‘Is Help Coming?’ Communal Self-Protectionduring Genocide

Despite the rhetoric of the Responsibility to Protect principle (R2P), vulnerable groups continue to experience genocide. Some, such as the Yazidis in Iraq, have tried to mitigate genocide through communal self-protection. The dominance of R2P in contemporary normative discussions about responding to genocide, however, means that there has been a lack of research into the lived realities of such experiences. This article explores the phenomenon of communal selfprotection during genocide, through a multiple case study analysis. It examines the pre-eminent examples of communal self-protection during three cases of modern genocide — the experiences of the Armenians at Musa Dagh during the 1915 Armenian genocide, the Tutsi at Bisesero during the 1994 Rwanda genocide, and the Yazidis in Sinjar during the 2014 Yazidi genocide. It presents a typology of communal self-protection strategies during genocide, developed from the case study analysis. The article finds that communal self-protection is only feasible as a strategy in exceptional circumstances. Even in a best-case scenario, communal self-protection offers a temporary reprieve, rather than sustainable living conditions. Vulnerable groups attempting communal self-protection are ultimately reliant on external rescue for their survival, which may not be forthcoming. Communal self-protection should therefore not be regarded as a viable strategy to mitigate genocide in any circumstance.


Introduction
Trapped on Mount Sinjar, besieged by Daesh, 23-year-old Shihab Balki had his mobile phone with him and little else. 'Is help coming?' Balki asked, as he used precious battery life to detail his surroundings for The Washington Post (Morris 2014: A01). He reported that at least 17 children had died on the mountain, and he feared an old lady and child lying on the ground next to him were also dying. Shihab and his family, along with 50,000 or so other Yazidis, had fled up Mount Sinjar to escape an attack by Daesh. With Daesh declaring that converting to Islam or execution were their only options, there was little other choice. Atop the mountain, the Yazidis watched the weakest among them die of dehydration and exposure. They lacked critical supplies essential for lifeincluding food, water, and shelter -and had no means with which to defend themselves and no route for escape. Prospects for surviving the genocide being perpetrated by Daesh seemed remote. Yet, through an extraordinary series of events, the Yazidis' attempt at communal self-protection led to the vast majority of those who fled being able to reach safety. Airdrops of emergency supplies, followed by the opening up of a safe passage allowing their escape, resolved the crisis within days (Shand 2018). In the wake of these events, the experiences of the Yazidi minority elicit important questions about the role of communal self-protection in mitigating genocide.
For the Yazidis, resorting to self-protection from the Daesh genocide was literally a matter of life or death. Despite increasing international focus on genocide prevention in recent decades, and the Responsibility to Protect doctrine (R2P), they found themselves in the position of being targeted for extermination and without protection. Their position was not unique. In genocides spanning the last century or so, a very substantial portion of those who have escaped have done so largely through their own efforts (Mégret 2009: 583). In recent times, other minorities facing mass atrocities, such as the Rohingya in Myanmar, have similarly encountered an international environment in which assistance has not been forthcoming. These lived realities contrast starkly with academic discussions around genocide and mass atrocities. Contemporary normative discussions are dominated by R2P, and a paradigm of external, 'top-down' intervention. Communal self-protection during genocide has received little attention, despite its repeated use as a strategy (Hirschel-Burnes 2014). This article seeks to examine this phenomenon. It begins by looking at current scholarship on R2P, and communal self-protection. The merits, and limitations, of each are considered in light of ongoing international failures to protect vulnerable populations. The article then conducts an exploratory analysis of communal self-protection during genocide. It examines three prominent examples: from the Armenian genocide, the Rwandan genocide, and the Yazidi genocide. The final section of the article explores the findings from these case studies. Each case represents a numerically large attempt at communal self-protection, involving a population not previously displaced, and with some degree of strategic advantage. Yet even in these relatively conducive circumstances, communal self-protection offered highly tenuous and temporary protection at best. Those communities that ultimately survived did so only as a result of external rescue. This suggests that communal self-protection during genocide is not a viable strategy for survival, absent external intervention.

The Responsibility to Protect and Communal Self-Protection
Since the genocidal violence in Rwanda and Srebrenica in the 1990s, the issue of preventing and responding to mass atrocities has been the subject of sustained international attention. Following these genocides, the UN and member states were heavily criticized for their inaction. The lack of political will that framed the response to the Rwandan genocide, in particular, was recognized as a key impediment (UN Security Council 1999: 3). In response, the International Commission on Intervention and State Sovereignty (ICISS) sought to create a new political landscape around intervention that moved away from debates about the 'right to intervene,' refocusing on the concept of sovereignty as responsibility (ICISS 2001). At the World Summit in 2005, UN member states endorsed the R2P principle. R2P emphasizes that states hold the primary responsibility for protecting their populations from genocide, war crimes, crimes against humanity, and ethnic cleansing (often collectively termed mass atrocities). The international community also has a responsibility to assist in this process. Furthermore, should national authorities 'manifestly fail' to protect their population, UN member states agreed that 'the international community, through the United Nations…are prepared to take collective action, in a timely and decisive manner…on a case-by-case basis' (UN General Assembly 2005: 30).
Since its endorsement, R2P has dominated normative discussions around preventing and responding to genocide. This is reflected in the former UN Office of the Special Adviser on the Prevention of Genocide becoming the United Nations Office on Genocide Prevention and the Responsibility to Protect, for example. Substantial progress can be identified in the area of prevention (Stimson 2014). This includes an improved understanding of risk and resilience factors for genocide, greater capacity for and use of diplomacy, and other preventive strategies in at-risk nations, and several cases in which escalating risk of mass atrocities appears to have been stabilized and/or reduced through timely intercession. It is in the area of responding to genocide and mass atrocities, however, that the greatest limitations of R2P are apparent. The UN Security Council remains a central obstacle. Despite the mass atrocities associated with the violence in Syria since 2011, for example, the Security Council has not taken concerted action to protect the people of Syria. Similarly, the 'textbook' case of ethnic cleansing of the Rohingya in Myanmar did not lead to a meaningful response from the Security Council (UN Security Council 2017; UN News 2017). Historically, many genocides have proceeded uninterrupted, and where intervention has occurred it has often been motivated by strategic reasons rather than human rights concerns. It is unclear whether R2P will change that pattern. While there have been some examples of intervention since its advent, most notably in Libya (2011), for communities like the Yazidis, R2P has not led to protection from mass atrocities.
Arguably, R2P is further weakened by its passive conceptualization of vulnerable populations. As Mégret (2009: 579) has argued, R2P appears embedded in a paradigm ' of "salvation,"' in which ' outsiders' rescue ' others' or 'strangers.' He noted: Despite notable openings to civil society in the ICISS report (which do not find much of an equivalent in the much shorter World Summit Outcome document), it is as though formulations of R2P all stopped short of recognizing that 'victims' (or intended victims) of atrocities might have a role in averting atrocities at the point when they are being committed. (Ibid.: 580) Research demonstrates that vulnerable populations are far from passive, both prior to and during genocide (Mayersen 2015). They can be highly motivated to pursue selfprotection strategies, at both the individual and the communal level. This willingness can reduce reliance on the political will of external actors. A critical factor, however, is that the agency of vulnerable populations is typically heavily circumscribed by their situation. Nonetheless, historical examples demonstrate that, even in the most desperate of circumstances, vulnerable populations may retain some capacity to mitigate against genocidal outcomes (Mayersen 2016). The current conceptualization of R2P neither acknowledges, nor takes advantage of, this potential capacity.
Recognizing the agency of vulnerable populations, an alternative approach to preventing genocide and mass atrocities has focused on community-based interventions. An increasing body of scholarship documents the benefits of a localized, 'bottom-up' rather than 'top-down,' approach to prevention (Anderson and Wallace 2013;Barrs 2016). Bridget Moix (2016: 64-65), working with the NGO Peace Direct, has highlighted the success of community-based approaches to preventing mass violence in multiple locations. In Burundi, for example, local NGOs have collaborated to create an early warning/early response system (ibid: 64). In the Democratic Republic of Congo, hundreds of community mediators have been trained to resolve conflicts locally before they escalate (Peace Direct 2018). In Mauritania, village committees have been created to reduce the risk of mass atrocities (Strauss 2018). Unarmed civilian protection is also gaining traction as an approach to protecting civilians in violent conflict (Nonviolent Peaceforce 2020). The US government allocated funds in 2020 to ' consider the future use of unarmed civilian protection organiza-tions…to provide direct physical protection to civilian populations [and] to strengthen the local non-violent peacekeeping capacity of communities currently experiencing or at risk of violent conflict' (Nonviolent Peaceforce 2020a). Nonetheless, this is a nascent field, in which a strong evidence base has yet to be established.
Several typologies examining civilian self-protection during violence have been developed. Jose and Medie (2015) grouped self-protection strategies into three categories: non-engagement, nonviolent engagement, and violent engagement. Non-engagement includes flight, taking shelter, silence (such as 'playing dumb'), and strategic movement (such as moving between a home and safer locations to sleep at night). Nonviolent engagement includes negotiations with belligerents, paying taxes/tolls, adopting a position of neutrality, and peaceful resistance (Jose and Medie 2015;Anderson and Wallace 2013;Hirschel-Burns 2014). Violent engagement includes creating self-protection militias, joining armed groups, or supporting armed groups (Jose and Medie 2015: 525). Baines and Paddon (2012: 236) also identified three overarching self-protection strategies adopted by civilians 'to protect themselves and their families from violent attacks.' These are taking actions to appear neutral; avoidance tactics, such as running and hiding; and accommodation tactics, such as providing information or assistance to belligerents. Barrs (2010) has also examined how civilians survive violence, and developed a detailed inventory of tactics. Kaplan (2013) has developed a theory as to how civilians can influence norms of civilian protection during armed conflict. To date, however, these approaches have not been examined with respect to self-protection from genocide. The unique nature of genocide suggests that models focused on violent conflict more broadly, or even mass atrocities, may be of limited utility in understanding this specific crime.
This limitation is indicative of a wider issue within the scholarship: the conflation of 'genocide' and 'mass atrocities,' driven by the lexicon of R2P. The prevalence of the umbrella term of mass atrocities fails to recognize fundamental differences between mass atrocity crimes. Yet different types of atrocities might render very different approaches for self-protection appropriate. In a civil war in which crimes against humanity are occurring, for example, a strategy of neutrality or accommodation may have some prospects of success (Barrs 2010;Baines and Paddon 2012). During genocide, however, such strategies may prove fatal. Additionally, much of the scholarship on civilian selfprotection focuses on violent conflict more generally. While this offers many insights, it is potentially inapplicable to genocide, during which the violence is uniquely propelled by the desire to destroy a civilian population.
A further issue, identifiable in both scholarly and practitioner approaches, is inadequate distinction between prevention and mitigation. The phrase 'preventing and mitigating' mass violence, or mass atrocities, is used as a catch-all, sometimes without adequate recognition that the types of strategies appropriate to prevention may be entirely different from those appropriate to mitigation. Local initiatives to promote tolerance may be wholly appropriate as a preventive mechanism in states at risk of genocide, for example, but totally inappropriate for mitigating genocide in a state in which it is already under way. Perhaps in part as a result of these issues, to date there has been almost no research specifically into civilian self-protection during genocide, despite its ongoing use as a strategy by targeted populations (Hirschel-Burns 2014;Mégret 2009). This article focuses directly on this phenomenon.
Before proceeding, it is important to consider how communal self-protection during genocide is conceptualized. There are a number of definitions of civilian self-protection that can provide insight. Jose and Medie (2015: 516), for example, define it ' as activities undertaken during armed conflict (international or non-international) to preserve physical integrity in which the primary decision maker is a civilian or group of civilians.' Similarly, Baines and Paddon (2012: 234) ' define self-protection as the ability of civilians to assess and determine the best strategies to avoid, mitigate or thwart violence by armed groups.' Such definitions have close parallels with earlier work examining resistance to genocide. Gottlieb (1983: 40), for example, defined resistance as 'the attempt by the oppressed to limit, thwart, or end that oppression.' None of these definitions necessarily distinguish between nonviolent and violent activities, the latter of which is often described more narrowly as self-defence. While there may be a clear demarcation between the appropriateness of nonviolent and violent strategies for self-protection at the preventative stage, this distinction breaks down during genocide. Self-protection goals, such as personal survival, the survival of family and friends, and a reduction in oppression, may not be achievable through exclusively nonviolent measures. For the purposes of this article, therefore, communal self-protection is defined as cooperative communal activities undertaken by civilians to avoid or mitigate genocidal oppression. Such activities may incorporate violent and nonviolent measures, and may be undertaken independently, or in conjunction with outside assistance.
This article proceeds from the recognition that, despite R2P, vulnerable groups continue to experience genocide. In these circumstances, their only option other than to succumb to the violence may be to attempt self-protection. Yet there has been little scholarly analysis of how communities do so. To investigate this phenomenon, therefore, the following section of this article presents an exploratory multiplecase-study analysis. Pre-eminent examples of (attempted) communal self-protection during three cases of modern genocide are analysed: the Armenians at Musa Dagh during the Armenian genocide (1915); the Tutsi at Bisesero during the Rwandan genocide (1994); and the Yazidis at Mount Sinjar during the Yazidi genocide (2014). These cases have been carefully selected based on a number of criteria. First, each was wholly led by civilians. Cases in which armed groups (within the vulnerable population) were active prior to the onset of genocide (such as the resistance to the Armenian genocide in Van) were excluded due to this complicating factor. Second, cases in which communities were displaced prior to the onset of genocide (such as the Warsaw ghetto uprising) were also excluded, due to the negative impact of displacement on capacity for selfprotection (Baines and Paddon 2012). Third, numerically large examples were specifically selected. Their scope means they potentially offer the best insight into whether communal self-protection can be a viable strategy during genocide, despite the asymmetrical nature of the violence.
The methodology is specifically designed to explore communal self-protection in relatively conducive circumstances. This is based on two underlying assumptions. First, that if self-protection is a viable strategy, these large examples are those most likely to offer insights into what factors contribute to successful outcomes. Second, that if communal self-protection is not viable in these cases, it is unlikely to be viable in any circumstances. A limitation of this methodology, however, is that it does not consider small-scale attempts at communal self-protection, or evaluate their likelihood of success. Following the case study analysis, the article presents a typology of the strategies used in communal self-protection during genocide. It examines the prospects and limitations of communal self-protection, and the critical role of external support. The findings have significant policy-relevant implications for potential future genocides, and the prospects of selfprotection as a strategy to mitigate them.

Communal Self-Protection in Response to Genocide: A Case Study Analysis Case Study 1: Musa Dagh
The events of Musa Dagh stand out as the most famous attempt at communal selfprotection during the Armenian genocide. When the genocide commenced in April 1915, at first it appeared that the Armenians along the coastal areas of Aleppo -where Musa Dagh is located -would be exempt from deportation (Kevorkian 2011: 611). At the end of July, however, the seven Armenian villages scattered around the base of Musa Dagh received an order to prepare for deportation in a week's time (Thorgom 1915: 522). The villagers met, and 'sat up all night debating what it would be best to do' (Andreasian 1915: 514). The terrible experiences of a local reverend, Dikran Andreasian, who had previously been deported from another region but later allowed to leave the deportation columns after the intercession of American missionaries, helped the villagers decide on a course of action (Kevorkian 2011: 611). While each family made its own choice, the vast majority of the villagers decided to retreat to Musa Dagh and attempt self-defence rather than take their chances with deportation. Altogether, more than 4,000 villagers began to climb the mountain, taking with them a meagre supply of weapons, and all the food and supplies they could carry (Pambookjian 1973: 24-25;Andreasian 1915: 515;Darrieus 1915: 21).
Musa Dagh is a coastal mountain rising 915 metres (3,000 feet) above sea level. At the time of the genocide it was accessible through just five difficult mountain passes, known only to locals (Pambookjian 1973: 22). The Armenians had only just begun to establish themselves on the mountain when they faced the first attack from the Turks. A company of 200 soldiers advanced. After a day of battle, the Armenians were victorious, providing a great boost to their morale (Andreasian 1915: 515-16;Pambookjian 1973: 31). Two days later, however, the Turks returned with a much larger force. At dawn on a dark, foggy morning, the Turks mounted a massed attack, ' appearing at every pass in the mountain' (Andreasian 1915: 516). The Turks made substantial advances. The Armenians were scattered by the multiple fronts, and severely hampered by pouring rain, which wet much of their gunpowder and rendered it useless. Andreasian recalled: 'By sundown the enemy had advanced three companies through the dense underbush and forest to within four hundred yards of our huts. A deep damp ravine lay between, and the Turks decided to bivouac rather than to push on in the darkness' (ibid.). It was clear to the Armenians that a crisis point had been reached. The leaders of the group met, making a plan to silently encircle the Turkish camp and mount an attack in the dead of the night (Andreasian 1915: 517;Pambookjian 1973: 31-33). The surprise attack gave the Armenians the advantage. The Turks took heavy losses, and withdrew before dawn. The Armenians of Musa Dagh won a temporary reprieve.
Following this battle, both sides regrouped and developed new tactics. The Turkish forces recruited thousands of Muslims from the area and used them to develop a siege around Musa Dagh on the landward side. At the same time, they continued to make periodic attacks on the mountain. The Armenians used their growing battle experience to develop a more militaristic approach. A military body was created, military guards were systematically deployed on all fronts, further trenches and barricades were built, and systems for logistics and communications devised (Pambookjian 1973: 34). Using this approach, they were able to keep the Turks at bay during the continuing battles. At the same time, the Armenians reviewed their dwindling food supplies in light of the enveloping siege, and recognized their tenuous position. A constant watch was set up on the seaward side of the mountain, for any passing battleship (Andreasian 1915: 518;Pambookjian 1973: 35). Three swimmers were given a petition, sealed in a tin, in order that upon sighting a ship they might swim out and seek rescue for the group. Women made two very large flags, one white with a red cross, and the other with black lettering declaring -in English -'Christians in Distress: Rescue' (Andreasian 1915: 519). These were tied to trees in the hope that they might be seen by a passing ship. Day after day, however, the watchers reported only the empty sea.
On the fifty-third day of the battle for Musa Dagh, the French cruiser Guichen sailed nearby, and seeing the flags, came closer in order to read their message. The plight of the Armenians was then brought to the attention of the admiral of the fleet, who sought to permission to undertake their rescue. As this process was under way, however, the Turks launched a new attack on Musa Dagh, and the position of the Armenians -by now critically low on ammunition -became increasingly desperate (Pambookjian 1973: 48-49;Darrieus 1915: 15). Admiral Dartige du Fournet then ordered the evacuation of the Armenians himself, sending four additional ships to take them on board. After a difficult evacuation, it became apparent that no port wished to receive the refugees. Admiral du Fournet solved this problem by simply disembarking them at Port Said, much to the displeasure of the Egyptian authorities there (Varnava and Harris 2018: 852). It was in this way that approximately 4,200 Armenians refugees managed to safely escape (Thorgom 1915: 521). For the Armenians of Musa Dagh, communal self-protection was a strategy that helped save them from genocide.

Case Study 2: Bisesero
Whereas the arrival of the French at Musa Dagh led to the rescue of the Armenians there in 1915, in 1994 the arrival of the French at Bisesero led to the abandonment of Tutsi hoping to be saved from genocide. When genocide erupted in Rwanda on 6 April 1994, Kibuye prefecture in the west was almost immediately engulfed in the violence. Within just the first week of the genocide, more than 9,600 of the 75,000 Tutsi in Kibuye had been killed (Verwimp 2004: 236, 240). The Tutsi of Bisesero, a mountain range in western Kibuye, were the Abasesero, known for defending themselves from previous outbreaks of violence (African Rights 1999: 5-6;Wallis 2014: 150). At the outset of the genocide, large numbers of Abasesero fled their homes and took refuge on the steep hills of Bisesero, undertaking the largest attempt at resistance in Rwanda. There, they were joined by Tutsi from surrounding areas, who fled there seeking protection as the genocide progressed (Morel 2014: 107). At their height, numbers at Bisesero are estimated to have reached 50,000 (African Rights 1999: 2).
Among the hills, Tutsi hid in the forests, in caves, and in mineshafts (Wallis 2014: 151;Morel 2014: 107). Almost immediately, they were attacked, suffering heavy losses in early battles (African Rights 1999: 15). Survivor Uzziel Ngoboka recalled 'the assassins were armed with guns and grenades, and their supporters with machetes and clubs, whilst we had only stones and a few machetes to our name' (ibid.). The Tutsi regrouped, however, and began to develop strategies for more effective self-defence. These included stockpiling stones; mingling with the enemy so the latter could not use grenades or guns without risk to themselves; and positioning themselves on the most strategically advantageous hills (ibid.: [16][17][18][19]. At the same time, the Tutsi had to contend with a lack of shelter, heavy rain, and very little to eat. As April progressed, despite the relative success of the resistance, conditions became increasingly difficult. Yet as April gave way to May, the constant attacks seemed to stop. By this time, the majority of Tutsi in Kibuye had already been killed (Verwimp 2004: 240). At Bisesero, some began to think the genocide was over (African Rights 1999: 31).
For the Hutu militias, however, the break was only a pause, enabling them to develop a more effective genocidal strategy. On 13 May they launched a massive assault. Local militia were reinforced with killers bussed in from around Rwanda, many armed with guns in addition to machetes. One survivor recalled, 'there were too many heavily-armed militiamen and we had nothing to repulse the attacks with. Our spears and stones had no effect, so each person just had to find a way to escape' (ibid.: 35). It is estimated that in this single attack, between 3,654 and 25,000 Tutsi were killed (African Rights 1999: 35, 42;Verwimp 2004: 240). For the survivors, there was no respite, even as they searched for the bodies of their loved ones. Assaults continued almost daily, and attempts at resistance became increasingly desperate and futile.
In June, the situation became even more dire. Survivor Claver Habarugira, whose wife and four children had all been killed by this stage, recalled 'We had nothing to eat. We slept in the bush amongst the rotting bodies. We were dying of thirst and were obliged to drink water from the stream which had bodies in it' (African Rights 1999: 58). Hutu militia continued to attack, even as the Rwandan Patriotic Front gained control of more and more of the country, halting the genocide as they did so. Meanwhile, on 22 June, the UN Security Council approved 'Operation Turquoise,' a French-led humanitarian mission to Rwanda, that has subsequently been heavily criticized (Dallaire 2004;Jones 2006;UN Security Council 1999). Ahead of their anticipated arrival, the extremist Rwandan government provided extra support and resources to intensify the attacks at Bisesero (Morel 2014: 109-110). According to one source, they wanted to complete the annihilation of the Tutsi there, leaving no witnesses for the French to find (ibid.).
By the last week of June, there were only around 2,000 Tutsi still alive, trying to survive by hiding. Even as the first Operation Turquoise troops arrived in Rwanda on 23 June, the French military was already aware of the ongoing massacres in Bisesero (Morel 2014: 112). By 24 June, the first French troops reached Kibuye, but it was not until 27 June that they proceeded to Bisesero. When the soldiers arrived, the Tutsi came out of their hiding places, hoping for rescue. Survivor François recalled: 'They were the ones who summoned us with microphones, asking us to reveal ourselves and come out of our hiding places -we'd been hiding in some of the mine holes in Bisesero. They told us we had to come out since they were coming to rescue us, and intended to transport us to a place of shelter' (Wallis 2014: 152-153). Yet once many Tutsi had emerged from their hiding places, survivor Eric Nzabahimana recounted, 'The soldiers looked at us and asked us to continue hiding. They told us that they would come back in three days' (African Rights 1999: 61). After the French drove away, Hutu militia resumed their killing with a new ferocity. Many of the Tutsi who had revealed themselves in hope of rescue were slaughtered before they could even hide again. It is estimated around 1,000 of the 2,000 Tutsi still alive in Bisesero were killed between 27 and 30 June, when French forces returned (Morel 2014). The remaining 1,000 or so were the only survivors out of the 50,000 who tried to resist the Rwandan genocide there. While those at Bisesero lived longer than those caught up in the genocide elsewhere in Rwanda (Verwimp 2004: 239), ultimately their attempt at communal selfprotection was unsuccessful.

Case Study 3: The Yazidi Genocide
The most significant aspects of the genocide against the Yazidis in Iraq are the speed with which it took place, and the speed with which the international community responded. In 2014, the rapid rise of Daesh posed an everincreasing threat. In June, Daesh seized the city of Mosul, and the particular threat they posed to Yazidis became clear. Whereas other minorities were offered one of three options -(1) accept protected but inferior ' dhimmi' status and pay a special tax, (2) convert to Islam, or (3) face the death penalty -for Yazidis only the latter two options were provided (van Zoonen and Wirya 2017: 10). The Yazidis in Sinjar now found themselves in a region that lay in between Daesh-controlled areas in Iraq and Syria. The town of Sinjar, located at the base of the barren Mount Sinjar mountain range, and hundreds of surrounding villages, were home to a large portion of Iraq's Yazidi minority (Tharoor 2014; van Zoonen and Wirya 2017: 7). Most of the population, however, was reassured by two factors: the area was perceived as being of little strategic value to Daesh; and it was protected by the Peshmerga, the highly regarded Kurdish fighting forces (Shand 2018: 41).
On 3 August, the situation changed suddenly and radically. In the early hours of the morning, Daesh launched an attack against Sinjar from its base in Mosul. Acting in concert, hundreds of Daesh fighters first targeted the towns and villages surrounding Sinjar town. Peshmerga forces offered little or no resistance, in many cases withdrawing ahead of the Daesh advance (UN Human Rights Council 2016: 6). The Yazidis experienced this as a shocking betrayal, compounded by the fact that the withdrawal was undertaken without being communicated to the local population (UN Human Rights Council 2016: 6-7; van Zoonen and Wirya 2017: 10; Shand 2018: 43). The villages on the southern outskirts of Sinjar region were the first to be attacked. In Siba, several hundred Yazidi men banded together and attempted to fight off the advancing Daesh forces (Shand 2018: 43; UN Human Rights Council 2016: 7). Vastly outgunned, their resistance was quashed within hours (Shand 2018: 43). As the Daesh advance unfolded, mobile phone communication was critical to the ability of the Yazidis to respond to the rapidly changing situation on the ground. Yazidis were alerted to Daesh's advance and the withdrawal of the Peshmerga by the frantic phone calls of those from the first villages attacked, and by the rapid movement of villagers fleeing north.
Almost immediately, the consensus within Yazidi villages was to flee. Reports from the Daesh takeover of Mosul, and elsewhere, meant Yazidis were aware of the likely fate that awaited them under Daesh rule. The flight north quickly became a flight to Mount Sinjar. For the Yazidis, Mount Sinjar is a sacred place, and a place where they have sought refuge from persecution since the twelfth century (Shand 2018: 54). Yet the mountain could offer only limited protection. Barren, rocky, with few trees for shade and little water, it reached temperatures of well over 40 degrees Celsius (100 degrees Fahrenheit). Speed was critical for those who sought safety on the mountain (Amnesty International 2014: 8-16). Among those who escaped, the Yazidis who fled immediately upon the Daesh attack had the best chance of reaching the mountain. There was little to no time to gather supplies, and some villagers fled without food, water, or even shoes for the journey (UN Human Rights Council 2016: 7). Many were caught as they tried to escape (Human Rights Office of the High Commissioner for Human Rights [OHCHR] and United Nations Assistance Mission for Iraq [UNAMI] 2014: 13). The elderly, sick, and those otherwise unable to leave found themselves at the mercy of Daesh. Such was the speed of the advance that Daesh had captured the villages surrounding Sinjar town and entered the town itself by 10:00 a.m. In those same few hours, around 50,000 Yazidis fled to Mount Sinjar. 1 The situation for the refugees quickly became dire. They had little food or water in the harsh conditions. Children began dying from dehydration and heat exhaustion almost immediately. By 5 August, UNICEF reported that some 40 children had died on the mountain, with thousands more in need of urgent assistance (OHCHR and UNAMI 2014: 13). The Yazidis were besieged, surrounded by Daesh forces, including some fighters attempting to ascend the mountain and attack (ibid.). As The Times (2014) put it, the Yazidis were 'facing a grim choice: to die, croaking for water, or to stumble downhill and be murdered at the hands of savage jihadists.' But they did have one powerful weapon at their disposal: mobile phones. As long as their batteries lasted, they called whomever they could, begging for help.
The Iraqi government, quickly alerted to the plight of the refugees, began a humanitarian airdrop over the mountain on 4 and 5 August. As he spoke, the first US airdrops of emergency supplies had just been completed. They successfully delivered critical supplies of food and water to the besieged Yazidis.
Even before the US response, however, another international response had already commenced, that would prove even more critical to resolving the crisis. The Syrian Kurdish Party -Syria's largest Kurdish group -and its militant wing, the Yekîneyên Parastina Gel (YPG) (People's Defense Units), had been fighting Daesh in Syria since 2013. In response to a call from leader Abdullah Öcalan to protect the Yazidis, both the Partiya Karkerên Kurdistanê (PKK) (Turkish Kurdistan Workers' Party) and the YPG entered Iraq and commenced fighting Daesh there (Shand 2018: 91-92). Within a day, the YPG began to create an escape corridor for the besieged Yazidis on Mount Sinjar. US forces, working to prevent the further advance of Daesh towards Erbil, conducted airstrikes against the Daesh forces at the base of Mount Sinjar. This effectively contained Daesh and facilitated the YPG's efforts, although the US was unaware of this effect of their actions at the time. The YPG was able to open a safe passage that enabled the Yazidis to escape off the mountain, enter Syria, and then return to a nearby Kurdish-controlled region of Iraq (Amnesty International 2014: 26). Even as the US began to prepare for a logistically complicated attempt to create a safe passage, which Obama expected to take 'some time,' the YPG was guiding -sometimes literally carrying -thousands off the mountain each day (Shand 2018: 155). The following week, when US special forces representatives were airdropped onto Mount Sinjar to conduct an assessment of the situation and provide some desperately needed intelligence, they were shocked to find the siege had been completely broken (Shand 2018: 190-91;Kikoler 2015: 17). The US provision of humanitarian supplies, the US bombing of Daesh positions around Mount Sinjar, and the opening of the safe passage by the YPG had together enabled the vast majority of the 50,000 Yazidis trapped on Mount Sinjar to survive their ordeal and escape to relative safety. For the Yazidis from Sinjar, communal self-protection bought enough time for external interventions that saved tens of thousands of lives.

Communal Self-Protection during Genocide: A Comparative Analysis
Building on the case studies, this section will tackle two questions: Under what conditions can communities attempt self-protection during genocide? And under what conditions can communal self-protection from genocide offer a reasonable prospect of saving lives? The case studies highlight that the possibilities for self-protection during genocide are far more constrained than during other forms of violent conflict. Strategies common in other forms of mass atrocities and violent conflict are absent from these narratives, suggesting they were not accessible to targeted groups. These include numerous strategies identified in typologies of civilian self-protection. Remaining neutral, for example, was rendered impossible by the perpetrators through their deliberate identification of civilians as targets. This also ruled out options for nonviolent engagement, including negotiation, labour provision, or peaceful resistance. Current typologies for civilian self-protection during violent conflict therefore do not provide good models for understanding self-protection during genocide. To identify the circumstances under which communal self-protection during genocide can be attempted, in the following paragraphs I outline a typology focused on these specific circumstances, developed from the case studies above.

Communal Self-Protection Strategies: A Typology
The role of intracommunal communication and cooperation was critical in facilitating self-protection in each case study. In the Armenian case -in which there was the most time for a decision -communal deliberations took place. In the other cases, communication was less formal, but in each case a broader decision to resist was made, alongside families and individuals having the opportunity to make their own, autonomous decision. In each case, it is clear that the agency of the targeted group, and its ability to work cooperatively, played a key role in making communal self-protection a (temporarily) viable strategy. The groups were determined to survive, and displayed a capacity to respond quickly and creatively to facilitate their survival. At Musa Dagh and Bisesero, community meetings were held to develop strategies for resistance, and use whatever resources were available most efficiently. Leadership teams were appointed, and strategies were reviewed and revised following battles. At Musa Dagh, military guards were organized to protect all fronts, with replacements every 24 hours. Logistics were arranged to supply food and water to the soldiers, and the 25 best fighters formed a group that could be dispatched to reinforce any unit in distress and in need of support (Pambookjian 1973: 34). At Bisesero, the tactic mwiuange sha, 'go and merge,' was developed. This involved lying down on the battlefield and mingling with the attackers, who were then unable to use grenades or guns without risking their own lives (African Rights 1999: 16-19).
On Mount Sinjar, the extreme speed and unexpectedness of the flight meant that the Yazidis had almost no resources. Nevertheless, they used their mobile phones to great effect. Yazidis trapped on Mount Sinjar rang their Yazidi member of the Iraqi parliament directly, begging for help (Shand 2018: 93-94, 98). Vian Dakhil's subsequent impassioned plea for help on the floor of the Iraqi parliament went viral on YouTube (Shand 2018: 98). Yazidis rang relatives and friends outside of Iraq, begging for help. These calls brought their plight to the attention of expatriate Yazidis in the US, who immediately began advocating for US assistance on their behalf. These strategies of intracommunal communication and cooperation helped the groups survive, at least temporarily.
Local knowledge appears to have played a role in facilitating the communication and cooperation necessary to enable communal self-protection. For the Yazidis, Mount Sinjar had been a place of refuge since the twelfth century. A small group of Armenians had previously taken a stand on Musa Dagh during the massacres in the 1890s, albeit unsuccessfully on that occasion; in Rwanda, the Abasesero had successfully resisted violence at Bisesero twice before in previous decades. This suggests that being able to draw on a past history of communal cooperation, and local knowledge of potentially effective selfprotection strategies, can be helpful. Local geographical knowledge -typically shared by the persecuted group, but difficult to access for the perpetrators -may also prove advantageous.
Avoidance strategies played a vital role in enabling each group to attempt communal self-protection. It is a striking similarity, and far from coincidental, that each sought refuge atop a mountain. This provided a short reprieve from engagement with the perpetrators. It enabled them to flee an initial, or anticipated, assault, providing a small window of opportunity for cooperation and the development of other strategies for selfprotection. Escape is well-recognised as a potentially effective strategy for civilian selfprotection during genocide. As Valentino (2004: 250) has noted, 'the ability of potential victims of mass killing to flee across borders probably has averted millions of deaths in this century.' Yet in many circumstances, including the three case studies herein, groups are unable to escape in this way. At Musa Dagh and Mount Sinjar, the Armenians and Yazidis were besieged. At Bisesero, some survivors of the massacre on 13 May tried to leave with the goal of reaching Burundi but were forced to return -there was simply no way out of the region (African Rights 1999: 37-38). Local avoidance strategies provided some reprieve. The specific goal of genocide to target civilians, however, meant that avoidance strategies were a very short term measure for self-protection.
Resistance was a key strategy utilized by the Armenians at Musa Dagh and the Tutsi at Bisesero. Many communities lack any means with which to resist during genocide, but in these cases the topographical advantage of the mountains played a substantial role in enabling resistance. They provided an area in which each targeted group could concentrate its forces and be numerically dominant, and in which there was space for some degree of safe, localized retreat from specific attacks. Any attack was intrinsically at a strategic disadvantage. Persecuted groups could see attackers approaching, then use their heightened positions and local geographical features to aid their defence. That the mountains themselves proved so crucial, however, is telling. It highlights how systematically these persecuted groups had been deprived of any other potential tools they might have used for self-protection. By the time of genocide onset, vulnerable groups have typically been disarmed, isolated, and sometimes financially ruined by a rising tide of persecution. Very often, community structures and leadership have been deliberately disrupted, sometimes with the removal of battle-age men. A whole range of resources the community might have used in pursuit of selfprotection have been made unavailable. In these cases, the groups each had access to a powerful resource that the perpetrators were unable to control. That it was such a crucial feature of each attempt at self-protection, however, suggests that absent such a resource, attempts at self-protection might not be feasible.
Avoidance and/or resistance did not lead to sustainable living conditions for the groups attempting communal self-protection. Despite the tremendous agency and resourcefulness of the Armenians, Tutsi, and Yazidis, it is clear that their circumstances were dire. In each case, communal self-protection can be described, at best, as a viable strategy on a temporary basis only. The Armenians at Musa Dagh were aware that -absent a turn in the tide of the war -their only hope in the longer term was for rescue. They embarked on multiple strategies to try and bring their plight to the attention of the international community, and it was only due to their eventual success in doing so that they survived. By the time they were rescued, their situation was critical. At Bisesero, the constant attacks took their toll. While the Tutsi could defend themselves with some success in the first month of the genocide, the massacre on 13 May was a turning point. Many thousands were killed on that day alone. For those who managed to survive, self-defence became increasingly difficult. On Mount Sinjar, the survival of the Yazidis was exceptionally time-critical given the absence of sufficient water. Communal self-protection offered a temporary reprieve from genocide, but in no case did it offer sustainable living conditions.
For each group, the ultimate success of their attempt at communal self-protection was contingent on their ability to obtain external rescue. The speed of this intervention was crucial. At Musa Dagh and Mount Sinjar it is not an exaggeration to say it came at a critical time, and that in either case had it been a few days or a week later, there might have been few still alive to rescue. At Bisesero, by the time the French arrived in the region, only four percent of the Tutsi were still alive. The French failure to immediately offer protection meant that by the time they returned, half of that number again had been killed. Yet one can also recognize that had a rescue mission been able to reach Bisesero before 13 May, potentially up to 50,000 Tutsi could have been saved. The Tutsi there were able to extend their survival by several weeks, but in the absence of external rescue it was not enough.
A typology of communal self-protection during genocide can therefore identify four key strategies: intracommunal communication and cooperation, avoidance, resistance, and appeals for external rescue. Unlike other typologies for civilian self-protection during violent conflict, however, this typology does not represent a 'toolbox,' from which individual strategies might be selected and used successfully. Rather, the options for self-protection during genocide are far more constrained. Communal self-protection can only be feasibly attempted in circumstances conducive to the employment of multiple strategies in close succession. Even when two or more of these strategies are deployed simultaneously, the targeted and deeply asymmetrical nature of genocide means that self-protection is only temporarily viable (at best). In each of the cases in which communal self-protection did save thousands of lives, it did so only due to the timely rescue of the targeted group by an external party. To some degree, successful communal selfprotection can and should be identified upon the basis of lives saved, and by this criterion two of the three case studies do offer examples of relative success. The highly contingent nature of these positive outcomes, however, necessitates deeper examination.
This brings us to the second question at the heart of this study: 'Under what conditions can communal self-protection from genocide offer a reasonable prospect of saving lives?' This case study analysis suggests that the answer is almost wholly dependent upon the prospects of external intervention. Yet, historically, political will for external intervention has been a critical problem preventing effective responses to genocide (Power 2003). It is far from clear that R2P has improved the prospects of meaningful international action in response to genocide. Moreover, the timecritical aspect of rescue for targeted groups undertaking self-protection is at odds with the often very slow pace of international deliberations and agreement on responding to genocide, as the international response to the Rwandan genocide overall so tellingly demonstrates. A further complication is that any rescue effort is likely to result in the rescued population becoming refugees. The current international climate surrounding refugees means that this issue could potential derail a planned rescue mission in the future. Any rescue of a vulnerable group attempting self-protection during genocide is thus contingent on a range of factors that have proven problematic in past instances, and it is reasonable to conclude that such rescue may not reliably be forthcoming. In short, this study has found that there are unlikely to be any conditions in which communal self-protection during genocide offers a reasonable prospect of saving lives.
These findings have significant policy implications. First, there is a need for universal recognition that communities experiencing genocide do not have the capacity to mitigate their circumstances. Their only prospects are to appeal for external assistance, or attempt escape (which may not be possible). Meaningful responses to mitigate genocide require external intervention by international actors. Second, the findings highlight the need for a much clearer distinction in policy circles between prevention and mitigation of genocide. Strategies appropriate for the prevention of genocide, including community-based initiatives, are fundamentally inappropriate for responding to a genocide that is under way. The often-used phrase 'prevention and mitigation' conflates these two vastly different circumstances, each of which require very different responses. Finally, there needs to be much greater recognition of the differences between R2P crimes, and the need for individualized approaches to each. Community initiatives to mitigate against crimes against humanity during outbreaks of violent conflict may have some prospect of success, for example, but are not appropriate responses to genocide. The particularities of genocide require tailored research and policy responses that cannot be adequately addressed within the umbrella of 'mass atrocities.'

Conclusion
Despite the rhetoric of R2P, the reality for vulnerable groups like the Yazidis in Iraq continues to be a lack of protection from genocide. With contemporary normative discussions on responding to genocide heavily focused around R2P, however, there is little research into how communities respond when the international community fails to meet its protection responsibilities. This paper examined three case studies in which communities attempted to mitigate genocide though self-protection. For the Armenians at Musa Dagh, the Tutsi at Bisesero, and the Yazidis at Mount Sinjar, this process involved employing a series of strategies. These included a combination of intracommunal communication and cooperation, avoidance, resistance, and appeals for external rescue. The case studies demonstrate that the ability to attempt communal self-protection required some degree of special circumstance, such as a protected location, like a mountain; and/or a history of or strong capacity for cooperation. Even in these circumstances, communal self-protection could offer a temporary reprieve from genocide at best, but in no case could it offer a lasting solution. In both cases in which communal self-protection during genocide saved substantial lives, it did so because the targeted group was rescued by external interventions. This highlights the fundamentally precarious nature of communal self-protection as a strategy to mitigate genocide.
Leaving people with no choice but to attempt self-protection during genocide is all but condemning them to death. While selfprotection may be an option with reasonable prospects of success at the preventative stage, and may also have some prospects of success during other forms of mass atrocities, communal self-protection is rarely a viable option during genocide. It is only in extraordinary circumstances that it can be undertaken, and only in even more exceptional circumstances that it can offer protection from genocide. Even in a best-case scenario, that protection is temporary, precarious, and ultimately dependent on external salvation. Given that, in our time, that scenario itself arises from the failure of the international community to meet its responsibility to protect, the prospect of such salvation being forthcoming is likely to be poor. In short, communal selfprotection during genocide must be recognized for the drastic measure of last resort that it is. It does not offer a reasonable prospect of success. If a state does not meet its responsibility to protect its population, the international community should take concerted action well before a vulnerable group is forced to resort to such extreme measures. In the event that a civilian group finds itself in this position, however, the international community must recognize the precariousness and time-critical nature of its position, and be prepared to act immediately to save as many lives as possible.

Note
1 This is the most common estimate; however, there is substantial uncertainty around the precise numbers.