In the late 1980s, Mozambique was suffering from a civil war that had destroyed the country’s infrastructure, resulted in severe violence against civilians, and contributed to widespread famines. It was at that time that a militia emerged, an armed group of volunteers from the civilian population that confronted the Renamo rebels who fought against the Frelimo government.Footnote ¹ “The people revolted. They were tired of the war, so they volunteered to confront those who were waging war and end [the fighting],”Footnote ² a local government representative explained to me in one of my many conversations about the origins of the group. Naparama, as the militia was called,Footnote ³ was created by a traditional healer in northern Mozambique, Manuel António, who claimed that he had received a divine mission from Jesus Christ to liberate the Mozambican people from the suffering of the war and learned of a medicine to turn bullets into water (Nordstrom Reference Raleigh and Kishi1997, 58).Footnote ⁴ António used the medicine to vaccinate militia members during an initiation ceremony. Naparama became the most important of many violent and nonviolent civilian resistance movements that emerged to stop the violence during a war that lasted sixteen years in total. The movement quickly spread across the country’s central and northern provinces, growing from a couple of hundred to several thousand members in at least twenty-six districts across two provinces within a year of its formation in 1988–89.Footnote ⁵ António went “on foot if necessary, to ‘wherever the people call me to help’” to train new members.Footnote ⁶ The people embraced this new force, and youths even dropped out of school to join the militia.Footnote ⁷ By 1991, Naparama controlled two-thirds of the northern provinces and returned stability to war-torn communities (Wilson Reference Wilson1992, 561).

Militias like Naparama are part of a broader phenomenon that is common across civil wars around the globe (Üngör Reference Üngör2020). By civil wars, I refer to armed conflict within a country between at least two parties subject to a common authority (Kalyvas Reference Kalyvas2006, 17). Militias, as defined in this book, are armed organizations that exist outside of the state’s security apparatus; they emerge as “countermovements” against insurgents either on the initiative of community residents or state representatives (see Jentzsch, Kalyvas, and Schubiger Reference Jentzsch, Kalyvas and Schubiger2015). Similar to communities in Mozambique, residents in Peru, Nigeria, and Sierra Leone formed militias to protect themselves against civil war violence. The civilian defense committees (rondas campesinas) fought against the Shining Path in Peru, the Civilian Joint Task Force (CJTF) countered Boko Haram in Nigeria, and the Kamajors quelled the Revolutionary United Front (RUF) rebellion in Sierra Leone (Zech Reference Zech2016; Bamidele Reference Bamidele2017; Hoffman Reference Hoffman2011). Sometimes, governments take the lead in mobilizing militias to counter armed rebellion. The National Defence Force, for instance, supports the rule of Syrian president Bashar al-Assad (Leenders and Giustozzi Reference Leenders and Giustozzi2019). The Iraqi government, together with the United States, collaborated with militias to counter Al-Qaida in Iraq (Cordesman and Davies Reference Cordesman and Davies2008; Ahram Reference Ahram2011). Sudanese president Omar al-Bashir, ousted in 2019 by the Sudanese Armed Forces, was notorious for delegating violence to the Janjaweed to fight the rebellion in Darfur (Flint and De Waal 2005). And the Afghan government has worked with warlords and militias to defeat the Taliban (Malejacq 2019).

This strikingly regular feature of civil wars, the presence of domestic “third actors,” and its significance for order and violence during civil war is neglected in conflict and security studies. Although militias were active in nearly two-thirds of civil wars between 1989 and 2010 (Stanton Reference Stanton2015) and 81 percent of the conflict-years between 1981 and 2007 (Carey, Mitchell, and Lowe Reference Chichava and Durán2013, 254), they remain under-researched and under-theorized. Civil wars such as the one in Mozambique are often understood as dichotomous forms of armed conflict between states and insurgents. While scholars have studied why insurgent groups factionalize and fragment (Bakke, Cunningham, and Seymour Reference Bakke, Cunningham and Seymour2012; Cunningham, Bakke, and Seymour Reference Cunningham, Bakke and Seymour2012; Christia Reference Christia2012; Woldemariam Reference Woldemariam2018), systematic analysis of armed groups formed to support the incumbent is more limited (Jentzsch, Kalyvas, and Schubiger Reference Jentzsch, Kalyvas and Schubiger2015; Malejacq Reference Malejacq and Mukhopadhyay2017; Carey and Mitchell Reference Carey and Mitchell2017; De Bruin Reference De Bragança and Wallerstein2020). We know why states form and delegate violence to pro-government militias (Carey, Colaresi, and Mitchell Reference Carey, Colaresi and Mitchell2015, Reference Carey, Colaresi and Mitchell2016; Biberman Reference Biberman2018, Reference Biberman2019). However, our understanding of how and why civilian communities organize to form militias is in its early stages (Zech Reference Zech2016; Blocq Reference Blocq2014).

I conceive of community-initiated militia formation as a type of collective action and an expression of civilian agency. Civilians are often seen as “extensions” of various armed groups in civil wars, not as actors in and of themselves. They are usually seen as facilitating armed group activities by providing access to resources such as food, money, recruits, and intelligence, rather than initiating any activity themselves. Recent research, however, has recognized the significance of “civilian agency” in civil war (Masullo Reference Masullo2015; Kaplan Reference Kaplan2017; Krause Reference Krause2018). Rather than passive victims, civilians respond to civil war in creative and organized ways, seeking to improve their own protection (Jose and Medie Reference Jose and Medie2015). While they may have nonviolent means at their disposal, civilians can also opt for violent ones, such as forming militias to ward off insurgent (and state) violence (Jentzsch and Masullo Reference Jentzsch and Masullo2019). By drawing attention to militias as a form of collective action, I emphasize the coordinated and organized nature of civilian responses to war and violence.

Recognizing how multiple non-state armed groups and civilian agency affect political violence, this book seeks to answer the following questions: Why do civilian-based, community-initiated militias emerge at particular times during civil war? What explains the spread of such militias across war-torn communities like Mozambique? Why are people drawn to participate in militias, even at considerable risk to life and limb?

1.1 When, Where, and How Militias Form

The book explains when, where, and how community-initiated militias form in irregular civil wars. While, based on existing scholarship, we might expect governments to mobilize militias when they are losing, or when they have the upper hand to retain their advantage, I argue that it is precisely when a military stalemate has been reached that communities themselves form militias. Stalemates pose significant risks for civilians as they find themselves between two forces, of which neither is able to protect them. Thus, we should expect communities to organize collectively and form militias when caught between evenly matched foes. This argument is important because it challenges our understanding of civil war as synonymous with rebellion or insurgency. Much of conflict research focuses exclusively on rebel groups and overlooks third actors such as militias (Metelits Reference Metelits2010; Weinstein Reference Weinstein2007). The lack of attention to domestic “third actors” in armed conflict theories is largely due to the simplifying assumption that civil wars take place between two sides – incumbents and insurgents (Pettersson, Högbladh, and Öberg Reference Pettersson, Högbladh and Öberg2019). Taking militias into account resists the tendency to portray state–rebel relations as purely dyadic interactions, with consequences for theorizing and modeling how rebellions emerge, evolve, and succeed or fail. For example, third actors can overcome a military stalemate and upend the military balance between rebels and the state and thus influence how the war evolves and ends.

In a civil war, militias rarely form independently of each other. To fully understand why community-initiated militia form, we need to consider how such forms of collective action diffuse across community-boundaries. Ethnic, ideological, and cultural bonds between communities and successful militia activity in neighboring communities promote the initial diffusion of militias. However, as I show in this book, militias only take root when community and local elites’ preferences overlap – a process I call “sustained diffusion.” A militia cannot establish itself in a community with elite conflicts, as it may be used to challenge local state authority. In emphasizing the diffusion of collective action forms and repertoires, this book is part of a research agenda that focuses on the endogenous dynamics of armed conflict (Arjona Reference Arjona2016; Balcells Reference Balcells2017; Kalyvas Reference Kalyvas2006; Krause Reference Krause2018; Staniland Reference Staniland2014; Steele Reference Steele2017; Wood Reference Wood2003). For example, I show that wartime collective action cannot be reduced to prewar structural factors such as ethnic group fragmentation, inequality, poverty, or state capacity (Cederman, Gleditsch, and Buhaug Reference Carey, Mitchell and Lowe2013; Collier and Hoeffler Reference Collier and Hoeffler2004; Fearon and Laitin Reference Fauvet and Mosse2003). Communities adopt successful repertoires from other communities over the course of a war, and inter-elite relations affect whether a community routinizes certain forms of collective action.

Finally, once they are formed, community-initiated militias need to grow to establish themselves. The peculiar nature of civil war gives rise to a context of uncertainty in which the consequences of civilian actions are difficult to calculate. Under uncertainty, people tend to make use of familiar knowledge to make decisions and plan their actions. Applying this insight to militia mobilization, I argue that community-initiated militias successfully mobilize members if they appeal to familiar preexisting social conventions. Reminding people of their own available resources provides community residents with the opportunity for self-empowerment and thus encourages participation. Militias’ rootedness in the social and political fabric of the communities in which they mobilize brings about a considerable advantage over rebel groups in terms of recruitment numbers and support, which may explain militias’ powerful impact on counterinsurgency (Peic Reference Peic2014). Thus, though the book emphasizes the endogenous nature of civil wars, it also demonstrates that armed conflict is not a state of exception completely disconnected from prewar social and political institutions (Balcells Reference Balcells2017; Ellis Reference Ellis1999).

Beyond these theoretical arguments, I make two main contributions related to the study of war on the African continent. For the study of Mozambique, and southern Africa more generally, the book shows the limits of a perspective that privileges elite politics and the external interference of neighboring states in domestic affairs to analyze armed conflict. The historiography of the war in Mozambique has adopted a macro-perspective, paying particular attention to RhodesiaFootnote ⁸ and South Africa’s goal to destabilize Mozambique through the funding and training of the Renamo rebels (Vines Reference Vines1991; Minter Reference Minter1994; Cabrita Reference Cabrita2000; Emerson Reference Emerson2014). However, ethnographies of the war in central and northern Mozambique and recent research in conflict studies more generally have demonstrated that local conflicts, rather than the “master cleavage” of war, shape how communities experience and respond to civil war (Geffray Reference Geffray1990; Nordstrom Reference Raleigh and Kishi1997; Kalyvas Reference Kalyvas2006; Balcells Reference Balcells2017; Cahen, Morier-Genoud, and Do Rosário Reference Cahen, Morier-Genoud and Do Rosário2018a). Access to new sources about the war in Mozambique has made this analysis possible and fruitful (Cahen, Morier-Genoud, and Do Rosário Reference Cahen, Morier-Genoud and Do Rosário2018a). By making use of such new sources and studying community responses to the war, the book thus contributes to a broader debate on violent orders and state formation in the Mozambican context (Macamo Reference Macamo2016; Bertelsen Reference Bertelsen2016).

The second contribution relates to how the Naparama militia organized and mobilized fighters. I build on works in African Studies on prophetic armed movements whose access to (traditional) religious practices shapes the way they organize, mobilize, and fight. Such movements are strongly embedded in the social fabric of particular communities (Kastfelt Reference Kastfelt2005; Nicolini Reference Nicolini2006). The Zimbabwe African National Liberation Army’s (ZANLA) close links with the peasantry, for example, provided guerrillas with access to powerful spirit mediums, and embedded guerrillas into the local popular imagination (Lan Reference Lan1985). The Holy Spirit Mobile Forces led by Alice Lakwena is perhaps the most iconic example of such a prophetic armed movement whose successes on the battlefield were tied to the spiritual powers of its leader (Behrend Reference Behrend1999).Footnote ⁹ I join Danny Hoffman (Reference Hoffman2011) in understanding such movements as developing and adopting an “experimental [military] technology,” and define them as an innovative response to wartime violence that is shaped by developments on the battlefield (Jentzsch Reference Jentzsch2017). In Mozambique, the spiritual dimension of the war and Renamo’s use of spirit mediums provided the background to the formation of Naparama, who reinvented preexisting social conventions to help people cope with the war (Wilson Reference Wilson1992).

1.2 Prevailing Approaches to Studying Militias

Conflict scholars’ narrow focus on rebels and the state has obscured the fact that incumbents and insurgents are rarely unified actors. Civil wars in Central and East Africa and East Asia have seen a proliferation of insurgent groups within the same war, often splitting from the same previously existing insurgent organizations (Stearns Reference Stearns2011; Woldemariam Reference Woldemariam2018; Staniland Reference Staniland2014). Scholars who analyze how and why armed groups fall apart have demonstrated that insurgent factions and changing alliances have an important impact on violence and the dynamics of war (Pearlman Reference Pearlman2009; Pearlman and Cunningham Reference Pearlman and Cunningham2012; Cunningham, Bakke, and Seymour Reference Cunningham, Bakke and Seymour2012; Staniland Reference Staniland2012; Christia Reference Christia2012; Bakke, Cunningham, and Seymour Reference Bakke, Cunningham and Seymour2012).

The fact that state armed forces may be fragmented and states may rely on multiple auxiliary forces to fend off opponents from within the regime or outside of it has recieved less attention (De Bruin Reference De Bragança and Wallerstein2020). The overall number of “pro-government militias” – militias with a clear link to the state – rose during the 1980s and 1990s and peaked at over 140, and then fell during the 2000s (Carey, Mitchell, and Lowe Reference Chichava and Durán2013, 254). Political and military developments in Indonesia, Iraq, Iran, Pakistan, and India over the last decades have demonstrated the extent to which incumbent forces involve militias in counterinsurgency operations (Ahram Reference Ahram2011; Biberman Reference Biberman2019). However, in much of civil war research, incumbent forces are still assumed to be unitary actors (Carey, Mitchell, and Lowe Reference Chichava and Durán2013, 250) and treated as completely separate from civilian actors (Mazzei Reference Mazzei2009, 6).Footnote ¹⁰

Militias may form during peace or wartime, and their activities may be defensive or offensive in nature. In times of elections, for example, political elites may form or sponsor nonmilitary death squads, party militias, or youth militias to intimidate opponents (Campbell and Brenner Reference Campbell and Brenner2000; Carey, Mitchell, and Lowe Reference Chichava and Durán2013; Raleigh Reference Raleigh, Linke, Hegre and Karlsen2016; Raleigh and Kishi Reference Salehyan and Gleditsch2020). But militias can also form to protect civilians against crime or other sources of insecurity. When the state is not able or willing to protect a community, vigilante groups form as a self-help mechanism to (violently) oppose “criminals and others whom the actors perceive as undesirables, deviants and ‘public enemies’” (Abrahams Reference Abrahams1998, 9). In democratizing states such as South Africa, vigilante groups can be an expression of unease over the strengthening of human rights, as crime fighters consider releasing a suspect on bail as a source of insecurity and state “failure” (Smith Reference Smith2019). Vigilantism is therefore “an exercise in power” (Bateson Reference Bateson2020, 1) and is closely linked to how elites or communities attempt to shape political order.

During war, militias are involved in counterinsurgency and the protection of communities. While the majority of militias operate in peacetime, they are very common in civil wars. Carey Mitchell, and Lowe (Reference Chichava and Durán2013, 255) find that 43 percent of all pro-government militias are active when the country experiences a civil war, and in 81 percent of country-years during which the country experiences a civil war, there is militia activity. In contemporary civil wars, militias have been cost-effective force multiplicators and help the state deny accountability for violence, as they can outsource such violence to militias (Carey, Mitchell, and Lowe Reference Chichava and Durán2013). But also in the past, during the anticolonial wars, for example, occupying forces frequently created and collaborated with local forces who knew the terrain well and were able to collect crucial intelligence (Coelho Reference Fearon and Laitin2002; Branch Reference Branch2009; Bennett Reference Bennett and Elman2013). Militias have always been important tools for state repression. States have delegated mass violence against civilians to militias throughout history (Ahram Reference Ahram2014; Üngör Reference Üngör2020). In addition, communities themselves often form militias to protect themselves, such as during the long and violent civil wars in Colombia, Guatemala, and Peru. These groups were encouraged or co-opted by the state or formed in cooperation with social and political elites (Mazzei Reference Mazzei2009; Romero Reference Romero2003; Remijnse Reference Remijnse2001; Starn Reference Starn1995). Usually, community-initiated militias recruit residents for nightly patrols, collecting intelligence, and warning the population of imminent attacks. In many cases, such militias professionalize and militarize over time, at times collaborating with the government or even substituting the army.

In this emerging research agenda on domestic third actors, many issues remain unexplored. Most scholars have focused on how and why states form militias and delegate violence to them. Such research implies that political elites form village guards or death squads to support their strategic efforts in counterinsurgency and state building (Kalyvas and Arjona Reference Kalyvas, Arjona and Rangel2005). However, militias can also form (and evolve) independently of such state initiatives. We know little about why communities form militias, what form these groups take, how they evolve during war, and what that implies for the dynamics of civil war.Footnote ¹¹

It is important to fill this research lacuna as militias have important implications for the dynamics of civil war and its aftermath, with seemingly contradictory effects. They are an important resource to defeat insurgents but contribute to the fragmentation of armed groups, which hampers negotiated settlements to end civil wars (Peic Reference Peic2014; Staniland Reference Staniland2015; Stedman Reference Stedman1997). They empower civilians to protect themselves and at the same time fundamentally restructure the social and political order by providing what has been denoted as non-state governance (Blocq Reference Blocq2014; Malejacq Reference Malejacq2016). They form to limit political violence but become violent actors themselves, often increasing the length and lethality of civil wars (Clayton and Thomson Reference Clayton and Thomson2014, Reference Clayton and Thomson2016; Hoffman Reference Hoffman2011; Starn Reference Starn1995; Mitchell, Carey, and Butler Reference Mitchell, Carey and Butler2014; Aliyev Reference Allen and Vlassenroot2020a, Reference Aliyev2020b). Finally, though militias are often formed during war, they shape the political process in the postwar era, in particular when they are excluded from demobilization and reintegration processes and co-opted by political elites (Acemoglu, Robinson, and Santos Reference Acemoglu, Robinson and Santos2009; Mazzei Reference Mazzei2009; Coelho and Vines Reference Coelho1992; Hoffman Reference Hoffman2003; Daly Reference Daly2016).

1.3 How to Study Militias

The focus on community initiatives to form militias requires detailed evidence from subnational units on how armed groups originate and evolve. To develop a theory of militia formation and explore its validity, this book builds on a subnational research design that allows for within-case comparisons of militia formation over time and controlled comparisons of geographical areas with and without militia activity within the context of one civil war. With both analytical strategies, I carefully trace the causal processes to check the subnational evidence of competing cases against alternative explanations and identify the causal mechanisms at work. In this way, the analysis both develops original arguments and explores their validity in a wider context.

The research design builds on other works that have brought together two recent methodological trends in the study of armed conflict, namely, the within-case analysis method of process tracing and the use of subnational evidence for controlled comparisons (Petersen Reference Petersen2001; Wood Reference Wood2003; Arjona Reference Arjona2016, Reference Arjona, Snyder, Moncada and Giraudy2019; Lynch Reference Lynch2013; Daly Reference Daly2016).Footnote ¹² Bennett and Checkel (Reference Bennett, Checkel, Bennett and Checkel2014) define one of the standards of good process tracing as the combination of process tracing with case comparisons to improve the test of alternative explanations and check for omitted variables.Footnote ¹³

Subnational evidence for controlled comparisons (or quantitative analysis) has greatly advanced research on armed conflict by studying the micro-foundations of violence and order (Kalyvas Reference Kalyvas, Kalyvas, Shapiro and Masoud2008b; Arjona Reference Arjona, Snyder, Moncada and Giraudy2019). A subnational focus can improve data quality, test causal mechanisms, improve the fit between concepts and measurement, and control for variables that can be held constant within the boundaries of smaller units of analysis (Kalyvas Reference Kalyvas, Kalyvas, Shapiro and Masoud2008b; Snyder, Moncada, and Giraudy Reference Snyder, Moncada and Giraudy2019). I use the unit of analysis of the community for subnational comparison, a population in a defined geographical space in which regular face-to-face interactions take place. These interactions create stable direct social interactions that are maintained by common institutions, such as markets, schools, the police, and the administration.Footnote ¹⁴ Empirically, in this book, communities are rural villages.

Combining subnational comparisons and process tracing is particularly valuable because they help uncover causal mechanisms: subnational evidence provides the fine-grained data necessary to conduct successful process tracing (Checkel Reference Checkel, Klotz and Prakash2008, 122). Subnational comparisons also strengthen the validity of the findings from process tracing, as alternative explanations are not only checked against evidence within one case, but also across cases (Lyall Reference Lyall, Bennett and Checkel2015). In addition, combining fine-grained comparative evidence with a process-oriented lens allows for “comparison with an ethnographic sensibility” (Simmons and Smith Reference Simmons and Smith2017). Oral histories help to critically interrogate conventional explanations arising from comparative evidence and analyze how interlocutors themselves experience the formation of militias across communities.

1.4 How to Study Militias in Mozambique

The book focuses on the formation of a certain type of militias – community-initiated militias – and their diffusion across community boundaries. I make use of subnational variation in the formation and diffusion of community-initiated militias in northern and central Mozambique, known under the name of Naparama, during the country’s civil war (1976–92). The insurgent group Renamo emerged shortly after Mozambique’s independence in 1975 among disgruntled Mozambicans with the help of the Rhodesian intelligence service in Rhodesia. Rhodesia and Apartheid South Africa, hoping to end Mozambique’s support for liberation movements challenging their governments, backed Renamo, and disenchantment with Frelimo’s authoritarian policies and one-party system further nourished the movement. By the early 1980s, Renamo extended the war into northern Mozambique. Facing increased levels of violence and abduction, and inspired by the war’s spiritual dimensions, Naparama formed and spread rapidly across two-thirds of the northern territory.

By focusing on subnational variation within one civil war, the book improves upon cross-national studies, which implicitly assume that militias’ presence extends to the entire country. I combine process tracing to analyze how the Naparama militia formed over time with a structured-focused comparison of how the militia diffused across districts. I identify mechanisms of mobilization by comparing the militia’s mobilization success with the less effective mobilization of state-initiated militias. Using an in-depth approach, I facilitate theory building in a thematic field within civil war studies that has remained limited in scope (Jentzsch, Kalyvas, and Schubiger Reference Jentzsch, Kalyvas and Schubiger2015).

I collected different sources of evidence during extensive fieldwork in Mozambique over thirteen months between 2010 and 2016. I collected quantitative and qualitative evidence from over 250 oral histories and semi-structured interviews with former militia members, rebel combatants, soldiers, government representatives, community leaders, and civilians, and more than 10,000 pages of government documents from the Zambézia and Nampula provincial governments’ archives in northern Mozambique. I systematically analyze interviews for narrative patterns, and construct a dataset of violent events from government reports for the province of Zambézia to trace how and why the militia formed. This combination of evidence allows me to triangulate information and overcome challenges of studying a war that ended over twenty years ago.

The war in Mozambique serves as an appropriate context in which to build a theory of community-initiated militia formation as it resembles other civil wars where community residents organized militias for self-defense. In Peru, for example, community-initiated militias were crucial in defeating the insurgency in the 1980s and 1990s (Zech Reference Zech2016). During the second civil war in Southern Sudan, tribal militias emerged to counter the Sudan People’s Liberation Army (SPLA) and settle local conflicts (Blocq Reference Blocq2014). In Indonesia’s Aceh province, ethnic minorities formed militias to protect themselves against rebel coercion (Barter Reference Barter2013). In Sierra Leone, where rebels encountered a severely weak state, community residents resisted violence by both state and rebel forces through the formation of militias (Hoffman Reference Hoffman2011). What these wars have in common is the state’s inability or unwillingness to protect its population, high levels of violence against civilians, and a fragmented nature of war in which local conflicts partly replace the master-cleavage of the war. I therefore expect the findings from Mozambique to apply to a larger set of cases.

1.5 Chapter Overview

The book is organized as follows. In Chapter 2, I develop a theoretical framework to analyze how militias form. I suggest a definition and typology of militias for the purpose of this book, introduce the theory that guides the subsequent analysis, and provide an overview of the research design of the study.

In Chapter 3, I reflect on the unintended consequences of fieldwork in polarized societies, which may affect the autonomy of both the researcher and the researched. In a context of past violence and intractable conflict, research participants often have concerns about how the research impacts the autonomy of their daily life by potentially compromising their safety. On the other hand, research participants may try to make use of the researcher for their own political and economic objectives, compromising the autonomy of the project. In analyzing the simultaneous empowerment and disempowerment of research participants, the chapter discusses the methodological and ethical challenges of power and neutrality during fieldwork and joins others in showing that conflict research needs to be understood as a form of intervention in local affairs.

In Chapter 4, I argue that the warring parties’ strategic aim of controlling the population provided the background for the formation of militias in Mozambique. The control of the population became an end in itself rather than a strategy to control territory. As a consequence, the population’s forced resettlement became a major weapon of war. The war’s focus on the people contributed to the rising level of community responses to the violence, which culminated in Naparama’s formation.

Chapters 5, 6, and 7 are organized along the three parts of the theory – when, where, and how militias form. Chapter 5 shows that, while community responses to the violence were widespread, the Naparama militia formed at a strategic moment in time, when “community-empowering military stalemates” emerged. Tracing the process of how Naparama formed over time, I find that local stalemates shaped community residents’ and local elites’ preferences and gave rise to windows of opportunity for militia formation. Community residents were willing to engage in armed responses to insurgent violence as other options appeared inviable. Local administrative elites complained about insufficient support from the provincial government and supported alternative military solutions such as Naparama. This chapter draws on evidence from an over-time analysis of Naparama’s formation in Zambézia province in Mozambique.

In Chapter 6, I assess why – though facing similar stalemates and other structural challenges – two adjacent districts in Zambézia province experienced the diffusion of militias so differently. I find that communities learned from neighboring communities about how militias formed and “diffusion agents” migrated to spread the message of militia success, which helped initiate militia diffusion. However, “sustained diffusion” – the persistence of militia activity in a district and integration of the militia into the local security apparatus – depended on the cohesion of elites. I explore the validity of my argument by analyzing Naparama’s diffusion to a district in Nampula province.

Chapters 5 and 6 are based on the group level of analysis to explain formation and diffusion of militias, but they do not answer the question of why individuals participate in militias despite the risk. I demonstrate in Chapter 7 that militias successfully mobilize members when they appeal to common social conventions, create innovative institutions and provide an opportunity for self-empowerment. In particular, I show that the appeal to preexisting social conventions such as traditional healing facilitated the mobilization process as the new militia institution resonated with local communities and created a belief in agency, which enabled the large-scale mobilization of members. I develop these arguments with evidence from Nicoadala district in Zambézia province and explore their validity with evidence from the main district of militia activity in Nampula province, Murrupula.

The concluding chapter reviews my theory and evidence and derives implications for how civilian agency, violent resistance, and the rise of third actors affect the dynamics of civil war. I explore how the arguments shed light on similar developments in past and contemporary armed conflicts and reflect on Naparama’s legacy for postwar politics in Mozambique. Overall, I show that, while this book explains when, where, and how militias originate, there is much work to be done to understand how militias evolve and develop their relations with governments, rebels, and civilians. Militias are important third actors in civil wars, but we do not yet completely understand the challenges that come along with their rise.