ANALYTICAL ESSAY Ceaseﬁre Violations: Why They Occur and How They Relate to Strategic Decision-Making Processes

Almost all ceaseﬁres experience violations, yet we know little about how such violations relate to the military and political aspirations of conﬂict parties. This article builds on ceaseﬁre and bargaining literature to un- derstand why ceaseﬁre violations occur and how they relate to strategic decision-making processes. Building on these theoretical insights, it pro- poses a typology of four main types of ceaseﬁre violations: strategic violations serve to strengthen the military advantage of a conﬂict party, retaliatory vi- olations seek to ensure ceaseﬁre compliance, spoiling violations aim to undermine the efforts of leaders, and localized violations are delinked from strategic decision-making processes. A case study of a major ceaseﬁre violation in the Bangsamoro peace process illustrates how we may use infor- mal Bayesian reasoning to empirically distinguish between these different types of violations. Treating ceaseﬁre violations as part of wider military and political processes enables us to better understand the causal conditions under which ceaseﬁre violations occur and identify strategic interests of different actors to carry out these violations. This helps explain the varied responses to ceaseﬁre violations and sharpens our understanding of how to address them. mieux comprendre les conditions de causalité qui se cachent der-rière l’apparition des violations de cessez-le-feu et identiﬁer les intérêts stratégiques des différents acteurs dans la conduite de ces violations. Ainsi, nous pouvons plus facilement expliquer les différences entre les réponses aux violations de cessez-le-feu. Nous pouvons également afﬁner notre com-préhension des manières d’y répondre.

Almost all ceasefires experience violations, yet we know little about how such violations relate to the military and political aspirations of conflict parties. This article builds on ceasefire and bargaining literature to understand why ceasefire violations occur and how they relate to strategic decision-making processes. Building on these theoretical insights, it proposes a typology of four main types of ceasefire violations: strategic violations serve to strengthen the military advantage of a conflict party, retaliatory violations seek to ensure ceasefire compliance, spoiling violations aim to undermine the efforts of leaders, and localized violations are delinked from strategic decision-making processes. A case study of a major ceasefire violation in the Bangsamoro peace process illustrates how we may use informal Bayesian reasoning to empirically distinguish between these different types of violations. Treating ceasefire violations as part of wider military and political processes enables us to better understand the causal conditions under which ceasefire violations occur and identify strategic interests of different actors to carry out these violations. This helps explain the varied responses to ceasefire violations and sharpens our understanding of how to address them.

Introduction
Ceasefire violations differ from other forms of hostilities during armed conflicts, in that they demonstrate the unwillingness or inability of conflict parties to adhere to their own commitments. Yet, despite their common occurrence, and their often critical effect on the trajectory of a conflict, ceasefire violations remain a blind spot in conflict research ( Wiehler 2021 , 418). Ceasefire research points to the interrelated relationship between political negotiations and ceasefire compliance ( Höglund 2011 ;Åkebo 2016 ) and highlights strategies for preventing or addressing violations ( Fortna 2003( Fortna , 2004Clayton and Sticher 2021 ;Pinaud 2021 ;Wiehler 2021 ). However, the varying causal conditions that lead to violations remain underexplored. This is an important gap, as reasons for why violations occur shape the responses of conflict parties and condition the effectiveness of strategies to contain further escalation. A broader body of literature explores impediments that occur during peace processes-from informational problems to spoiling behavior-but to date, these insights have not been systematically linked to ceasefire violations.
This article connects insights from these two strands of research, showing how ceasefire violations relate to the larger military and political aspirations of conflict actors. Two basic propositions stand at the core of the analysis: (1) that the moti-vation to engage in a ceasefire shapes conflict party leaders' interest in ceasefire compliance and (2) that commanders and lower ranks may not always pursue the same goal as the conflict party leadership. These two propositions relate to two main layers of strategic decision-making: decisions taken at leadership level and those taken by other members of the party. At leadership level, actors may have a vested interest in ceasefire compliance, or they may use or allow violations to achieve a military advantage. At the level of commanders and lower ranks, actors may follow the chain of command, or they may seek to undermine leaders through spoiling behavior. Importantly, not all violations are linked to strategic decision-making processes: members of conflict parties may violate an agreement involuntarily or for reasons not linked to the wider conflict.
This leads me to identify four basic types of ceasefire violations: strategic violations form part of conflict party leaders' cost-benefit calculations and serve to achieve a military advantage; retaliatory violations equally form part of leaders' cost-benefit calculations, but serve to preserve the continuation of the ceasefire; spoiling violations are deliberate efforts by factions or lower ranks to undermine the efforts of a conflict party leadership; and localized violations are transgressions that are delinked from strategic decision-making processes.
It is challenging to empirically differentiate between these different types of violations, as actors have an incentive to misrepresent their actions and intent in certain situations. I propose a Bayesian approach to address this challenge, identifying observable implications that are more common for certain types of violations than for others-including characteristics of the conflict context, the violation itself, and responses by the conflict parties. I then demonstrate how to apply this approach empirically, using a major ceasefire violation in the peace process between the Philippine government and the Moro Islamic Liberation Front (MILF) as a case study.
Placing the analysis of ceasefire violations within a strategic decision-making framework, this article represents a significant departure from the way ceasefire violations are commonly treated, namely as idiosyncratic events. The main takeaway is that we should view ceasefire violations as events that form part of larger military and political processes. The theoretical insights and typology enable us to better understand the causal conditions under which ceasefire violations occur and identify strategic interests of different actors to carry out these violations. This not only helps explain the varied responses by the targets of a violation, but also sharpens our understanding of how to effectively address prevailing violations.
The article proceeds as follows. After clarifying key terminology and the scope conditions of my framework, the first section provides a brief overview of what we know about ceasefire violations from the existing research. The second section contains the theoretical framework for why ceasefire violations occur, discussing military and political calculations at leadership level, when and why commanders and lower ranks may engage in violations to undermine incumbent leaders, and why violations may occur delinked from strategic decision-making processes. The third section introduces the four main types of ceasefire violations and observable implications for each type, and the fourth section illustrates how to apply the framework empirically. The concluding section discusses implications for theory and policy and identifies future research avenues.

Terminology and Scope Conditions
This article focuses on violent, political conflicts between organized armed actors. It is based on an instrumental understanding of war, assuming that conflict parties are not interested in conflict violence per se, but rather use it in the pursuit of a specific goal (see Slantchev 2003 , 623). Similarly, this article assumes that conflict party leaders use ceasefires as strategic instruments in the pursuit of their larger Downloaded from https://academic.oup.com/isr/article/24/4/viac046/6701908 by guest on 22 September 2022 military or political aspirations. In such an instrumental view, ceasefires are not an aim in themselves, but form part of strategic decision-making processes. 1 There is no consensus on what constitutes a ceasefire ( Bara, Clayton, and Aas Rustad 2021 , 332). In the media, the term is often used as shorthand for any halt in fighting ( Lyons 2016 ). In peace processes, the terminology tends to be highly contested, with actors using terms such as cessation of hostilities , truce , or humanitarian pause to imply specific meanings (see Åkebo 2013 , 22). Scholars also use the term "ceasefire" in varying ways, focusing on an agreement that sets a process in motion ( Kreutz 2010 ;Åkebo 2013 ), the process itself ( Clayton et al. 2019 ;Sticher and Vukovi ć 2021 ), or viewing a ceasefire as the outcome of a process ( Beardsley et al. 2006 ;DeRouen, Bercovitch, and Pospieszna 2011 ). What the different approaches share in common is an understanding that a ceasefire addresses conflict violence, and not the main disagreement between conflict parties ( Clayton et al. 2019 , 2;Kreutz 2010 ).
Following Clayton et al. (2019 , 2), this article defines ceasefires as "arrangements by or between conflict parties to stop fighting from a specific point in time." This includes unilateral, bilateral, and multilateral arrangements that vary greatly in terms of their geographical coverage, time period, formality, and provisions. It does not include arrangements that de-escalate rather than suspend conflict violence (such as a declaration to stop targeting critical infrastructure). This relatively broad definition enables the proposed theory of ceasefires to be widely applicable, rather than focusing only on one specific type of arrangement.
The proposed theory focuses on ceasefires that are put in place prior to the conclusion of a peace agreement (i.e., non-definitive ceasefires). While accounting for third-party pressure, the theory primarily relates to conflicts in which the parties are free to decide whether to engage in consensus-based arrangements (such as engaging in peace negotiations, accepting a bilateral ceasefire agreement, or signing a peace agreement).
I also use a broad understanding of ceasefire violations, including both specifically prohibited activities and failure to carry out the activities prescribed (such as withdrawing troops from a demilitarized zone). A ceasefire violation can involve multiple acts of transgression that relate to a single conflict event, 2 and may encompass violent and nonviolent acts, although the focus here is on violent acts (see discussions in Nilsson and Söderberg Kovacs 2011 , 610-12). Again, this relatively broad understanding of a ceasefire violation ensures that the proposed theory is broadly applicable, possibly at the expense of specificity. Future research could refine the framework and typology by, for example, investigating the prevalence of different types of violations for different types of arrangements.

What We Know about Ceasefire Violations
This section provides an overview of what extant research tells us about ceasefire violations. It starts with insights into the causes and effects of ceasefire compliance more generally, before discussing existing characterizations of ceasefire violations in the literature. 1 In theory, an instrumental understanding of ceasefires excludes ceasefires that are purely humanitarian in nature.
In practice, however, humanitarian ceasefires also often serve a larger political goal (see Wiehler 2020 ).
2 For example, a conflict party may launch an attack in an opponent's zone of control. Such a ceasefire violation comprises two (or more) acts that violate an agreement, that is, the attack itself and the movement of troops into the opponent's zone of control.

Causes and Effects of Ceasefire Compliance
A large and growing body of literature focuses on ceasefires in armed conflicts, discussing various aspects of the complex relationship between conflict contexts, ceasefires, and peace negotiations. Scholars have identified a wide variety of reasons why conflict party leaders may refuse to engage in a ceasefire, including to avoid looking weak, misjudging the military balance, or the absence of political negotiations (see Smith 1995 ;Dukalskis 2015 ). Reasons to engage in a ceasefire are even more manifold and range from initiatives targeted at gaining a military advantage through re-arming, re-organizing, or gaining breathing space ( Crocker, Hampson, andAall 2004 , 158, Chounet-Cambas 2011 , 7-8, 20), to demonstrating military strength and resolve ( Mahieu 2007 ;Höglund 2011 ;Åkebo 2016 ;Wiehler 2021 ), signaling good faith ( Höglund 2011 ), or creating a conducive environment for peace negotiations ( Greig and Diehl 2005 ;Clayton et al. 2019 ). Importantly, these underlying objectives do not necessarily align with the stated purpose of a ceasefire ( Clayton, Nathan, and Wiehler 2021 ). Given the variety of reasons why conflict parties engage in a ceasefire, and the different contexts in which they take place, it comes to no surprise that the effects of ceasefires vary greatly, both in terms of addressing conflict violence (see Lundgren, Karakus, and Svensson 2019 ;Karakus and Svensson 2020 ;Clayton and Sticher 2021 ) and the contested issues between conflict parties (see Höglund 2011 ;Åkebo 2013. While studies into the effects of ceasefires tend to focus on ceasefire compliance, they also make pertinent assumptions about and observations on the nature of ceasefire violations, including why such violations occur and how they may affect conflict. Importantly, these studies show that while violence is an accepted part of armed conflicts, violence during a ceasefire has a different function: it implies that conflict parties are not willing or not able to follow through on their commitment to suspend conflict violence. This may undermine trust and confidence in the ability to find a negotiated conflict settlement ( Åkebo 2013 , 201-203).
Ceasefire violations may also expose conflict party leaders to internal criticism and play into the hands of hardliners, as they fuel the sentiment that a negotiated settlement with an enemy is not possible ( Höglund 2011 ;Åkebo 2013 ). They possibly indicate a lack of command and control, making it difficult to build confidence in a negotiating partner that may not be able to deliver ( Höglund 2011 ).
Research into the responses to ceasefire violations shows that states can prevent exploitative behavior by building up a history of retaliations ( Long 2014 ). Conflict party leaders interested in ensuring ceasefire compliance have an incentive in choosing an appropriate response, to disincentivize future violations while mitigating the risk of involuntary conflict escalation ( Wiehler 2021 ).
Yet, retaliations are not the only measure conflict party leaders can take to increase the chances of ceasefire compliance. Those with an interest in ensuring compliance may self-commit to certain measures that mitigate the risk of ceasefire violations or decrease the fallout from such events. In her seminal work on the effects of ceasefire provisions in interstate conflicts, Fortna (2003Fortna ( , 2004 outlines three main causal mechanisms through which conflict parties may lower the risk of ceasefire violations: (1) raising the costs of attack to lower the incentive to violate a ceasefire; (2) reducing uncertainty by signaling willingness to accept costly measures; and (3) containing escalatory dynamics. She then shows how ceasefire agreements that include monitoring provisions are associated with a longer absence of conflict violence. Others argue that enforcement problems are secondary if conflict parties struggle with distributional problems, suggesting that a ceasefire agreement breaks down more easily if it is imposed by a third party and does not reflect perceptions of the relative strength between the conflict parties ( Werner and Yuen 2005 ).
A similar argument is made in the context of intrastate conflicts, that is, that the objective and subjective quality of ceasefire design matters, but only in contexts where ceasefires are not imposed on the conflict parties ( Nathan and Sethi 2020 ). An analysis of all written ceasefire agreements in intrastate conflicts between 1990 and 2019 finds that monitoring mechanisms, among other agreement provisions, increase the likelihood that a ceasefire arrangement holds ( Clayton and Sticher 2021 ). These findings reflect both an independent effect of monitoring provisions and an increased willingness to comply with a ceasefire by those who sign on to such provisions ( Clayton and Sticher 2021 , 15-16). This highlights the need to pay attention to the strategic situation in which monitoring arises when assessing various aspects of compliance mechanisms.

Characteristics of Ceasefire Violations
Beyond these more general insights that relate to ceasefire violations, existing studies highlight various violation characteristics that may shape the trajectory of conflicts. One key distinction is between violations that are intentional and those that are involuntary (see Fortna 2003 , 345). Related terms are premeditated , to contrast violations that were planned in advance from those that were accidental or carried out in self-defence ( Fortna 2003 , 340;Long 2014 , 3), or referring to a (violent) violation as an attack or ambush as opposed to a clash . Happymon (2013 , 3) distinguishes between violations committed at the orders of a conflict party leadership and those occurring in response to local dynamics. Such information can be key to gauge the potential effects of a violation on peace negotiations (see, e.g., The Economist 2015 ). The separate but related literature on spoiling in peace processes discusses such dynamics more generally (see Stedman 1997 ;Zahar 2008 ;Nilsson and Söderberg Kovacs 2011 ;Whiting 2016 ).
Conflict parties and monitoring missions tend to introduce their own terms to categorize ceasefire violations. Such terminology is often negotiated and reflective of a specific purpose ( Clayton et al. 2022 ). The Organization for Security and Cooperation in Europe (OSCE) ceasefire monitoring mission in Ukraine, for example, listed violations by a number of injured and fatalities or by the type of weaponry that was used. The Joint Security Committee monitoring the implementation of the 2002 ceasefire in Aceh (Indonesia) listed violations with regard to the potential impact they had on peace negotiations ( Huber 2004 , 36). While some ceasefire agreements simply evoke a suspension of hostilities, others, such as the Nuba Mountains ceasefire agreement or the nationwide ceasefire in Myanmar, specify prohibited and prescribed activities (see Forster 2019 ).
The literature also shows how in some cases, it may not be clear whether a specific act qualifies as a ceasefire violation. Information may be lacking; for example, if it is unclear whether the violation was committed by a party to the ceasefire or an actor outside the arrangement ( Åkebo 2019 , 483; see also Höglund 2005 ). A ceasefire agreement may be vague or ambiguous, making it harder to determine whether an act was indeed a violation ( Fortna 2003 ;Wiehler 2021 ). The 2005 Memorandum of Understanding between the Government of Indonesia and the Free Aceh Movement distinguishes between violations in letter and spirit (see Chounet-Cambas 2011 , 14, 34), to ensure that parties not only stick to the text of the agreement, but are also bound by a broader understanding of how compliant behavior looks like. In this context, violations in letter may be understood as acts that are specifically prohibited in a ceasefire agreement or noncompliance with prescribed activities, while violations in spirit relate to acts that may not be specifically forbidden, but that undermine the stated purpose of a ceasefire.
In sum, from existing ceasefire research, we know that ceasefire violations play a vital role in shaping armed conflicts and their resolution, and we know that there are multiple reasons for why violations occur. What we lack is a framework that ties these reasons together and systematically links them to wider military or political aspirations of the conflict parties. We also lack a comprehensive ceasefire typology that speaks to the causal conditions and strategic interests of conflict actors. The next sections address both gaps, first proposing a strategic decision-making framework to identify main reasons for ceasefire violations and then building on these reasons to develop a systematic ceasefire typology.

Why Ceasefire Violations Occur
To understand the varying nature of ceasefire violations and their effects, we need to understand why they occur. Building on bargaining theory and ceasefire research, this section theorizes how ceasefire violations relate to elite decision-making processes. It is divided into three sections: first, it assesses how perceptions about the bargaining situation affect strategic interests of leaders to comply with a ceasefire; second, it discusses when and how internal divisions may incentivize actions to undermine a conflict party leadership; and third, it shows when and how actors engage in violations that are delinked from wider conflict issues.

Strategic Decision-Making at Leadership Level
Conflict party leaders take the decision to engage in a ceasefire deliberately, considering the expected costs and benefits of a unilateral or reciprocal arrangement. Expected compliance or noncompliance forms part of this cost-benefit analysis. To assess these decisions, we have to understand whether leaders are interested in conflict settlement or pursue a military approach to the conflict. As clearly argued below, I do not assume that leaders are either inherently good or inherently bad or opportunistic. Instead, I develop a situational typology in which preferred approaches depend on perceptions of relative capabilities and resolve-which may change over time as they update their expectations or as the context changes.

Imperfect Information and Interest in Compliance
A mainstay of the bargaining theory of war is the problem of imperfect information. The problem offers a key explanation of why wars occur, even though ex post , they are always inefficient ( Schelling 1960 ;Fearon 1995 ;Powell 1999 ;Walter 2009 ). The basic idea is that, because a negotiated settlement reflects the distribution of military power, both sides have an incentive to misrepresent information about their own capabilities and resolve, and by extension do not trust the information provided by an opponent. As a result, actors-at least those that end up going to waroverestimate their own relative strength and underestimate the costs of war. This makes a military approach appear attractive ( Schelling 1960 ;Fearon 1995 ;Walter 1997 ;Powell 1999 ;Wagner 2000Wagner , 2009. Through fighting ( Wagner 2000 ;Powell 2004 ;Werner and Yuen 2005 ;Walter 2009 ), peace negotiations ( Wagner 2000 ;Filson and Werner 2002 ;Slantchev 2003 ;Powell 2004 ), and ceasefires ( Sticher and Vukovi ć 2021 ), conflict parties reveal information that leads them to converge in their assessments of a military outcome. However, even when their assessments sufficiently converge to reveal a space for a negotiated settlement, bargaining problems often persist: expectations about a mutually acceptable agreement may still considerably diverge ( Werner and Yuen 2005 ;Beardsley 2008 ;Findley 2013 ), or actors face other obstacles, such as credible commitment problems ( Fearon 1995 ;Walter 1997Walter , 2009 ), indivisible stakes ( Pillar 1983 ;Hassner 2003 , 24-26), or insufficient constituent support ( Haass 1988 ;Putnam 1988 ;Sticher 2021 ).
Ceasefires can serve as strategic bargaining instruments in both situations-when parties pursue a military outcome or when they begin to see the benefits of a negotiated settlement. However, leaders will shift their use of ceasefires in line with their underlying motivation for engaging in a ceasefire ( Sticher and Vukovi ć 2021 ).
As long as leaders continue to pursue a military outcome, their key concern is to strengthen their fighting capabilities and resolve, and diminish the capabilities and resolve of the opponent. In these situations, leaders will pursue a ceasefire if doing so promises to strengthen them militarily, for example, by providing their troops breathing space, allowing them to re-arm and re-group, or to attack an opponent when it is most vulnerable ( Crocker, Hampson, and Aall 2004 , 158;Chounet-Cambas 2011 , 7-8, 20). Leaders who prefer a military outcome over conflict settlement have no vested interest in ceasefire compliance. They may not care about violating an agreement if it serves them militarily, or they may even engage with the intention to violate an agreement.
Leaders who recognize the benefits of a political approach and are genuinely interested in conflict settlement assess the benefits of a ceasefire differently. They are primarily interested in a ceasefire if such an arrangement increases their bargaining position or the chances of reaching a settlement ( Sticher and Vukovi ć 2021 ). In contrast to leaders who prefer a military outcome, leaders with an interest in settlement need to ensure that their behavior does not impede current or future negotiation efforts. Repeated noncompliance may send a signal that they are not genuine about a peaceful resolution of a conflict, or that they cannot be trusted with the implementation of an eventual agreement ( Åkebo 2013 ;Nathan and Sethi 2020 ). In contrast, conflict party leaders can signal that they are both willing and able to comply with their promises through ceasefire compliance. They can show control over their forces, which may increase their bargaining position ( Höglund 2011 ;Åkebo 2016 ). More generally, successful ceasefires can create a conducive environment for peace negotiations, thereby increasing the chances of settlement ( Clayton, Nathan, and Wiehler 2021 ).
To be perfectly clear, leaders who prefer settlement over a military outcome also care about their military strength-not least because an eventual settlement will likely reflect the relative strength between the conflict parties ( Fearon 1995 ). They may refuse a ceasefire if they believe that it will primarily benefit the opponent. However, while fighting becomes an accepted part of wars, committing to a ceasefire and later violating it are often seen as unacceptable, especially if a violation is a major transgression and seen as intentional.
All these suggest that leaders who are genuine about pursuing conflict settlement have a vested interest in compliance once they commit to a ceasefire . Major ceasefire violations threaten to undermine their ceasefire objectives, by casting doubt on their good faith, indicating command and control issues, decreasing public support of an agreement, or leading to distrust from the opponent ( Höglund 2011 ;Åkebo 2013. Figure 1 visualizes the theorized relationship between the assessment of the bargaining situation and vested interest in ceasefire compliance. Independent of their assessment of the overall bargaining situation, conflict party leaders may decide to engage in a ceasefire or refuse to do so, but the logic shaping this decision differs greatly when they prefer a military outcome over conflict settlement or vice versa. Once leaders who prefer conflict settlement engage in a ceasefire, they have a vested interest in ceasefire compliance, as ceasefire violations undermine rather than further their bargaining position and the chances of settlement. This is not the case when leaders pursue a military outcome of the conflict. Note that the assessment of a bargaining situation usually changes over time, as leaders learn new information about each other, or context factors shift. Parties may update preferences incrementally or experimentally, constantly reassessing what approach may be most beneficial. Importantly, shifts in perception may occur as a result of major or repeated ceasefire violations; they may also reflect internal calculations (pressure by factions or internal challenges), failure to progress in peace negotiations, or changes in context, such as increased external pressure.

No Interest in Compliance
A first reason for ceasefire violations thus relates to the ceasefire logic: if leaders engage in a ceasefire not to calm a situation or negotiate a settlement, but to pursue a military advantage, they have no vested interest in ceasefire compliance. Indeed, leaders may engage in a ceasefire precisely with the intention to violate it, either in letter (violating parts of a formal agreement) or in spirit (engaging in activities that go against the stated objective of an arrangement). Announcing and then defecting from a ceasefire arrangement entail potential audience costs, so those pursuing such objectives have to carefully weigh the benefits and costs of engaging in a ceasefire when intending to defect ( Fortna 2003 , 343). They may still decide to do so, if the advantages offered by defection override the audience costs that may occur, or if they believe they can maintain a pretense of compliance while violating an agreement whenever it suits them militarily. In some cases, leaders may be pressured into a ceasefire arrangement by third-party actors. In these cases, it may be "cheaper" to sign onto an agreement and not (properly) implement it, rather than bearing the costs of refusal (see Nathan and Sethi 2020 ).
In sum, a key reason for ceasefire violations to occur is because leaders are not willing to comply with an arrangement. This may appear almost tautological, but it is important to include in the analysis, as it is distinct from other types of reasons for violations as outlined below. As an empirical example, in the conflict in eastern Ukraine prior to the Russian invasion, a ceasefire was in place, but neither side appeared committed to it. Ceasefire violations often went up when political tensions between Ukraine and Russia increased, suggesting that the surge in violations was used as a signaling instrument rather than outside the control of the conflict party leadership. 3 Another example are air strikes by the Sri Lankan government during a ceasefire with the Liberation Tigers of Tamil Eelam (see Reliefweb 2006 ). Absent major problems in chain of command, it seems implausible that an air offensive would happen without the orders or knowledge of those in charge.

Vested Interest in Compliance
A second reason for ceasefire violations to occur is the precise opposite: leaders may decide to violate an arrangement to ensure the continuation of a ceasefire arrangement. This occurs primarily in response to a (real or perceived) ceasefire violation by the opponent.
Leaders respond to ceasefire violations by an opponent in a number of ways, including by ignoring it, by delegating a response to a dedicated body (such as a joint ceasefire committee), by retaliating, or by suspending the ceasefire and potentially other consensus-based arrangements, such as peace negotiations. Retaliation may be motivated by different reasons and take different forms. There are two distinct reasons why leaders may want to use retaliations to ensure a continuation of a ceasefire. First, they may want to avoid exploitative behavior of an opponent in the future, as an appropriate response disincentivizes further violations (see Wiehler 2021 ). More broadly, a conflict party leadership may seek to build up a "history of retaliation" that helps prevent future violations, including by actors that are not currently part of a specific ceasefire arrangement ( Long 2014 ). Second, leaders may engage in retaliatory action not to condition a response of opponents but to appease members of their own conflict party, who might read a non-response as a sign of weakness or leniency toward the opponent. In this latter case, a decisive response by conflict party leaders may prevent members taking the matter into their own hands, and thus mitigate the risk of spoiling behavior.
It is important to note that violence often increases grievances, and leaders who clearly prefer settlement may want to avoid violent retaliation that could trigger further escalatory dynamics. They may often prefer alternative responses and will seek to carefully balance internal calculations (such as pressure to retaliate) against the risk of involuntary escalation.
In sum, a second-perhaps counterintuitive and often reluctantly pursuedreason to engage in ceasefire violations is to increase the chances that a ceasefire holds. One example of this is a series of targeted shelling, attacks, and air strikes by the Philippine Government and the MILF in the days following the Al Barka ceasefire violation, as discussed in more detail in the case study below.

Internal Divisions
Conflict party leaders are not the only strategic actors in conflicts: commanders at various levels and lower ranks may have their own preferences and assessments and should not be treated as blind followers of their leaders ( Walter 2017 , 11). They may seek to undermine the efforts of a conflict party leadership for either of two reasons. First, they may disagree with a decision taken by a conflict party leadership because their assessment of the strategic situation diverges from the assessment of the leadership. Second, they may act opportunistically, for example, if they realize that a ceasefire or negotiation efforts are unpopular with certain constituents, in order to sideline or undermine a conflict party leadership. In both of these situations, they have an incentive to derail the efforts of the conflict party leadership through spoiling behavior (see Stedman 1997 ;Zahar 2008 ;Nilsson and Söderberg Kovacs 2011 ).
Spoiling may occur through violent or nonviolent acts ( Nilsson and Söderberg Kovacs 2011 , 610-12). In the context of ceasefires, violent acts refer to the use of conflict violence during the ceasefire, while nonviolent acts may include any other prohibited act (such as recruitment) or noncompliance with prescribed activities (such as refusal to withdraw from a specified area). While violent acts of spoiling are largely ineffective as long as conflict violence is left unaddressed, they become highly visible once parties engage in a ceasefire ( Sticher 2022 , 156). Violent acts of spoiling are thus particularly pertinent in the context of ceasefires (see also Kolås 2011 , 790). Moreover, once conflict party leaders have a vested interest in ceasefire compliance, violent acts of spoiling become particularly disruptive. This suggests that the incentive of potential challengers or lower ranks to engage in ceasefire violations is highest once leaders view a settlement of a conflict as their preferred outcome.
Disagreement or political opportunism is not sufficient for such spoiling behavior to take place. To constitute a ceasefire violation, individuals or factions also need to act on their (genuine or stated) disagreement and engage in an act that violates the ceasefire. A leader with strong leadership skills and a high level of command and control may be able to avoid spoiling even in the presence of disagreeing members. Yet, certain circumstances-such as nonreciprocity of a unilateral ceasefire by the opponent-may make it difficult even for a popular leader to keep factions and lower ranks on board over a prolonged period of time.
In sum, a third reason for ceasefire violations to occur is spoiling behavior due to internal divisions. An example of this were the repeated ceasefire violations committed by the 105th Base Command of the MILF, led by Ameril Umbra Kato, during the ceasefire between the Philippine Government and the MILF (see, e.g., ICG 2008 ). Kato eventually formed a splinter group that remained outside the Bangsamoro peace process.

Defections Unrelated to Leadership Decisions
The first three reasons for ceasefire violations are linked to leaders' strategic decision-making processes and committed deliberately, either by the conflict party leadership or by individuals seeking to undermine the efforts of their leader. A fourth reason for ceasefire violations are events unrelated to leadership decisions. There are at least three types of situations in which such violations occur.
First, some individuals may not be aware of a ceasefire and commit acts as part of the "normal" hostilities of armed conflicts. Communication may be an issue, particularly for a guerrilla movement that is stationed in hard-to-access areas of a country. For example, when members of the Revolutionary Armed Forces of Colombia (FARC) targeted energy infrastructure during a self-declared unilateral FARC ceasefire, the leadership stated that the orders had not reached the troops in time ( Americas Quarterly 2012 ).
In some cases, lower ranks may be aware of a ceasefire, but not its exact content, and commit acts that are specifically prohibited in an agreement without the intention to violate it. For example, a low-level commander may understand that a ceasefire is in place and stops launching attacks, instead focusing on recruitmentwhich may also be specifically prohibited in an agreement. Another example is a commander of a mobile unit who moves her unit to stage attacks outside a declared ceasefire area, which constitutes a ceasefire violation if such movement is explicitly prohibited in the ceasefire agreement (as was the case in the Nuba Mountains ceasefire).
Second, lower ranks may act in what they believe to be self-defense. Such actions are here treated differently from retaliations, as they are an "immediate, in-crisis response" rather than a calculated response to a perceived threat ( Long 2014 , 3). Some ceasefire agreements, such as the ceasefire in the Bangsamoro peace process, specifically allow for defensive acts, while others do not specify this (see Forster 2019 ).
Third, ceasefire violations by lower ranks may be deliberate, but not related to the larger conflict dynamics. Such violations are often driven by local dynamics ( Happymon 2013 , 3). For example, in the conflict in eastern Ukraine-prior to the Russian invasion-violations along the contact line often increased after troop rotations, as local commanders wanted to make a statement to their counterparts on the other side. 4 Another example comes from the Bangsamoro conflict in the Philippines: local clans often engaged in violent feuds. These feuds had nothing to do with the larger conflict, but they were sometimes perceived-or developed into-ceasefire violations, as members of the opposing clans often belonged to different parties to the ceasefire (see Åkebo 2019 , 487).

Typology of Ceasefire Violations
Building on the four key reasons why ceasefire violations occur, this section proposes a typology of ceasefire violations and draws observable implications to empirically distinguish between the different types of violations.

Four Types of Violations
The theory of the previous section suggests that we distinguish between four basic types of ceasefire violations.
The first type of ceasefire violations are those ordered, or at least implicitly sanctioned, by the conflict party leadership in order to create a military advantage. They are here called strategic violations because noncompliance with the ceasefire forms part of the cost-benefit calculation of the conflict party leadership to commit to a ceasefire.
The second type of ceasefire violations are those that seek to safeguard the continuation of a ceasefire in the wake of a violation by an opponent. Leaders explicitly order this type of violations to guard against future exploitative behavior or to appease members of a conflict party who might otherwise start to challenge the strategic course of the conflict party leadership. These violations are here called retaliatory violations , as they respond to a violation by an opponent (see Long 2014 ;Wiehler 2021 ).
The third type of ceasefire violations are those that occur against the orders and the will of a conflict party leadership. These violations occur when factions or lower ranks seek to undermine the efforts of conflict party leaders to negotiate a settlement-either because they disagree with their leader's course of action or because they seek to challenge her. This type of violations are here called spoiling violations (see Stedman 1997 ;Christensen 2006Christensen , 2008Zahar 2008 ;Nilsson and Söderberg Kovacs 2011 ).
The fourth and last main type of ceasefire violations refers to violations that are not linked to leaders' strategic decisions. They may arise because troops on the ground are not aware of the ceasefire or its content, believe that they act in selfdefense, or engage in hostilities for specific reasons not related to the overall conflict dynamics. These violations are here called localized violations , as they tend to be driven by local dynamics. 5 Table 1 summarizes how the types of violations relate to strategic decisionmaking processes. The two main dynamics driving deliberate violations-leaders' perceptions of the bargaining situation and internal dynamics-are shown as the main axis of the 2 × 2 matrix. Because localized violations are delinked from strategic decision-making at the leadership level, they may occur in either bargaining situation.
In reality, the boundaries are not always so clear-cut. For example, spoiling violations may also occur in situations where leaders have no vested interest in compliance. 6 There are also some violations that do not neatly fit into one of the four violation categories. For example, non-state actors may reluctantly and counterproductively engage in ceasefire violations to protest stalling in a peace process by a 5 It is possible to further refine the typology by distinguishing between the three main reasons why localized violations occur. They are here clustered together as the focus lies on linking violations to strategic decision-making processes, and localized violations are characterized by an absence of such a link. 6 However, incentives for spoiling are highest when those acts are most visible and potentially most disruptive, which is the case once leaders prefer a negotiated settlement and have a vested interest in ceasefire compliance. government actor. Such violations may resemble retaliatory violations, in that the actor is generally interested in a functioning peace process, but they are driven by a preference for progress in negotiations over ceasefire compliance, and often reflect growing internal discontent. The typology is deliberately kept simple to guide the identification of causal conditions and strategic interests, while hoping to provoke a discussion about nuances and context-specific manifestations.

Observable Implications
Empirically distinguishing between different types of ceasefire violations is challenging, as actors who deliberately engage in ceasefire violations often have an incentive to misrepresent their intention and events on the ground (see Fearon 1995 ;Richmond 1998 ). This means that it is difficult to assess whether an actor pursues a military approach or prefers a negotiated settlement. The same dynamics also make it difficult to determine if a ceasefire violation was ordered or sanctioned by a conflict party leadership or took place against its desire. I use a Bayesian approach to address this challenge. The idea behind such an approach-as described in more detail in the case study-is that while any individual observation may not be necessary or sufficient to determine a type of violation, certain observations are more likely for specific types of violations than for others. If we look at many observations together, the "body of evidence" points to a certain direction, increasing or decreasing our confidence that a specific violation has indeed occurred ( George and Bennett 2005 , 189-92;Beach and Pedersen 2016 , 201-13).
The following subsections therefore discuss characteristics in terms of probabilities, showing what observations are more likely for one type of violation than for others. I divide observations into three categories-the characteristics of the conflict context, characteristics of violations, and responses of the violating and violated parties-and note how incentives to misrepresent information may decrease our confidence related to some information compared to others. Some observations are more pertinent for major violations, as these tend to be impossible to politically ignore. For smaller violations, applying these insights may require assessing several violations together, and looking for patterns rather than determining the type of a single incident.

Context in Which a Violation Takes Place
In situations where leaders prefer a military outcome over a negotiated settlement, we expect strategic violations to be prevalent, whereas we expect retaliatory and spoiling violations to be more prevalent when leaders prefer a negotiated settlement. We can thus learn something about the potential type of a ceasefire violation by assessing the bargaining context in which it took place.
Assessing the nature of fighting in the lead-up to a ceasefire can reveal important information about the bargaining context. Fighting may be used when leaders pursue a military outcome and when they prefer settlement, but the nature of violence likely differs. As long as parties pursue a military outcome, they care primarily about crushing the capabilities and resolve of the opponent and may use brute force if this suits their purpose. Parties may also try to force concessions through the use of force, but more often, we see more targeted violence during negotiations, as actors want to demonstrate strength while keeping the doors open for an eventual agreement (see Sticher and Vukovi ć 2021 , 9). The use of targeted versus indiscriminate violence can therefore offer a possible indication regarding settlement preference.
Another indication of settlement preference are costly signals to express commitment to peace talks. For example, leaders who accept preconditions that are unpopular with their constituents send a strong signal that they are genuinely interested in exploring negotiations ( Kaplow 2016 ).
Third-party pressure may provide further information with regard to the preferred outcome. As argued above, accepting a ceasefire and not following through entail third-party audience costs. It is more likely that actors without a genuine interest in settlement accept a ceasefire under strong third-party pressure than without such pressure (see also Beardsley 2008 ).
A ceasefire that is unilateral can provide further information regarding the strategic interests at play. Leaders who pursue a military outcome engage in a ceasefire with a hidden agenda, for example, using a ceasefire to re-group, re-arm, or reorganize ( Crocker, Hampson, and Aall 2004 , 158;Toft 2010 , 15;Chounet-Cambas 2011 , 7-8, 20). These objectives can only be achieved when an arrangement is reciprocal. Leaders who prefer a military outcome may still engage in a unilateral ceasefire, in the hope that the opponent follows suit. However, they are not likely to follow through with such an arrangement if the opponent fails to replicate. A unilateral ceasefire that holds without being replicated is thus an indication that a leader sees the benefits of a negotiated settlement ( Sticher and Vukovi ć 2021 , 12).

Characteristics of the Violation
The location of an incident offers potential information about the probability of a specific type of violation. An incident that occurs in an area known to be under the command of aspiring leaders, or where factions have a vested interest in the continuation of a conflict, is more likely to have resulted from spoiling than violations in other parts of a country. In areas where multiple armed groups are present, the risk of localized violations is higher compared to areas without such presence. Similarly, communication problems (e.g., in hard-to-access conflict areas) increase the risk of localized violations.
In some geographical areas, there may be local incentives to engage in acts that constitute localized ceasefire violations. For example, troops placed in border areas are more prone to engage in illicit cross-border activities. Such activities may constitute ceasefire violations, but they are not necessarily linked to wider conflict issues or designed to undermine the efforts of the conflict party leadership.
The timing of a violation may provide further information. The chances of an involuntary defection due to late or incomplete information decrease over time, as lower ranks are more likely to become familiarized with a ceasefire arrangement and its content ( Åkebo 2016 , 42;Pinaud 2021 , 476). If a ceasefire is linked to ongoing peace talks, the incentive for spoiling may be heightened if negotiating delegations appear close to a major breakthrough.
Retaliatory violations always occur in direct response to another violation, and we should find evidence of a corresponding violation by the opponent, as leaders will refer to these violations to justify their own actions. However, just because a violation occurs in response to an opponent's violation, it does not necessarily imply that it is a retaliatory violation. A leader may use any (perceived) violation as an excuse to score military points or escalate back to open conflict. If there is evidence of a related ceasefire violation, we may assess the appropriateness of the violation (see Wiehler 2021 ). We may also assess whether there was constituent pressure to retaliate and signs that leaders appeared to engage reluctantly. Leaders with an interest in compliance seek to engage in appropriate responses, while a disproportionate reaction may be designed to lead to conflict escalation and is therefore indicative of a strategic violation.
The number of casualties on both sides provides information on whether a violation was premediated or resulting from an opportunity-driven violation or a local clash. Premediated violations tend to involve a more unbalanced death toll, because the party that is attacking has the advantage of being able to plan and use elements of surprise in their attacks. Retaliatory violations, as defined above, are planned in advance to send a very targeted message. They thus tend to involve an unbalanced death toll. Strategic violations, spoiling violations, and localized violations may also be planned in advance and highly targeted, but they may also be more opportunity driven (or in the case of localized violations, arise from clashes and self-defense). This means that on average, retaliatory violations are more likely to involve an unbalanced death toll, although an unbalanced death toll in and of itself in no way suggests that a retaliatory violation took place.

Responses to Violations
Statements and actions of both conflict parties in the aftermath of an incident provide further information about the type of violation that may have taken place.

Response by Violating Party
Those involved in a ceasefire violation will often seek to shift blame to other actors or to the circumstances (see, e.g., Wiehler 2021 , 425). However, a conflict party leadership may be more likely to admit wrongdoings on their own side in some situations than in others. In order to interpret statements by conflict parties, we need to understand incentives for presenting true motives or misrepresenting intent and objectives.
A persistent challenge for those involved in a peace process is to distinguish between strategic and spoiling violations. Conflict party leaders who seek to exploit cooperative behavior by an opponent have an incentive to misrepresent their intention: they will not likely admit that they ordered or sanctioned a ceasefire violation (see also Richmond 1998 ). They may instead attribute such violations to spoiling behavior or blame local dynamics or actors who are not part of the ceasefire arrangement.
Those committing spoiling violations similarly have an incentive to misrepresent the nature of a transgression: their acts will be most effective in undermining efforts by their leadership if the opponent believes that the leadership acted in bad faith and intentionally violated an agreement. However, aspiring leaders that seek to undermine a conflict party leadership may also declare their disagreement publicly or assume responsibility for attacks.
Leaders facing spoiling violations within their own ranks will likely blame spoilers. The chances of finding such statements for spoiling violations are higher than for other types of violations, although of course, leaders may also blame spoilers for strategic violations, to avoid third-party audience costs.
In the case of retaliatory actions, leaders have an interest in taking responsibility for such a violation, for a number of reasons: to ensure that the message is received by the opponent, to signal to the own constituents that violations are dealt with appropriately, and to avoid or lower third-party audience costs.
For localized violations, leaders have no incentives to misrepresent violations, and the actors involved in a violation may even admit their wrongdoings, especially when they defected involuntarily (e.g., if they received incomplete instructions about the content of a ceasefire arrangement). Yet again, actors may claim that they were unaware of specific instructions if it serves their purpose.
Given these overall incentives, statements have more weight as evidence if actors claim responsibility for their actions compared to if they deny it.
Statements that are followed up by concrete actions are more reliable indicators of the type of violation that has likely been committed. If a leader calls for an investigation of a ceasefire violation and cooperates in such an investigation-by granting third-party observers access to the location and witnesses, providing important context information, etc.-such a statement is much more credible.
Another action that lends credence to a leadership's commitment to a ceasefire is addressing ceasefire violations within their own ranks. We should not expect leaders to do so for retaliatory violations, as these are explicitly ordered to send a signal to the opponent. For strategic violations, leaders may act opportunistically, attempting to shed blame on individuals at lower ranks to convince the opponent to continue with the ceasefire and resume cooperation. However, punishment not only sends a message to the opponent, but also to their own ranks, making clear that violations will not be implicitly tolerated. It is therefore more likely that we see punishment of own troops for spoiling and localized violations than for strategic violations, especially when we look at a pattern of behavior over time.

Responses by the Violated Party
Broadly speaking, a leader may respond to a violation in several ways: by not responding at all or in a symbolic manner, by calling for an investigation by a thirdparty or ceasefire mechanism, through retaliation, escalation, or by suspending the ceasefire and other consensus-based arrangements, such as peace negotiations. These responses are not mutually exclusive and may be combined.
Evidence at this unit of observation may be unreliable for at least two reasons. First, the response to a ceasefire violation may be shaped by leaders' own interests and their assessment of the bargaining situation, rather than by their interpretation of the violation. As discussed above, leaders who pursue a military outcome may use ceasefire violations by an opponent as an excuse to escalate back to violent conflict or to launch counterattacks while averting the blame. They may also directly call the end of a ceasefire or peace negotiations, not because they necessarily doubt the good faith of the opponent, but because they had no interest in moving toward a negotiated settlement in the first place. These responses are informed primarily by their desire to gain militarily, rather than their assessment of the violation or a desire to condition the opponent's behavior.
Second, even if they want to respond in a manner appropriate to the violation, imperfect information may make it difficult for leaders to correctly interpret an event. This problem is further accentuated by the fact that leaders may be expected to respond within a short time period, especially when a violation triggers a public outcry.
The interpretation of a ceasefire violation may also be affected by cognitive biases, such as the confirmation bias (see Nickerson 1998 ;Kelman 2007 ). Leaders who already established confidence in the good faith of an opponent are more likely to give the opponent the benefit of the doubt and attribute ceasefire violations to spoiling, compared to leaders who enter a ceasefire with great uncertainty about an opponent's intent.
In short, a response to a ceasefire violation is shaped by bargaining assessments and interpretations of an incident, rather than directly by the type of violation that occurred. That said, we can still learn something from observing the reaction by the violated party. Efforts by the violated conflict party to contain the escalation of a conflict through collaboration with the opponent are a strong indication that a conflict party leadership sees the violation as spoiling, localized, or retaliatory and not as strategic. A statement of continued commitment to a peace process in the aftermath of a major violation by an opponent similarly suggests that the violated party does not interpret an incident as a strategic violation. Finally, all else equal, we are more likely to see an escalation of conflict violence, a suspension of a ceasefire, or a suspension of peace talks if a strategic violation indeed occurred, although all these responses could be shaped by other factors.

The Al-Barka Ceasefire Violation
A case study example illustrates how we may apply the theoretical insights to assess the type of ceasefire violation that was committed. It focuses on the peace talks between the Philippine Government and the non-state actor MILF that started in 1997 and culminated in a comprehensive peace agreement in March 2014. As a violent conflict between a state and a non-state actor pursuing a political goal (establishing an independent Islamic state) with a ceasefire in place before the conclusion of a peace agreement, the case matches the scope conditions of the theory.
The case study focuses on a major ceasefire violation in October 2011. The specific violation was chosen for multiple reasons. It followed a long period of relative calm and involved a high number of casualties, putting considerable stress on the ongoing peace talks. From an analytical perspective, it is particularly interesting to study major ceasefire violations rather than minor skirmishes, as large-scale violations are more likely to lead to a critical junction in a peace process. In addition, due to the high number of casualties involved, the violation was widely reported on, making it more likely that evidence of specific factors is found, if these factors indeed exist.

Method Description
The case illustration uses a case study approach underpinned by informal, narrativebased Bayesian reasoning (see Beach and Pedersen 2016 , 154-226;Fairfield and Charman 2017 , 364). Bayesian logic proposes that we start with an initial confidence about a hypothesis being true, and systematically update this confidence by assessing a collective body of evidence. Such an approach is well-suited to distinguish between different types of ceasefire violations, both from a theoretical 7 and from an empirical perspective. 8 In and of itself, the existence or nonexistence of a particular piece of evidence does not confirm or disconfirm that a specific type of violation occurred, but when taken together, they form a collective body of evidence that points into the direction of a specific violation type ( Beach and Pedersen 2016 , 201-13). Table 2 translates the implications discussed in the previous section into a format that can be used for informal Bayesian updating. What is of primary interest to us is an assessment of the probability of finding a specific type of evidence for a specific type of violation, compared to the probability of finding that piece of evidence for any other type of violation. For each type of evidence, table 2 notes (in the columns corresponding to different ceasefire violation types) if the probability of finding that evidence is higher or lower compared to the likelihood of finding it for other ceasefire violations. For example, we are more likely to find evidence that a violation "occurs in direct response to a violation" if a ceasefire violation is retaliatory, compared to the chances of finding such evidence for other types of violations. Leaders that engage in strategic violations may have an incentive to claim that they act in direct response to an opponent's violation, but a strategic violation in no way requires an opponent's violation. Thus, without any prior knowledge of the circumstances, we are more likely to find that a violation occurred in direct response if it is indeed retaliatory. Importantly, not each type of evidence has the same effect on our confidence that a violation is a particular type of violation (see Beach and Pedersen 2016 , 154-226;Fairfield and Charman 2017 , 368-72 for comprehensive discussions). The so-called inferential strength of a piece of evidence depends on a number of factors, most importantly how unique it is to a specific violation. For example, the combination "occurs in direct response" and "appears to be an appropriate response" is relatively unique to retaliatory violations, as leaders who use an opponent's violation as an excuse for a strategic violation will likely seek to escalate or create a military advantage through their response (see Wiehler 2021 ). Accordingly, finding such a combination of evidence has high inferential strength. Other factors that shape how strongly we can update our confidence is how similar a piece of evidence is compared to evidence that has already been evaluated (with similar evidence having less impact on updating our confidence) and how confident we are in the accuracy of the evidence.
The next subsection discusses and assesses all the potential evidence listed in table 2 , and uses this collective body of evidence to update our confidence that the ceasefire violation under question constituted a specific type of violation. Sources used to assess the evidence include English-language news articles, published statements by the conflict parties, and think-thank reports, demonstrating how we may apply the framework with limited information (and in particular, without knowledge of the intent of conflict parties or orders given by a command structure).

Evidence for the Al-Barka Incident
The peace process between the Philippine Government and the MILF started in early 1997. In July 1997, the conflict parties reached an agreement on a ceasefire. The ceasefire collapsed at three moments of the peace process-in 2000, 2003, and 2008-but was always reinstated, remaining one of the cornerstones of the peace process ( Åkebo 2019 ). The Al-Barka incident took place on October 18, 2011, on the Basilan island of Southern Philippines, when members of the armed forces clashed with the MILF in a known MILF area ( ICG 2012 , 10-11). It later emerged that the armed forces had entered the area without prior coordination through the ceasefire mechanism, as would have been required ( ICG 2012 , 10). Twenty-five people died as a result of an hour-long battle, making it by far the largest incident since the ceasefire reinstatement in 2009.
Before assessing the evidence in the case illustration, we need to clarify the hypotheses and our initial confidence in them. There are four distinct types of ceasefire violations, so we have four hypotheses, each being that an incident constitutes a specific type of ceasefire violation. Without any prior knowledge of the circumstances in which a violation took place, we have no reason to believe that an incident constitutes a specific type of violation. Our prior confidence in any particular of the four hypotheses being true is therefore the same and relatively low, around 25 percent.

Context Factors
To update our initial low confidence in the four hypotheses being true, we first assess the different types of evidence listed under context.
After a surge in hostilities following the ceasefire collapse in 2008, there were only a limited number of deadly clashes between the conflict parties in the months leading up to the ceasefire arrangement, and none with a high number of civilian casualties (the strongest indicator of an indiscriminate use of violence). Once the ceasefire became effective, compliance was generally high, with five battle-related deaths recorded by the Uppsala Conflict Data Program between its reinstatement in July 2009 and the October 18, 2011, Al-Barka incident ( Sundberg and Melander 2013 ;Högbladh 2019 ).
In August 2011, shortly before the ceasefire violation, President Aquino met with the MILF chairperson in Kuala Lumpur. The first such high-level meeting since the beginning of the peace process in 1997, this was seen as a signal of commitment to the talks. Although it was not a particularly costly signal, the meeting indicated that leaders were willing to bear some ex post audience costs should they not be able to deliver on the peace talks (see Fearon 1997 ;Kaplow 2016 ).
Third-party involvement in the process gradually increased over the course of the peace process, but there is no evidence that the ceasefire was imposed by a third party (see, e.g., Coronel Ferrer, Evangelista, and Åkebo 2022 ). The ceasefire was not unilateral (which might have provided more ground for spoiling).
The collective body of evidence related to context factors decreases our confidence that the violation was strategic and increases our confidence that it was retaliatory. Supporting and contradicting evidence for a spoiling violation level out. Our confidence that this was a localized violation remains unaffected by the context assessment, as this type of violations is not linked to strategic decision-making processes at the elite level.

Ceasefire Violation Characteristics
The location and timing of the incident provide further information about the type of violation that might have occurred. Al-Barka is a municipality on Basilan, an island of the Sulu Archipelago. In a 2012 report, the International Crisis Group identified local politics in this archipelago as a potential spoiler for the peace process. The report identified the Basilan governor among key figures with a vested interest against a peace agreement (although making clear that the main challenge would come from the Sulu governor) ( ICG 2012 , i-ii). There is thus evidence that the area is home to potential spoilers on the government side, although not from within military ranks. 9 Besides the presence of the MILF, Basilan is home to another armed group, Abu Sayyaf, and to a small number of foreign jihadists. In addition, there is an ongoing rivalry between local clans based in Basilan, which increases the risk of localized violations ( ICG 2012 , i).
There is no reason to believe that communication with the armed forces stationed in Basilan was more difficult than across the archipelago (see, e.g., Philstar Global 2005 ). The ceasefire had been continuously in place for more than two years, increasing the chances that information about the ceasefire mechanism was widely known. According to a statement by a military officer, coordination between belligerents in the area often happened in an informal way, rather than through the official ceasefire mechanism ( ICG 2012 , 10). Such a statement suggests that while they may not have followed the official mechanism, they were aware of the need to coordinate their actions in this area.
The location of the incident contains an additional, important piece of information. It was in the same municipality that in 2007, MILF members ambushed a marine convoy, killing at least fourteen soldiers. 10 The incident had been met with a strong public outcry, exacerbated by the fact that ten of the soldiers were beheaded. We should not interpret the 2011 Al-Barka incident as a direct response to the incident: more than four years had passed, and the Philippine government did not declare the Al-Barka incident as a retaliation for the 2007 ambush (which we would expect, if it was intended as a retaliatory violation). However, the 2007 ambush provided local incentives for the violation: several individuals in Al-Barka had bounties 9 The report identified other potential spoilers-Christian settlers, conservative nationalists, and some members of another insurgent group in the South-but did not refer to factions within the military. 10 See ICG (2012 , 8). Six MILF members and an imam were also killed. Entities of the ceasefire monitoring architecture complete investigation and release confidential report Armed forces chief of staff recommends court martial proceedings against four officers March 2012 MILF issues statement that it had sanctioned three commanders implicated in the incident April 2012 Armed forces form general court martials 2012-2013 Of the four officers that were court-martialed, two are found guilty and convicted, while two are acquitted on their heads, including the MILF commander targeted by the soldiers ( ICG 2012 , 10; Mendoza 2012 ). There is some disagreement about the original targets of the special forces and the role bounties played in the incident, but "many . . . believe that the commanders involved were motivated by the rewards on the targets' heads" ( ICG 2012 , 10; see also ABS- CBN News 2011 ;Elusfa 2011 ;Santos 2012 ).
There is no evidence that a breakthrough in peace talks was expected, which might have heightened the risk of spoiling. 11 Curiously, the death toll of the incident was highly imbalanced, but not in the direction we might expect. The main violating party, the government, suffered nineteen battle-related deaths, whereas the MILF lost five or six members, depending on the source. 12 Several factors appear to have contributed to this devastating toll from the attacking force. The operation appeared to have been undertaken at short notice ( ICG 2012 , 10), and many of the special forces deployed were trainees without the necessary combat experiences ( Mendoza 2012 ;Santos 2012 ). The number of MILF members swelled over the course of the hour-long battle ( Malig 2011 ;Dida 2012 ). No reports were found of casualties with other armed group identities, but there is evidence that members of Abu Sayyaf joined the MILF during the battle ( ICG 2012 , 11;Mendoza 2012 ).
Adding to the information we have about the context, the evidence about ceasefire characteristics leaves us with a very low confidence that this was a strategic or a retaliatory violation. It increases our confidence that this was a localized violation, while the estimated probability of a spoiling violation remains at a low level.
Responses Table 3 provides an overview of key events following the Al-Barka ceasefire violation. In the aftermath of the incident, hostilities re-erupted for a limited period of time. On October 19, the day after the original violation, the government shelled an MILF base on the main island of Mindanao, and on October 20, the MILF retaliated by attacking military and police convoys. On October 24 and 25, the government used air strikes against the MILF-a first since the reinstatement of the ceasefire ( Associated Press 2011 ;Calica 2011 ;ICG 2011 ). Thousands of villagers were displaced and at least ten security service members and six members of the MILF were killed during these follow-up incidents.
While this is undoubtedly a high death toll, it appears that both sides were still somewhat restrained, and the response designed to be "appropriate" to the original violation (which had led to twenty-five battle-related deaths). Leaders on both sides stated their commitment to contain an escalation of the conflict. Despite high pressure from the public (and even his own military) to return to war, the president stated his continued commitment to the peace talks and even criticized the military ( Associated Press 2011 ; ICG 2012 ; Mendoza 2012 ). On the MILF side, a spokesperson told the media that the MILF "leadership [was] working very hard to prevent this from escalating" and ordered their fighters "to stay put" ( Associated Press 2011 ).
Such statements in isolation do not constitute strong evidence, especially when they align with incentives to avoid blame. However, actions undertaken by both sides aligned with their rhetoric. Less than three weeks after the Al-Barka incident, both sides met in Kuala Lumpur to continue the peace talks. Both sides called for an investigation, which was carried out by numerous entities of the ceasefire monitoring architecture (see Junio 2012 , 34-35). Importantly, both sides initiated disciplinary action against their own members who were involved in the incident ( Mendoza 2012 ). The armed forces court martialed four of its own members and the MILF temporarily suspended three of its members, sending a strong signal that they were serious about violations within their own ranks. 13 No evidence was found that the government called the incident an act of spoiling.
In sum, evidence related to both sides' responses to the Al-Barka incident further decreases our confidence that this was a strategic violation and increases our confidence of a localized violation, while again, the confirming and contradicting evidence for a spoiling violation levels out.

Assessment
The collective body of evidence points into the direction of a localized violation. For each cluster of evidence-the context in which the incident took place, ceasefire characteristics, and responses-there are pieces of evidence that suggest that a localized violation may be at play, whereas there is no strong contradicting evidence. We can therefore be reasonably confident that this was indeed a localized violation. Similarly, we can quite confidently state that this was not a strategic or retaliatory violation. The possibility that this was a spoiling violation appears somewhat higher, although not as high as for a localized violation.
This assessment helps explain why the Al-Barka incident, which led to a short resumption of hostilities and considerable stress on the peace talks, did not lead to the breakdown of the ceasefire or a collapse of the peace talks. Of course, localized violations are no guarantee for violence containment. A high degree of confidence in each other's commitment to the talks and in the ceasefire monitoring mechanism certainly made it more likely that the parties contained a fall-out from such a major incident (see also Åkebo 2019 ).
In some situations, and for some purposes, it may not be necessary to evaluate all available evidence to gain a good sense of the type of violation that has likely taken place. For example, even with the relatively limited information we have about the violations that followed the Al-Barka incident, we can be relatively confident that these were retaliatory violations, as they occurred in direct response to a ceasefire violation and appeared designed as an appropriate response to the original 13 That both sides punished members involved in the incident indicates that while the government was the main violating party, the MILF had also engaged in violations. While the official report on the violation is classified, the MILF stated that it had been cited for one ceasefire violation, while the government had been cited for ten ( ICG 2012 , 10-11). violation-a combination that is relatively unique to retaliatory violations and has high inferential strength.

Conclusion
Major ceasefire violations represent a critical junction in peace processes. Building on bargaining theory, ceasefire research, and deductive reasoning, this article theorizes why violations occur and how they relate to the wider military and political calculations of conflict parties. Treating ceasefire violations as part of larger conflict dynamics allows us to observe and learn from them in a way that can inform both theory and practice.

Implications for Research
This article fills two important gaps in ceasefire research. First, by linking ceasefire violations to strategic decision-making processes, it offers a framework to identify causal conditions and strategic interests that can lead to ceasefire violations. Second, based on this framework, it proposes a typology of four main types of ceasefire violations. This conceptual contribution enables cross-case comparison and allows ceasefire scholars to identify clear scope conditions for their theories. By focusing on causal conditions and strategic interests, this article also helps explain why we see such varied responses to major ceasefire violations. Future research could study these responses more systematically, disentangling the role of strategic interests, ceasefire violation type, and perceptions and cognitive biases in interpreting conflict events.
Beyond its contribution to ceasefire research, this article speaks more broadly to a key uncertainty in armed conflicts: the question of intent. One of the most persistent challenges for actors involved in a peace process is gauging the intent of their opponent. Scholars struggle to operationalize intent, even though it plays a key role in many theoretical arguments. This article provides a concrete example of how we may assess the strategic interests of actors in conflict-and relatedly, their intentacknowledging the methodological challenges this entails. The hope is to provoke a debate on how we may infer intent from actions in the battlefield, including ceasefire violations. The above discussions suggest that if the collective body of evidence related to ceasefire violations repeatedly points toward strategic violations, a leadership is not seriously committed to peace talks and continues to pursue a military outcome. 14 If other types of violations are prevalent, and the likelihood of strategic violations seems low, we can be more confident that leaders are committed to a negotiated way out of a conflict.
This article is primarily concerned with theory-building. Future research could test and refine the theory, in particular with regard to the observable implications. To complement the deductive approach used in this article, such endeavors should ideally be empirically grounded. Interviews with conflict actors reflecting on past violations, on their own decision-making processes, and on possible manifestations of different types of violations in their own and other contexts could be one way to approach such research.

Implications for Policy and Practice
For policymakers, understanding how different types of violations correspond to different strategic situations helps identify what type of third-party engagement is most effective in a specific conflict context. Fortna famously identified three main mechanisms for addressing ceasefire violations: increasing the costs of attack, reducing uncertainties about intent and behavior, and containing escalatory dynamics ( Fortna 2003( Fortna , 2004. This article suggests that some mechanisms are more effective for some types of violations than for others. Increasing the costs of violations alters cost-benefit calculations at the leadership level and will therefore be more effective in reducing strategic and retaliatory violations than spoiling or localized violations. Conversely, reducing uncertainty will not likely prevent strategic violations, as these are driven by preferences for a military outcome rather than uncertainties about the opponent's intent (although the two can be related). Similarly, third-party measures to contain escalatory dynamics are designed for involuntary escalation, and therefore will not likely be effective for strategic violations. Policymakers should be sensitive to these dynamics and adopt measures that address the type of violation that is most prevalent in a conflict context.
Practitioners may adapt some of the insights of this article for use in dialogue and ceasefire capacity-building workshops. One challenge that conflict actors face is how to respond to an opponent's violation, especially in situations where they are committed to a ceasefire but uncertain about the opponent's intent. Cognitive biases make it more likely that a violation is interpreted as strategic, even when other dynamics are at play ( Kelman 2007 ). Third-party experts could introduce the ceasefire violation typology in such settings-not to replace context-specific terminology but to sensitize actors to the different strategic interests that may apply. This also presents an opportunity for actors to reflect on empirical manifestations of different types of violations in their own context, anticipate related challenges, and discuss measures to address them.