Conceptualizing Civil War Complexity

Abstract Civil wars that appear to observers to be the most complex—even using a colloquial understanding of the concept—are also those that seem to register the most intense fighting, the most prolonged spells of war, and the most resistance to durable conflict resolution. But what does it really mean for a civil war to be complex? We currently lack a concept of “civil war complexity” that can help us better understand the most important variations in civil wars across time and space. To address this gap we develop a conceptualization of “civil war complexity” consisting of three dimensions—“actor complexity,” “behavior complexity,” and “issue complexity”—and demonstrate how they manifest empirically. We also highlight this conceptualization’s utility—and the danger of overlooking it—through the case of Darfur. This conceptualization paves the way for a new research agenda that explores how civil wars differ in terms of their complexity, the causes and consequences of civil war complexity, and how to refine conflict resolution techniques and strategies.

civil war-widely perceived as one of the most complex of the last decade-displays a myriad of domestic, regional, and international actors fighting each other, with rivalries stretching from the local to the global. Allies in one conflict are belligerents in others, and side-switching is so rife it creates an intricate and shifting patchwork of dynamics. Countless issues are under dispute, such as the removal of Bashar al-Assad, the establishment of a Kurdish state, and the foundation of a caliphate. 2 Is the Syrian civil war more complex than other civil wars, such as the ones in Afghanistan or Yemen, or the historical case of the Spanish Civil War? The latter seemingly shares some of Syria's characteristics in terms of the many actors involved, rivalries from the local to the global, and shifting alliances and behavioral dynamics. Yet we currently lack a clear conceptualization of civil war complexity that would help facilitate analyses of how civil wars vary in terms of their complexity. Though it is true that all conflicts are complex to some extent, this does not entail that the degree and type of complexity are equal across cases.
Theorists of international relations have stressed the importance of complex systems for decades. Kenneth N. Waltz, for example, argued that relations between states need to be looked at as a complex system of interactions to be understood. 3 Almost twenty years later-despite the popularity of Waltz's work-Robert Jervis saw the necessity of arguing anew that our view of how politics function needed to be updated to understand and predict the social world around us, stating that "we need to take more seriously the notion that we are in a system and to look for the dynamics that are at work." 4 However, such insights have not been sufficiently integrated into the study of civil war. Some recent work demonstrates how a complexity perspective can enhance our knowledge about civil war trajectories, but the full potential of relying on such a perspective has not been realized. 5 Applying a complex systems perspective to civil wars, as well as providing a clear conceptualization of civil war complexity, should help advance civil war research. As Franciso Guti errez-San ın and Elisabeth Jean Wood have argued regarding the concept of repertoires of violence 6 and as Kristin M. Bakke and colleagues have on rebel group fragmentation, 7 how we choose to conceptualize civil war's core concepts matters greatly for how we as a discipline are able (or not) to approach it in terms of research.
We make three main contributions to the field in this article. First, we introduce a conceptualization of "civil war complexity" that makes this elusive concept more concrete. Drawing on Peter Wallensteen's definition of conflict as consisting of actors, behavior, and contradictions-joined together in the so-called conflict triangle 8 -we conceptualize "civil war complexity" as consisting of actor complexity, behavior complexity, and issue complexity. This tangible definition opens the possibility for research exploring how the degree and type of complexity fluctuates across civil wars, as well as what effects these variations may have on, for instance, human security. Second, although research has made recent advances by disaggregating civil wars-for example, by examining different forms of violence such as communal conflict or interrebel fighting-these spheres have often been studied separate from each other. This increased compartmentalization risks our findings from such ventures becoming piecemeal, with a comprehensive understanding of civil war dynamics lost, much as Jervis argued against reductionism in the study of politics more generally. 9 By synthesizing the important insights made by different disaggregation endeavors, our conceptualization facilitates a more encompassing analysis of civil wars along their most important dimensions. Third, by enhancing our understanding of the key components of civil war complexity and demonstrating how these dimensions manifest themselves empirically, we offer insights into how to understand the causes, dynamics, and resolutions of complex civil wars.
Applying a complexity perspective to civil wars yields important insights for specific research strands across the broader civil war research field. To illustrate: first, it enables new opportunities for the literature on the onset, intensity, and duration of civil wars. 10 A fruitful area for future research is to explore if civil wars with a higher degree of complexity are associated with more intense and protracted wars. On its face, complexity seems to be associated with grave human insecurity. In the past decade, a handful of civil wars have produced some 30-60 percent of global fatalities from organized violence. 11 These are wars such as those in Afghanistan, the Democratic Republic of Congo (DRC), South Sudan, and Syria: wars that are not only highly violent but also ones that from a cursory look appear to be characterized by a high degree of complexity. Future work would benefit from studying if civil wars that are more complex also are more intense or tend to drag on longer.
Second, this article also speaks to the literature on organizational cohesion and multiparty civil wars. Whereas previous work within this strand has tended to focus only on the actor or the behavioral dimension, 12 our conceptualization emphasizes three dimensions' interconnected natureactors, behaviors, and issues. By applying a complexity perspective, scholars can, for instance, explore how both local-and national-level conflict issues-what nonstate actors fight for-matter for how and when groups factionalize, merge together or become allies, and spawn as functions of their changing civil war environment. As we illustrate later in this article via the case of Darfur, complexity in one dimension can influence complexity in the other two.
Third, applying our concept opens new avenues within research on third parties, spanning not only negotiation and mediation but also third-party interventions in the form of peacekeeping. Regarding conflict resolution, by exploring variations across our three dimensions of complexity, scholars can seek to assess whether civil wars that exhibit complexity in a particular dimension require certain tailored solutions for resolution. For example, such analyses can help us understand if peace agreements that better address the different dimensions of civil war complexity are more likely to prove successful. Similarly, a complexity perspective can be useful for analyzing the specific challenges confronting peacekeeping. This literature is increasingly accounting for various facets of civil wars, providing more nuanced analyses of peacekeeping effectiveness. 13  war complexity may require certain types of responses-for instance, in terms of overall mission characteristics, as well as their activities on the ground. 14 In sum, our new conceptualization of civil war complexity is useful for several different research areas, which should help both inform and challenge our current understanding of civil wars. Next, we discuss civil war and complexity theory and then present our conceptualization. Thereafter, we unpack our framework's core components and delineate their interlinked nature. We then proceed to illustrate our concept empirically and demonstrate what insights we can gain from applying the concept as a tool for analysis. We conclude by pointing toward avenues for future studies.

Civil War and Complexity
Before we approach the main task of this article-conceptualizing civil war complexity-we need to discuss what we mean by civil war, explicate the meaning of complexity, and stipulate why it is a useful tool for capturing variations in civil wars. The main rationale for wedding civil war and complexity is how complexity theory offers analytical insights into identifying both a phenomena's most vital components and how these are interlinked and interdependent. Complexity theory thus offers a structured and systematic way of approaching and understanding civil wars.
Civil war is commonly defined as a situation where a government faces a violent challenge by one or more rebel actors over some articulated goal of political change. Several scholars have criticized such a definition's simplicity, declaring that it obscures vast variation within the phenomenon. They suggest that wars should be viewed as systems of related conflicts, containing several contested stakes linked across levels of analysis, or be seen as networks of interlinked actors. 15 However, although scholars have usefully highlighted the complexity of civil wars, they have not unpacked what complexity means in the civil war context or what precise consequences it would have and why.
A common feature of these perspectives is that they emphasize relationships and interdependencies between actors at different levels. As such, these portrayals of civil war resemble depictions of complex systems. We, therefore, turn to complexity theory to see what this strand of thought can add to our understanding of civil war.
In this article, we-in line with both the colloquial and dictionary meanings of complexity 16 -view a phenomenon as complex when it consists of many different and connected parts. Complexity theory constitutes the opposite of the reductionist trend of separating components to understand how the whole works, as in, for instance, categorizing different types of violence in a civil war rather than trying to grasp how they are all interconnected. 17 It proposes that the whole is more than the sum of its parts and views civil war as a system of interacting parts that together produce the totality of the war, rather than a series of simultaneously and independently occurring processes.
Complexity theory seeks to comprehend how systems are organized and has been applied to everything from the functioning of ant colonies, 18 to forays into conflict analysis, 19 to the inner workings of peacebuilding missions. 20 Although there is no agreed-upon definition of complex systems, there is consensus on some basic criteria: (1) all systems display collective behavior stemming from a large number of individual components; (2) these components send signals to each other; and (3) the system adapts to its environment. 21 In a complex system, a large number of components (ants, rebel actors, brain cells, etc.) interact with each other and their environment through different linkages or signaling mechanisms (ant communication, alliances and fighting, neurons), and adapt when the system changes (find new places to forage, expand territory upon defeating a rival actor, create new neural pathways if one is broken). 22 Some categorizations of complex systems are based on the interactions between components rather than such components' number. 23 The above identifies which criteria we should apply to classify something as a complex system but tells us little about how to think about degrees or levels of complexity. There is no consensus on how to measure a system's level of complexity, and suggestions range from measuring n components or levels of hierarchy (systems within systems), to network theories' focus on the number and density of interactions between components. 24 An application of basic complex systems theory to organized violence is not the place to settle that dispute. To understand variations in complexity two intuitive criteria are, however, useful: (1) the number of components that constitute the system/network; and (2) the density and type of interactions that link these components. These two criteria allow us to link components and interactions to observable phenomenon such as rebel actors or the number of conflict issues (components), as well as the degree to which actors share issues in dispute or the number of actors in multiple dyads (density and type of interaction). To apply complexity theory to civil wars thus entails, in a first step, identifying (1) its base components; and (2) how these components are interlinked.

Conceptualizing Civil War Complexity
To identify a civil war's base attributes and to meaningfully capture different degrees and types of complexity, we turn to classic conflict theory. Building on Johan Galtung's 25 conflict triangle, Wallensteen 26 defines conflict as consisting of three elements: (1) actors, (2) behaviors, and (3) contradictions/issues at stake. As such, building our conceptualization on the latter's conflict triangle brings nuances to how civil wars vary and identification of its base components. 27 Allowing civil wars to vary in the type and number of actors, behaviors, and issues thus helps define a system's parts and their interlinkages. 28 We define civil war complexity as the degree to which the core components of conflict-actors, behaviors, and issues-are manifold in number and type, and the degree to which these are interlinked. To delineate the 23  complexity of a given civil war one thus has to consider (1) actor complexity-that is, the number and type of armed actors related to it; (2) behavior complexity-that is, the number of hostile ties in the war, as well as the type of violence used (different repertoires of violent conduct); (3) issue complexity-that is, the number and type of contested issues; and (4) the interlinkages between these three dimensions-that is, the degree to which each dimension feeds into the others. Figure 1 provides a graphical illustration of our model. Next, we unpack each component of our concept both theoretically and empirically. Intertwining theory and empirics allows us to move between Giovanni Sartori's "What is?" question regarding the boundaries of our concept, and the "How much?" question regarding empirical measurement. 29 Thereafter we discuss interlinkages between the different dimensions.

Actor Complexity
Our first dimension, actor complexity, concerns the number and type of warring actors engaged in a civil war. We contend that actor complexity should not be restricted to governments and rebels but instead include other actors that partake in the civil war, such as pro-or antigovernment militias, splinter groups, and foreign parties. 30 A strong focus in previous work on civil wars has been on the number of actors and their origin(s). Stathis N. Kalyvas, for instance, argues that civil wars foster action by actors at different levels of analysis. 31 This approach implies that a core variation in civil wars rests in the number of actors involved and their respective level of origin, such as the international, national, and subnational levels. Similarly, Christopher Mitchell proposes that one way a war can grow to become more "intractable" or "protracted" is if a conflict attracts additional parties. 32 Previous studies provide important insight into how an increase in the number of actors influences civil war dynamics. David E. Cunningham, 33 for instance, demonstrates how more actors prolong the duration of civil war, 34 and Desir ee Nilsson 35 shows how the number of warring parties matters for the duration of peace. 36 Research on rebel fragmentation also supports the notion that the number of actors influences civil war outcomes and dynamics. 37 We propose that the type of actors also affects complexity. Introducing different varieties of actors seems to have effects on civil wars that differ from the addition of more warring actors per se. Edward E. Azar 38 reminds us that we need to consider external actors in internal dynamics as they alter these, for instance, by prolonging civil wars. 39 Other studies reveal that communal conflict can be tightly linked to civil war dynamics, 40 and examinations of progovernment militias demonstrate that these "extradyadic actors" 41 affect both conflict processes and outcomes. 42 How do variations in numbers and types of actors manifest themselves empirically? The two civil wars in Syria (2011-) and Egypt (1992-1998) are illustrative. The Syrian civil war is a prototype of high actor complexity, with hundreds (perhaps thousands) of domestic actors involved in fighting between the government and mainly Sunni and Kurdish oppositions. Localized community-based mobilization and opposition fragmentation both cause the proliferation of domestic actors. Syria also harbors a high number of external actors, both transnationally mobile nonstate actors (the Islamic State [IS] and Hezbollah primarily) and those from other states. In fact, Iraq, Saudi Arabia, other Gulf States, the United States, and Western allies have all been active, either directly on the battlefield or as supporters of different sides. 43 Syria thus exhibits not only an abundance of warring actors but also actors of many different types. Egypt's first war with Islamist jihadists (1992-1998) does not display the same degree of actor complexity, neither on a strict numerical basis nor in terms of the types of actors. In this war the government confronted the al-Gamaa al-Islamiyya group that strived to replace the "apostate" government with an Islamic state. Though al-Gamaa al-Islamiyya split into two factions in 1997, al-Gamaa al-Islamiyya and the government were the only directly participating actors for most of the period. There were no external interveners, nor progovernment militias or self-defense militias of note. 44 The two cases thus display stark differences in the levels of actor complexity.

Behavior Complexity
Our second dimension-behavior complexity-focuses on the number of hostile behavioral ties and the types of violent behaviors between warring actors during a civil war. The higher the former-parties engaged in fighting each other-and the more varied the types of violence, the more complex a civil war is on this dimension.
The relationships between the warring actors-their behaviors-and how these change over time constitute a fundamental part of a civil war's dynamics. Several scholars have emphasized the importance of behavioral interactions in conflict systems, 45 with network approaches in international relations arguing that not only the number of actors, but also the relationships between them, matter and can alter conflict processes. 46 Such effects have, for instance, been found in how violent rebel-rebel interactions affect civil war processes and outcomes such as settlement, 47 the risk of war recurrence, 48 and targeting of civilians. 49 Kalyvas approaches behavioral interactions and their evolution through the concepts of "alliances" versus "cleavages," 50 and Paul Staniland details how changes in wartime political orders affect the conflict process. 51 In line with this thinking, our first component of behavior complexity focuses on the amount of hostile behavioral ties that exist in a civil war, as well as fluctuations in this n component.
A consequence of the different mosaics of rivalry and alliances is that two civil wars with an equal number of actors can differ drastically in the number of hostile ties. If several nonstate actors are all allied against a government, fewer hostile ties need to be resolved than in a war where these actors also fight each other. A civil war where the only behavioral ties are four rebel groups fighting against a government consists of four conflict dyads. In a scenario where all these groups also fight against each other, the number of hostile interactions increases to ten. Empirically, we can consider the Angolan liberation struggle (1961)(1962)(1963)(1964)(1965)(1966)(1967)(1968)(1969)(1970)(1971)(1972)(1973)(1974), where the three movements that fought the Portuguese were not allies but bitter rivals fighting each other and attacking the others' ethnic constituencies. 52 Behavior complexity, in terms of behavioral ties, is higher in that case than in, for example, the second Liberian civil war (approximately 1999-2003), where two rebel organizations fought against Charles Taylor's government but did not fight each other. 53 Patterns of alliances and rivalries (or, wartime political orders) are not constant, and Staniland underscores that how states and nonstate actors relate to each other often shifts over the course of a war: "There is extraordinary diversity in the rules of the game that structure interactions between state forces and non-state armed groups," and these "develop, evolve, and collapse." 54 These shifts matter for complexity too, with side-switching having occurred in at least a fourth of all multiparty civil wars. 55 Changed allegiances can radically alter balances of power and shape conflict evolution. Furthermore, whereas stable networks of alliances and rivalries vouch for some level of predictability, frequent shifts in whom is fighting whom do the opposite. 56 This unpredictability can affect chances for conflict resolution by increasing the risk of commitment problems, 57 strategic miscalculations, 58 and elusive ripe moments. 59 Empirically, this can be exemplified in how a war involving several international actors is more complex if the outside actors' relationships form an intricate web of behavioral ties than if the international actors line up strictly on two clearly defined sides. Hence, a civil war with rapidly changing rivalries can thus serve to increase behavioral complexity.
In addition to the number of hostile behavioral ties, behavior complexity also increases when actors engage in many different types of violent behavior within the confines of a civil war. As is evident by a cursory look at the civil war literature, a plethora of different types of violent behavior can exist in a civil war. Deliberate targeting of civilians by the warring actorsbe they governments, militias, or rebels-transpires in many such conflicts and comes in different forms, such as ethnic cleansing, forced displacement, genocide, or terrorist attacks. 60 Other forms of violence that some warring actors engage in include sexual violence, 61 attacks on aid workers, 62 and targeting of peacekeepers. 63 As Guti errez-San ın and Wood elegantly explain, all actors that conduct political violence have a distinct violence "signature" and display a wide diversity in the configuration of the "reportoire," "targeting," "frequency," and "technique" of the violence they use. 64 Consider the wide-ranging repertoire of IS violence that included, for instance, sexual assalts, bombings, and executions, in comparison to the African National Congress, which primarily engaged in sabotage attacks against targets associated with the apartheid system. 65 Various types of violence can feed into spirals of other forms of violent behavior, thereby enhancing complexity. In Iraq, for example, attacks on civilians (by coalition troops) provoked higher levels of insurgent retaliation. 66 Reed M. Wood and Jacob D. Kathman have also demonstrated how the likelihood of civilian targeting increases when nonstate actors are engaged in conventional fighting against each other, 67 while Alex Braithwaite shows how the deployment and violent actions of an intervening actor in a civil war lead to an increase in the use of terrorist attacks. 68 Dynamics such as these serve to increase the intensity with which a civil war is fought, as well as make conflict resolution more difficult due to the widening of the civil war.
A plurality of violent behavior types increases complexity, and thereby posits a challenge to conflict resolution. Not only do more forms of violence need to be halted for a conflict to be resolved, but the suitable management approach is also likely to differ between the forms of violence. Peacekeeping, for instance, could become more difficult if the warring actors engage in many different forms of violence. Stian Kjeksrud has demonstrated both quantitatively and qualitatively how the varying types of violence used by nonstate actors require vastly different approaches from peacekeepers to be able to protect civilians and fulfill their mandate. 69 In Darfur, peacekeepers have been ensnared in a situation where they have faced numerous types of violent demeanor. In addition to warfare in the militia-rebel-government nexus, the civil war in Darfur has also included violence such as burning of villages, ethnic cleansing, forced displacement, massacres, sexual violence, and intercommunal violence. 70 To further exemplify, intercommunal violence displays distinct features compared to rebel-government violence that are important for resolution strategies. The fact that communal conflicts generally are less asymmetrical, involve parties with a lower level of organization, and often are much briefer (in many cases the actual fighting only occurs during a few days) than civil wars means they require different approaches for solving them. 71

Issue Complexity
Carl von Clausewitz famously asserted that "the political object is the goal, war is the means of reaching it, and means can never be considered in isolation from their purpose." 72 This reminds us that all civil wars involve contested issues: the specific demands and grievances that warring parties have chosen to go to war over. Issues also have a prominent position in theoretical models of conflict. 73 Nevertheless, issues have drawn significantly less attention than actors and behaviors in the civil war literature. 74 Bringing issues back in not only allows for a more nuanced picture of what civil wars "are about"; this is also fundamental for conflict resolution and, as we argue, essential for grasping a civil war's complexity. We posit that both the number and type of conflict issues are important for complexity.
First, an increasing number of issues entails higher complexity. Conflicts are often classified according to whether the main incompatibility concerns territory or government. 75 However, wars are not binary, and the number of issues varies significantly across cases. Whereas a high number of issues may enable horse-trading, an increase in issues overall adds to complexity and can make a conflict more difficult to resolve, including by reducing the possibility of finding a solution that the parties prefer above continuing the war. 76 Variations in the number of issues exist both at the national and local levels. 77 At the former, a war may center on one conflict issue, such as rebel representation in government, while another war may focus on several issues, including demands for autonomy, a purge of the security forces, and recognition of minority rights. A varying number of issues may similarly exist at the local level. 78 The most complex civil wars thus have many issues at both levels, and they can be interconnected. 79 The Israel-Palestine conflict illustrates the importance of issue interconnections. Although land sits at the heart of the conflict and could be distributed between the parties, this issue is closely linked to others such as settlements, Jerusalem, holy places, the right of return, and security guarantees. 80 A solution to the land issue subsequently influences how to deal with settlements and with Jerusalem; the issues are connected and should be analyzed as such.
Second, the type of issues feeds into complexity. Civil wars can involve a high number of issues, but not all are equally important to actors and not all demands are equally extensive. Based on this reasoning, we propose that how much complexity an issue induces depends on its salience, divisibility, and totality. Issue salience denotes how important warring actors consider an issue. 81 Some civil wars contain dozens (or hundreds), but not all are equally important. The rebel group Fuerzas Armadas Revolucionarias de Colombia (FARC), for instance, has specified that it was fighting for the overthrow of the Colombian government, but also to stop the dam project in El Quimbo. We deem it uncontroversial to assume that overthrowing the government takes precedence over stopping the dam project; that is, the former is more salient than the latter. Relatedly, the parties' perceptions about divisibility matters. Issues deemed indivisible are intricate, as they hamper opportunities for cooperation. 82 For example, the issue of land in Israel-Palestine is again illustrative: though land as such can be divided, Jerusalem itself is not easily divisible, especially from a symbolic perspective-it is an important holy place for the three Abrahamic religions. 83 Conflict issues also vary in how extensive, or total, they are. Complete changes of political systems and demands for independence are, for instance, two issues that call for fundamental societal changes: they are of the total type. One manifestation of the fact that the degree of totality matters is that separatist wars often become highly intense. 84 In the middle of the spectrum, we find issues such as protection of minority rights, increased autonomy, or security sector reform. Issues that do not impact society in an equally extensive manner include demands for amnesties, release of prisoners, or recognition of wrongdoing, and are thus less total. The Israel-Palestine conflict covers issues across the whole spectrum of totality, from claims by both parties regarding full territorial sovereignty to less extreme issues of water management. By demanding more fundamental changes to society, more total issues add more complexity to a civil war than less extensive issues.

Interdependent Complexities
As implied by our choice of a triangle to illustrate civil war complexity, the three dimensions are not separate; they are linked. Guided by complexity theory's focus on interlinkages and interdependencies, we now turn to consider the synergism within the three dimensions and how complexities reverberate throughout a civil war system. Complexity in one dimension may lead to complexity in another or both, as the dimensions are interlinked.
First, a change in actor complexity can influence complexity in the two other dimensions. A rebel group breaking apart entails an increase in actor complexity, which in turn can cause increased behavior complexity if these groups engage in violence against each other. A typical example is the fractionalization of the Sudan Liberation Movement/Army in Darfur, which brought significant infighting between the many resulting factions. Increases in actor complexity can also spawn higher issue complexity. If the number of rebel groups increases, issues at both the national and local levels are likely to grow. At the national level, groups may have different political goals and foresee different futures for the country, as with the Sudan People's Liberation Movement/Army (SPLM/A)'s split in Sudan in 1997, which saw rival groups emerge with differing end goals. Rebel fractionalization, however, often relates to leadership struggles, and two previously united groups can maintain similar goals vis-a-vis the state. Even so, rebel splits may spur increased issue complexity at the local level, with new issues (such as local political representation, land rights, and resource distribution) emerging. 85 In this way, fractionalization can cause increasing complexity beyond the mere expansion in actors. Contrastingly, decreased actor complexity can reduce complexity in the other corners. For example, the merger of different SPLM/A factions (that is, decreased actor complexity) significantly lowered nonstate fighting and attacks against civilians. The reduced actor complexity-and its consequences-paved the way for the Comprehensive Peace Agreement (CPA) that terminated the war in 2005. 86 Second, a shift in behavior complexity can affect complexity in both the actor and issue spheres. For example, a more widespread use of violence, such as the targeting of civilians, can spawn new actors (who might emerge to counter these threats or to avenge the killing) and new issues (such as compensation and accountability). The full-fledged civil war that emerged in South Sudan in December 2013 offers an example. Though this war shattered large areas of South Sudan, Western Equatoria was largely unaffected during the conflict's first years. In 2016, however, government forces' use of violence contributed to the reactivation of dormant militias in this area. 87 And in the DRC, violence between rebels, between rebels and the government, and violence against civilians have continuously sparked the emergence of self-defense groups and new militia or rebel outfits. 88 Additionally, as Laia Balcells argues, revenge and honor serve as important drivers of conflict, directly related to communities being exposed to violence. 89 Moreover, government incursions into Western Equatoria not only spawned new armed actors but also unearthed old issues while creating new ones. Government soldiers (primarily Dinka) abused Equatorian civilians and gave support to raiding co-ethnic cattle herders. Dinka encroachment on their land enraged Equatorians, especially since government soldiers now supported Dinkas in local conflicts with old roots. Equatorian leaders soon concluded that their local grievances over land and cattle raiding could only be resolved by breaking Dinka domination at the national level; thus, changes in behavior reignited latent local conflicts and transformed/elevated them to national-level issues for Equatorian elites. 90 85 On the proliferation and importance of local issues in the "aftermath" of the war in the DRC, see, for instance, S everine Autesserre, "D. Intensified behavior complexity can thus spur more actor and issue complexity. In a situation where behavior complexity instead decreases, it reduces the need for self-defense and revenge; we should thus not expect to see the same proliferation of actors and issues. Hence the interlinkages between the complexity spheres can go in both directions. Finally, issue complexity can influence the other corners of our complexity triangle. Issues in civil war can "tie" actors together either in conflict or in cooperation. The desire to oust the government, for example, can result in rebel groups forming an alliance and thereby reduce actor complexity. Though such an alliance may prove unstable-as armed groups may have disparate societal goals outside a shared desire for regime change-sharing the same ultimate goal can also tie rival groups together. After the signing of the 2005 CPA in Sudan, a six-year interim period before possible independence followed. Analysts widely believed the referendum would never materialize and that the regime in Khartoum would take turbulence in the south as a precondition for its cancelation. However, despite fighting each other fiercely both before and after this interim period, a "remarkable display of unity by southern Sudanese of all tribes and political persuasions" characterized this interval. 91 The pivotal reason was a unanimous desire for independence. Sharing the same ultimate goal thus contributed to a period of relative peace by decreasing issue complexity.
An increase in issue complexity may also enhance actor complexity when new issues threaten the status quo and lead to actor mobilization. The more issues parties raise, and the more salient these are, the more likely it is that other actors will mobilize to protect their stakes. In Darfur, for example, some communities that initially stood outside the conflict perceived the rise of the rebel movements and their demands regarding land rights and political influence as threatening those communities' interests. Fear that their position would worsen made these groups more vulnerable to government manipulation; members of these groups subsequently joined the Janjaweed and other outfits. 92 The issues raised thus contributed to a higher number of actors involved.
Increased issue complexity may also exacerbate the risk of more behavior complexity. The first pathway between issue and behavior complexity goes via actors. With more actors entering the fray, the risk for additional hostile ties such as interrebel fighting increases. Furthermore, in our framework, we look beyond alignment along one single master cleavage, or incompatibility, and acknowledge that actors-via issues-can be aligned along several cleavages. 93 This perspective entails that higher issue complexity makes for complicated networks of alignments between actors that could affect behavior complexity by, for example, increasing the fluidity of alliances and rivalries. Syria is a case in point. The high number of issues (at the local, national, and international levels) results in highly fluid "alliances" where, for example, the Free Syrian Army and the regime both fought IS while simultaneously opposing each other on other issues. Similarly, Turkey fought the Kurdish movements in Syria, yet also opposed many of their enemies. Complex issue networks thus undermine stable alliances and cause rampant-at least in the Syrian case-side-switching and unpredictability in patterns of violence.
Discussing these interdependencies between our dimensions demonstrates that each is not freestanding, and there is thus a need to challenge a paradigm of increasing compartmentalization in the study of civil war. This is one of the major points of applying a complexity perspective: the realization and application of principles of interdependence between civil wars' core building blocks. Changes in the complexity of these blocks reverberate through the full civil war system in what complexity theory would call "feedback loops" 94 that can in turn cause even further complexity. This does not mean civil wars that display a higher degree of complexity necessarily score high on each dimension; as outlined above, dimensions are interdependent in various ways. This entails specific types of problems for conflict management and conflict resolution, including, for instance, how the resolution of the issues at stake need fine-grained consideration regarding how a solution resonates with other involved groups, as well as how the resolution of one issue affects the possible solution of others.

Charting Civil War Complexity
Above we have made the theoretical case for how to conceptualize our three dimensions. We then linked them to the general tenets of both complexity theory and conflict theory and illustrated how they are interconnected. We applied short empirical examples to illustrate how these dimensions manifest themselves. But how does our newly developed concept resonate with widely held ideas and intuitions about which civil wars are complex and which are not? To approach this question, we first chart civil war complexity using large-N data on some aspects of the different dimensions. This exercise serves to describe the concept empirically and provide a measure of face validity, as well as allow us to understand the universe of variation. Second, we delve deeper into the case of civil war 93 See also Stathis N. Kalyvas, "Ontology of 'Political Violence,'" 475-94; Autesserre, "D. R. Congo." 94 Mitchell, Complexity; Robert Jervis, "System Effects Revisited." complexity in Darfur to both underscore the real-world importance of taking the complexity framework on board as an analytical tool and to more fully illustrate empirically how it manifests itself.

Variation in Complexity across Cases
First, to demonstrate the variations of actor complexity, Figure 2A,B displays variations in the number of actors participating in violence in a country between 1989 and 2016. 95 Figure 2A includes state, nonstate, and communal actors, whereas Figure 2B shows rebel nonstate actors only.
These figures show large variations in how complex civil wars are on the actor dimension: some have only two or three warring actors; others, such as the conflict in Mali, count more than 60 actors at their peak. The many actors in Mali, Afghanistan, and Iraq (see Figure 2A) stem to a large degree, however, from the intervention of international coalitions. Just below these wars with high levels of external involvement, a second class emerges, containing wars such as those in India, Pakistan, Syria, Somalia, Nigeria, Sudan, South Sudan, and the DRC. These wars score high on actor complexity since many rebel groups and/or communal actors are active. On the lower end of the spectrum are wars with only two actors, which account for 34 percent of the country-years. This proportion includes conflicts such as the one in Spain with Euskadi Ta Askatasuna (ETA), the conflict over Western Sahara, and most years of the Mozambique government's war with Resistência Nacional Mocambicana (RENAMO). Figure 2B displays only rebel actors. The picture is quite different from Figure 2A and demonstrates the actor complexity of the Syrian civil war from a more "domestic" perspective. After Syria, the countries involved in civil war with the highest number of rebel actors are Afghanistan, the DRC, Libya, India, and Pakistan that all observed at least one year with around ten rebel actors. African conflicts tend to have the highest involvement of less organized communal actors, and these have been particularly prominent in Ethiopia, Nigeria, Sudan, and South Sudan (see Figure 2C in the online appendix). The n component of our actor complexity dimension thus has strong face validity, as the civil wars scoring high on actor complexity are those often mentioned in this context. It manages, however, also to demonstrate differences regarding types of actor complexities. Sudan and South Sudan's wars, for instance, emerge from a higher number of rebel and communal actors, while Mali's is mainly a product of foreign intervention. India, on the other hand, has seen very little foreign involvement but a high level of rebel complexity in its interwoven conflicts in the tribal areas of the northeast.
We turn now to behavior complexity and illustrate the number of hostile ties between actors in civil wars. We again rely on country-level yearly data from the Uppsala Conflict Data Program as the best available proxy, with Figure 3 displaying the number of dyads engaged in hostile relationships by country-that is, the number of actor pairs who engage each other directly in violence as part of an ongoing civil conflict (at least twenty-five deaths in a calendar year). Hence, here we capture situations, for example, in which rebel groups are in conflict not only with a government, but also with each other.
This exercise also provides a degree of face validity for this dimension, as cases such as Syria, the DRC, Nigeria, India, Libya, and Sudan all appear at the top of the graph. As was the case for actor complexity, stark differences exist across wars. Syria is a prominent case at the extreme end of both actor and behavior complexity. However, note that although there were at least 50 actors registered in 2014, "only" approximately 27 dyads were directly engaged in hostile relationships with each other. In other words, there were many actors in Syria in 2014, but the vast majority did not fight more than one other actor. In contrast, the DRC in the early 2000s registered roughly 15 conflict dyads out of approximately 10 actors, a proportion suggesting a relatively higher density of behavior ties. 96 The large variations identified, however, are of more general interest for our argument. These range from wars with only a single dyad to violent processes involving up to almost 30 such pairs. Bearing in mind what we know from previous literature regarding how rebel-rebel interactions and similar phenomena affect conflict processes, and what we here show in terms of face validity, it seems critical that we consider the behavior complexity dimension and the implications this may have for the conflict resolution tools needed to resolve such wars.
Lastly, we illustrate how the number of issues varies across conflicts. Since global data in this area is scarce, 97 we describe some patterns based on a new issue data collection, which covers Africa and the Americas. 98 Figure 4 displays the number of stated issues per country and year for armed conflicts between 1989 and 2016 in these regions. In this figure we thus do not consider the type of issues, only the number. A score of 1 (bottom of the chart) implies that only a single issue was mentioned as a stated goal by a rebel group in that year. Figure 4 clearly demonstrates that much as was the case for actors and behaviors, civil wars display significant variations in how many issues rebel groups say they are fighting over. The figure shows how, from around 2000 onward, between 30 and 50 different issues of contention have been voiced in the civil wars in Sudan and South Sudan. To a certain extent this can be linked to the many actors participating in these wars, but the picture is not fully reducible to that dimension. El Salvador, Peru, Mexico, and Colombia have few actors compared to Sudan and South Sudan, yet the former have registered between 20 and 45 issues during some years. This stands in stark contrast to many other wars in which only a handful of issues are stated as the bones of contention. Allowing the incompatibility side of civil wars to vary thus demonstrates the vast heterogeneity inherent in how this concept manifests itself empirically. It is more difficult to make a call on face validity regarding issues, seeing as how little relevant data has been available, and how little cross-case research has been conducted as a consequence. Nevertheless, the most visible cases are not likely to surprise those familiar with these wars. Deeper analyses of issues are, however, necessary before drawing any firm conclusions.
To sum up, we find it encouraging to see that there is some face validity to all three dimensions, as wars one would expect to display high levels of complexity are also those that score high on the respective complexity dimension. Studying all three visualizations of the dimensions demonstrates one point clearly: there is vast variation captured by fuzing together the concepts of complexity and civil war even though data limitations prevent us from empirically exploring the concept in detail. Being able to-through the use of more data, for instance-demonstrate the full range of interlinkages and types would yield even more variations of interest. This, we argue, illustrates that a reductionist view of civil wars that compartmentalizes or dichotomizes dyads, actors, or conflict issues risks defining or operationalizing away variations already identified as important for conflict dynamics and outcomes.

Darfur through the Complexity Lens
To illustrate the real-world importance of our conceptualization, and to show how our framework can facilitate new analysis of conflict resolution initiatives, we turn to Darfur. The civil war in Darfur is illustrative of how civil war complexity-and the failure to recognize it-impeded conflict resolution: our conceptualization provides a tool for grasping this.
Darfur's civil war began in 2003, and the gross human rights abuses the government and its affiliated Janjaweed militias conducted gave rise to extraordinary global attention that consequently put the international community under immense pressure to solve the crisis. After drawn-out negotiations the Sudanese government and one rebel faction signed the Darfur Peace Agreement (DPA) in 2006. However, rather than bringing peace to Darfur, the accord exacerbated the conflict: the fighting intensified, the rebel side fractured, and the civil war lingered on for more than a decade. Our purpose here is not to analyze the particularities of the DPA's failure in detail-that has been done elsewhere 99 -but to illustrate how civil war complexity provides a useful framework for third parties facing difficult conflict management situations.
Repeatedly stated reasons for the DPA's failure include that it did not address the root causes or consider the complexity of the situation. While correct, this provides limited analytical guidance. Adam Azzain Mohamed, in contrast, argues that the fundamental problem was the mediators' perception of the crisis "as simply a region-centre conflict, over equitable national wealth and power sharing," 100 thereby overlooking that the crisis also encompassed local resource-based conflicts and intercommunal elite conflicts. The international community thus ignored complexity and contributed to the accord's failure. 101 Issues fundamental for a durable solution to the Darfurian crisis (such as distribution of land and reconciliation) were left unaddressed or only treated superficially. A multitude of highly salient local issues thus remained unresolved between the majority of armed actors.
Overlooking the importance of local issues in this way was in turn linked to the international community's failure to fully appreciate the importance of actor complexity. This meant critical actors were unconsidered during the negotiation process, although their grievances and issues were strongly intertwined with the DPA's political consequences. The region had for decades experienced violent intercommunal conflicts over issues such as grazing, water, pasture, and political recognition, leaving a large host of armed and relevant actors. In fact, the government's biased intervention in these conflicts constituted a key motivation for the insurgency. 102 The DPA also brought about an extensive proliferation of the number of armed actors. Whereas the rebel side comprised three movements during the negotiations (with only one signing the agreement), the DPA was followed by a series of splits that generated dozens of new resistance movements. 103 A consequence of this fragmentation was that negotiations post-DPA needed to focus on talks between different rebel factions rather than between the opposition and the government. The challenge this brought to future attempts to solve the conflict is illustratively captured by Ambassador Jan Eliasson (UN secretary-general special envoy on Darfur in 2006-2008), who, together with Dr. Salim Ahmed Salim, acted as a chief mediator for Darfur: I never thought I would say that the Iran-Iraq negotiations were easy, but in comparison with Darfur they were. The main reason for the Darfur negotiations being of such an extreme difficulty is the mosaic of actors involved. To solve this conflict a solution is needed at several levels: within the resistance movements, within the Sudanese government, between the movements and the government, among the neighboring states (especially Chad, Eritrea, Libya and Egypt), and within the United Nations and African Union. Thus, peace in Darfur is so elusive because simultaneous solutions at several levels are needed. 104 Extensive behavior complexity also aggravated the situation. The conflict in Darfur encompassed a wide swathe of different violence types (such as ethnic cleansing, massacres, sexual violence, violence against peacekeepers), and many of the actors repeatedly shifted their allegiances. At the time, outsiders often labeled the war in Darfur as an Arab-African conflict where the government supported and armed the Darfurian Arabs. This overly simplistic perception meant disregarding important dynamics. For example, while the communal conflicts preceding the insurgency primarily pitted Arabs against non-Arabs, most communal conflicts after 2003 stood between communities that both identified as Arab: the solutions needed to end war went far beyond reconciling Arabs and Africans. There was thus a lack of attention to those critical behavioral ties that ran against the prevailing and tidy Western assumptions about who is fighting whom. Another important aspect for behavior complexity was that the government, as part of the DPA, committed itself to disarm and abandon the Janjaweed, which infuriated many communities and opened new fault lines and changes of allegiance. 105 Because the government had armed these militias, outside analysts believed they could also disarm them. To empower its insurgency, the government gave massive amounts of weapons to Janjaweed leader Musa Hilal. Later, however, when he was disappointed with how the government treated some Arab communities that had fought at its side, he used these weapons against the regime when he launched the insurgency group Sudanese Awakening Revolutionary Council. 106 The above expos e is telling in regards to how a complexity lens can provide a sound mode of analysis. Using our complexity concept identifies and places the limelight on those specific areas where other analysts have found the main shortcomings of the DPA. In other words, our complexity perspective highlights those variations in civil wars that are of critical import for understanding war dynamics. Accounting for the interlinkages between dimensions also provides a dynamic way of thinking about how the core components of actors, behaviors, and issues affect each other, and can help us better understand how to resolve conflicts.

Concluding discussion
Understanding civil war is pertinent to both the policy and academic communities. As several scholars have argued, how we choose to conceptualize our core concepts matters greatly for how we can approach them in terms of research. 107 Present conceptual limitations concerning civil wars can lead us to overlook the patterns and relevant comparisons needed to understand and resolve these wars. In this study we have contributed to remedying this by providing a conceptualization of "civil war complexity" and illustrating its empirical manifestations. Our descriptions show that our concept has both face validity and the ability to capture critical variations in civil wars that affect conflict processes and outcomes.
The ever-growing literature on the microdynamics of conflict suggests that both degrees and types of complexity play a key role in the evolution and settlement of civil wars. As we described, variations in the number and type of actors, conflict behavior, and the issues fought over have been linked to increases in the human cost of civil wars, difficulties in reaching peace agreements, and length of civil wars. 108 Many studies have underlined the importance of taking heed to not study civil wars as pure dyadic interactions. 109 In other words, the multiple actors that exist, the multiple behaviors they engage in, and the multiple issues they are fighting for play into how a civil war unfolds. Our concept unifies these findings and theories into a whole, and can thus be a highly useful tool for both theoretical and empirical analyses.
We believe our concept can prove useful in terms of better understanding the causes and consequences of civil war complexity. First, by more clearly delineating which civil wars exhibit more complexity than others, we can begin to understand if the civil wars with a high degree of complexity differ from others in terms of their causes. A fruitful application of our concept may thus be to revisit existing explanations for civil war onset to explore if these factors also hold explanatory leverage for why some conflicts spiral into a high degree of complexity while others do not. If we, for instance, approach complexity levels as portraying different classes of civil war, it is probable that more complex civil wars have somewhat different precipitating causes. Understanding a country's propensity for complex civil war before the outbreak of conflict seems a worthwhile exercise. Second, we can start to explore if those civil wars that exhibit greater complexity also follow certain trajectories. With a clear conceptualization of civil war complexity, we can study if and how the degree of complexity affects several conflict outcomes, such as conflict intensity or the duration of the conflict. It appears pertinent to study these dimensions not only separately-focusing on, for instance, fragmentation or the issues rebels fight for-but to unify them to consider if these effects are additive or multiplicative, as well as how they feed into each other. Third, and most importantly, a commonly stated reason for the difficulty in resolving some civil wars are their inherent complexity. Our conceptualization allows one to investigate whether the degree of civil war complexity influences a variety of conflict resolution outcomes, including the onset of negotiations and ratification of peace agreements. This can help us better understand if civil wars displaying variation in complexity may require specific tailored solutions. We could here consider the possibility that different degrees and types of complexity spawn different drivers of continued conflict. A civil war driven by a proliferation of local issues is likely to require a different form of peace agreement than a war where all compete for power at the center.
In sum, we see several useful applications for our new conceptualization of civil war complexity that should help inform and challenge our current understanding of civil wars. As such, we hope it can offer an opportunity for a shared vocabulary and terminology. Our aspiration is that making civil war complexity more concrete can serve to make our discussions on the variations and peculiarities of civil wars more precise, and be an important step toward a research agenda that explores causes and consequences of our most devastating wars.