suggests MILF has used landmines and other explosive devices mainly against the military, police, and government officials. To a lesser extent, MILF has also targeted specific groups of civilians (that is, Christian Filipinos), especially in the earlier years of the insurgency. However, internal regulations on landmine use prohibit the "use, stockpiling, production and transfer of APMs near population centers, places of worship, schools, business establishments, residential areas, farm and farm-to-market roads, and even areas inhabited by working animals." 117 The same guidelines ordered MILF fighters to map and visibly mark mined areas. Early reports from the Armed Forces of the Philippines (AFP) suggest that most mines were located in MILF-controlled territory close to bunkers, trenches, and abandoned camps. 118 However, some mines MILF supposedly laid were found along roads in disputed territories close to AFP checkpoints. 119 The justified use of antipersonnel mines for self-defense was finally dropped on 27 March 2000, when MILF further limited its landmine use policies and signed the "Deed of Commitment under Geneva Call for Adherence to a Total Ban on Anti-Personnel Mines and for Cooperation in Mine Action." 120 There is significant evidence that MILF's commitment was genuine. Following allegations of noncompliance, however, a Geneva Call verification mission confirmed instances of the use of improvised antipersonnel mines, arguably due to misunderstandings regarding such devices' technical specifications. The mission reiterated the technical requirements for anti-vehicle and command-detonated improvised explosive devices, which are not banned by the deed. 121 Consequently, MILF signed an expanded version of the deed 122 and issued written directives to its forces regarding compliance with the ban, mine awareness, and humanitarian law education programs for its members. 123 These changes came alongside MILF's other efforts to reject the use of terrorist tactics and renounce previous relationships with terrorist organizations, such as Jemaah Islamiyah and ASG. 124

Abu Sayyaf Group: Unrestrained Landmine Use
Concentrated in the southern Philippine islands, ASG is estimated to have had 1,250 fighters at its peak in 2000. 125 However, experts believe the group currently has fewer than 100 members. 126 ASG was organized by Abdurajak Janjalani in 1986 as a splinter group from MNLF. Janjalani and other dissatisfied MNLF members questioned their parent group's negotiation efforts with the Philippine government and its failure to appropriate Islamic concepts for the Moro struggle. The group's stated aim is to establish an independent Islamic state in Mindanao and expel Christian settlers who have migrated to the southern Philippines since the 1910s. 127 Despite ASG's stated goals, the group shows signs of becoming more materially motivated rather than ideologically driven, resembling a criminal gang. 128 ASG is still active and is often considered one of the top internal terrorist threats in the Philippines.
ASG's relations with local communities are characterized by coercive compliance. The group narrowly understands its core constituency as the Filipino Muslim minority aligned with the group's Islamic jihadist ideology. Popular support for ASG is limited, as moderate Muslims mainly refute its interpretation of Islam and the means of jihad. 129 Still, ASG has avoided direct military confrontation by seeking refuge among local communities. Recruitment to the group is voluntary, often following kinship lines, and is supported by limited family and village networks. Motivations to join vary, including ideological commitment, opposition to the state, personal enrichment, machismo, and physical survival. 130 Almost 98 percent of ASG's funds come from kidnapping and extortion, followed by shares of almsgiving and diverted resources from Islamic NGOs. 131 ASG's relationship with the Philippine government is conflictual, as the group rejects political negotiation in favor of violent jihadism. 132 ASG has not engaged in formal dialogue with the Philippine government and has used hostage and ransom negotiations to express some of its political demands. 133 The group's relations with human-rights-conscious international audiences are insignificant at best and confrontational at worst. Nondemocratic countries, such as Libya, have backed ASG, and the group has also garnered support from or declared allegiance to regional and international terrorist groups such as al Qaeda, the Islamic State, and Jemaah Islamiyah. 134 Unlike other Philippine Islamic groups, such as MILF and MNLF, the Organization of Islamic Cooperation does not recognize ASG. On top of not cooperating with international NGOs, ASG has recurrently targeted foreign aid workers for kidnap-for-ransom and extortion. 135 In general, ASG has shown a low level of restraint on landmine use. The group demonstrated no significant efforts to avoid collateral damage and has directly targeted civilians. ASG laid victim-activated landmines primarily for perimeter defense in its jungle and mountain camps and command-detonated devices in offensive operations. Explosive devices are also used in civilian terror attacks. ASG's victims include mostly Filipino citizens, from government officials to Christians and Muslims who oppose the group's interpretation of Salafi jihadism. 136 Out of 172 recorded ASG attacks using explosive devices, for example, none were directed against military targets, and 12 were against police forces. Most ASG attacks were against noncombatants in populated areas. 137 I found no evidence of precautionary measures implemented by ASG to avoid civilian victimization.

New People's Army: Selective Use of Landmines
The leading armed group of the communist insurgency is the Communist Party of the Philippines' (CPP) New People's Army. Formed in 1968 and active since 1971, the group wages a "protracted people's war" against the government and operates throughout the country, although it is concentrated in rural and remote areas. Its main objective is to overthrow the Philippine government and establish a people's republic. The group is estimated to have more than 10,000 members and poses the greatest armed threat to the Philippine government given its size, military capability, and activity. 138 The relationship between NPA and local communities takes the value of voluntary compliance. Following a communist ideology, the group broadly understands its core constituency as the Filipino working class, regardless of nationality, ethnicity, or religious affiliation. 139 During the 1970s and 1980s, many educated and ideologically motivated urban youth joined the insurgency. As those members progressively left due to the group's inability to advance their armed struggle, its ranks shifted to mainly poor rural youth. More recently, NPA has found more support prompted by the expansion of mass organizations in response to government corruption and inefficiency. 140 Most of NPA's funding comes from individual supporters, "revolutionary taxes, " and "permits to campaign. " "Revolutionary taxes" are a form of extortion applied to locals in their areas of operations, 141 and the group sells "permits to campaign" to politicians who wish to campaign in NPA-controlled territory. 142 In some areas, the group has established a "shadow government" to build support networks among the population. 143 These parallel governance structures include taxation systems, business ventures, farmers' cooperatives, and other development plans. 144 The record of NPA's relations with the Philippine government is mixed but primarily conflictual. NPA challenges the existing political structures and thus avoids negotiating through them. However, the CPP-NPA has sporadically engaged in peace talks with the Philippine government. For example, the insurgency has signed the Comprehensive Agreement on Respect for Human Rights and International Humanitarian Law 145 and committed to ceasefires during disaster relief and special truces. 146 The US Department of State's 2002 recognition of NPA as a terrorist group brought additional challenges to political engagement between NPA and the Philippine government. 147 NPA's foreign support is moderate. CPP-NPA is aligned with a global communist movement. Accordingly, the group attempted to establish working relations with a series of like-minded state and nonstate actors to attract foreign funds, though without much success. While only a small number of them allegedly provided financial aid, arms, training, or other types of support to NPA, most foreign assistance drastically decreased following the Soviet Union's collapse in the 1990s. 148 Still, NPA maintains an overseas diplomatic representation in the Netherlands. The group does not seem to engage with international NGOs significantly.
Overall, NPA has shown a moderate level of restraint on landmine use. The group has, since 1998, formally agreed to restrict the use of landmines. 149 Accordingly, NPA claims to follow a discriminate strategy in using landmines only as "legitimate offensive weapons against military vehicles of the AFP and Philippine National Police transporting troops and supplies across and within AFP-NPA battlefields." 150 CPP further attested that landmines "are carefully attended to by NPA fighters to ensure that these are fired only against legitimate military targets." 151 Indeed, the striking majority of mine casualties attributed to NPA, when details are known, were caused by command-detonated anti-vehicle explosive devices used against government forces. 152 However, several reports indicate the use of victim-activated explosive devices for both defensive and offensive purposes. For example, NPA strongholds and weapons production bases were mined. 153  on a rebel camp, AFP reports referred to the base as "booby-trapped." 155 Similarly, CPP acknowledged that NPA had used "contact-detonated [victim-activated] mines … for a limited time and limited range and under close supervision of NPA command concerned in order not to cause risk for civilians." 156 Although most mine victims attributed to NPA are Philippine military and police personnel, the group also targets civilians, namely government officials and politicians, journalists, foreign investors, and business owners. The group perceives these individuals as legitimate targets through a logic of "revolutionary justice." 157 The few civilians inadvertently killed in NPA attacks are usually family members traveling in military vehicles or those mistakenly identified as government forces. 158 In these cases, NPA has publicly apologized and offered to pay hospital expenses for the injured. 159 Still, reports from the Philippine military condemn the use of landmines by NPA around populated areas, such as roads, villages, and farms. 160 Further, according to government forces, mined areas discovered around NPA camps were unmarked. 161

Explaining Variation in Restraint on Landmine Use
The empirics support the tested hypotheses. H1 suggests that rebels seeking legitimacy from local communities are more likely to limit the effects of landmines on their perceived constituency. Both MILF and NPA relied on voluntary compliance by their local constituents, indicating high legitimacy-seeking behavior toward local communities. As expected, the two groups showed significant efforts to minimize civilian victimization due to landmine use. They prioritized employing command-detonated explosive devices in unpopulated areas and called for their fighters to record, mark, and control access to their minefields. ASG relied on coercive compliance by local communities and demonstrated no effort to avoid civilian casualties from its landmines. Instead, the group directly targeted civilians with terrorist tactics. Moreover, even though MILF effectively banned antipersonnel mines, NPA also seemed concerned with avoiding civilian casualties from its landmines. For example, NPA publicly apologized and paid for the medical costs of civilians injured by their landmines, clearly communicating its restraint behavior to its perceived constituency.
The difference among the groups in targeting noncombatants is also worth noting. While no group directly targeted its constituents with landmines and other explosive devices, the perception of who comprised their constituency varied greatly according to each group's ideology. NPA's broader understanding of its constituency, which includes the majority of the Filipino population regardless of geographic concentration, ethnicity, or religion, meant attacks against noncombatants were limited to select individuals, such as government officials and business owners. In contrast, MILF's more limited notion of its core constituency, restricted to the Mindanao Muslim population, arguably justified attacks against specific groups of civilians, such as Christian Filipinos. On the far end of the spectrum, ASG considers as constituents only the few who share the group's interpretation of Salafi jihadism. Therefore, ASG targeted a broader range of noncombatants.
The cluster of hypotheses that make up H2 states that rebels seeking legitimacy from the state government are more likely to (a) comply with national law related to landmine use; (b) reciprocate government behavior; and (c) limit the effects of landmines on the government's constituents. As expected, the groups' political goals directly influenced how they interacted with the national government. MILF had a more conciliatory relationship, from early political engagement in the late 1990s to the 2014 peace agreement. Because MILF had already committed to stop using antipersonnel mines in early 2000, it is hard to attribute MILF's restraint on landmine use only to its legitimacy-seeking behavior toward the state government. However, 2010 marked a shift in MILF's relationship with the Philippine government when the group changed its independence claims to autonomy goals. Peace talks had been stalled since 2008, and, as a confidence-building measure, the parties agreed to launch a joint mine and unexploded ordnance clearance project with national and international NGOs. 162 MILF's conciliatory posture has arguably leveraged its political capital toward the government, especially relative to other rebel groups in the country.
On the contrary, NPA and ASG displayed conflictual relations with the state. They rejected formal politics and sought to overthrow the Philippine government, either nationally or regionally. As predicted, the groups continued to use antipersonnel mines and booby traps. Whereas ASG directly targeted the Philippine government's constituency as its main tactic, NPA showed considerable restraint toward the Filipino population at large but targeted civilians perceived as enemies of the revolution, such as government officials and business owners.
Finally, H3 suggests that rebels seeking legitimacy from human-rightsconscious international audiences are more likely to comply with international law relevant to landmine use. Before the 2000s, MILF had been supported by foreign actors with records of human rights violations. During that period, MILF justified laying antipersonnel landmines in self-defense and against government forces. In the early 2000s, however, MILF experienced a drastic change in its legitimacy-seeking behavior. On 21 March 2000, the Philippine government launched an offensive campaign against MILF. Less than a week later, on 27 March, the rebel group signed the first version of Geneva Call's Deed of Commitment to stop using antipersonnel landmines. Contrary to the "poor man's weapon" and desperation arguments, which should predict less restraint in the face of a massive military challenger, MILF chose to exercise a higher level of restraint, perhaps to credibly signal its legitimacy to the international community and gain support against GRP.
During this period, following the US-led "War on Terror," MILF leadership demonstrated that the armed group abided by international law and was thus a legitimate political actor. Accordingly, MILF established more substantial relationships with human-rights-conscious foreign audiences via international NGOs, the IMT, or the ICG-SPPP. In line with the hypothesis, MILF showed more consistent compliance with its commitments to international norms, as depicted by its repeated commitments to the antipersonnel mine ban in 2000 and 2002, the welcoming of an international monitoring mission, and active measures to enforce the ban, including education and punishment of their cadres who disobeyed.
ASG, on the contrary, rejects international norms and is supported mainly by a few state and nonstate actors with a record of human rights violations. Unsurprisingly, ASG showed no effort toward compliance with the relevant bodies of international law on landmine use. In between MILF and ASG, NPA sought support from a wide range of like-minded state and nonstate actors with mixed records of human rights violations. Expectedly, although NPA formally signaled commitment to broad human rights and humanitarian norms, it at times flexed its compliance to suit military needs, justifying the use of antipersonnel landmines due to military considerations.
The variation in landmine use observed in the Philippines cannot be fully explained by alternative predictors of restraint. First, in line with Fazal and Konaev, 163 political goals and military strength alone do not seem enough to push rebel groups to exercise restraint on landmine use. 163 Fazal and Konaev, "Homelands versus Minelands. " MILF and ASG, both secessionist groups, should have more incentives to comply with humanitarian norms than the revolutionary NPA. However, MILF exercised the highest levels of restraint, ASG displayed the lowest, and NPA showed a moderate level of restraint. Further, Fazal and Konaev's argument 164 that militarily strong and secessionist rebels are more likely to exercise restraint may explain MILF and ASG's behavior, but not NPA's. Second, although ideology helps understand how armed groups define and relate to their constituencies and other audiences, it alone cannot explain restraint behavior. Among the Moro rebels, MNLF used landmines mostly selectively, 165 MILF committed to the total ban, and ASG is a major user with little restraint. On the communist front, NPA still uses landmines when necessary, but two other groups have committed to the ban. 166 Further, the claim that territorial control should lead to landmine use for defensive purposes does not hold for the analyzed cases. Whereas MILF and NPA controlled significant portions of territory and ASG did not, no group deployed extensive defensive minefields. MILF committed to the antipersonnel mine ban, and NPA prioritized employing command-detonated devices and landmines in offensive ambushes against government forces. Neither does the expectation that weaker groups lack access to alternatives to landmines hold. ASG, the smallest group in the sample, often employed more discriminate means, such as commanddetonated devices, but with lower levels of restraint. Finally, although there have been isolated clashes between MILF and ASG, the three groups have largely operated independently against the Philippine government, thus decreasing the relevance of their relative strength in explaining restraint on landmine use.
In summary, while different ideology and political goals shaped rebels' legitimacy-seeking behavior toward local communities, the state government, and international audiences, they did not seem to independently influence restraint on landmine use. Further, territorial control, group size, and relative strength between rebel groups cannot account for variation in restraint on landmine use in the Philippines.

Engaging Legitimacy-Givers
In this article, I build on the argument that legitimacy-seeking rebels are more likely to exercise restraint 167 and argue that how restraint is exercised depends on which audiences rebel groups rely on. A comparative analysis of three rebel groups in the Philippines corroborates the main argument: reliance on local communities led to a greater focus on the protection of the group's perceived constituency; stronger relationships with the government emphasized compliance with national law, reciprocity, and protection of the government's constituency; and human-rights-conscious foreign sponsors encouraged compliance with international law.
The article contributes to research on civil war restraint in two main ways. First, it provides an original conceptualization of restraint based on the practical application of humanitarian norms to landmine use. This more nuanced focus advances previous studies on commitment and compliance with the antipersonnel mine ban 168 and discloses unobserved details of how rebels exercise violence and restraint. Similar conceptualizations may prove fruitful in studying specific means and methods of war. 169 Second, it expands previous legitimacy-and reputation-based theories of rebel restraint 170 by focusing on how particular audiences shape restraint behavior in relation to landmine use. This article's limitations offer opportunities for future studies, which should continue to explore how different actors influence rebel behavior, including differences within audiences usually clustered together (for example, the "international community"). More work should be done to expand the findings' generalizability, investigate the suggested causal mechanisms, and look deeper into rebel decision-making processes. Lastly, future theoretical frameworks should account for potential interaction effects between the different influencing factors of restraint on landmine use, such as military considerations, organizational cultures, ideology, and political objectives.
This article suggests that engaging armed groups and their legitimacy-givers may be valuable for encouraging rebel restraint on landmine use. Active engagement can communicate different audiences' preferences to rebel groups, such as respect for humanitarian norms and civilian protection. Likewise, external actors' promotion of mine awareness for local communities, domestic actors, and foreign sponsors can pressure rebel groups to exercise restraint on landmine use. Although humanitarian engagement should not legitimize armed groups, it can educate and convince rebels that abiding by restraint norms may become an added source of legitimacy. It follows that rebel groups might capitalize on engaging with various actors over the landmine ban, for instance, to gather support and foster their armed struggle. Therefore, current and future engagements with nonstate armed actors should be transparent and constantly reinforce their humanitarian nature. While this article looked into legitimacy-seeking behavior as a driver of rebel restraint on landmine use, several other factors remain underresearched. For instance, when groups face a high number of casualties from accidents with their own mines, whether during production and emplacement or in movement, restraint might be the rational choice to maintain safety, morale, and continuity of operations. 171 Alternatively, there might be an economic reason not to use landmines. Groups may foresee potential impediments in conducting economic activities due to the denial of productive land and future mine clearance costs. 172 Finally, rebel engagement with and support of mine action activities remain a fruitful avenue for future research. For example, why do some rebel groups conduct humanitarian demining, an activity that is dangerous, time-consuming, and one that potentially diverts valuable military resources?