The waxing and waning of ethnic boundaries: violence, peace and the ubwoko in Burundi

Abstract

Violence based on identity constructs reinforces the experience of ethnic boundaries as felt distance between in-groups and out-groups. But what makes such an experience of rigid ethnic boundaries fade or disappear, if anything? We examined this in Burundi, a country characterised by repeated episodes of violence between Hutu and Tutsi since independence. We analysed the waxing and waning of ethnic boundaries through the (life) stories of 202 individuals collected through an iterative research process in two rural villages that were seriously touched by (ethnic) violence. Rigid boundaries between ethnic in- and out-group appeared to fade through non-violent interactions; when categorisations other than ethnic emerged; and when awareness of interstitiality, being in-between salient groups, contested the relevance and meaning of the ethnic boundary as such. These insights invite us to bring in multiple temporalities and identities when aiming to understand legacies of violence in conflict-affected societies such as Burundi. This would allow us to avoid treating groups as substantial entities, which reinforces boundaries between in-groups and out-groups.

INTRODUCTION

In 1969, Barth's Ethnic groups and boundaries challenged the essentialist conception of ethnicities. This work emphasised that ethnic boundaries should not be considered as fixed and well-established as suggested by primordial conceptions. Since it was ‘the ethnic boundary that defines the group, not the cultural stuff that it encloses’ (Barth 1998: 6), attention should be paid to the changing meaning and nature of ethnic boundaries. Ethnic groups thus started to be seen as social constructs, products of social dynamics that varied according to contextual factors. Barth's approach has been so influential that, nowadays, the use of a constructivist approach in the study of ethnicity is well-established (Brubaker 2004: 64; Jenkins 2008: 18; Eriksen & Jakoubek 2019: 7).

Yet, many analysts and analyses have not taken the social constructivist approach to its logical end point. A genuine social constructivist perspective implies that the experience of boundaries in social life can vary in intensity. From a social constructivist perspective, social groups do not appear and disappear but they are transformed into something experienced as a collective demarcated by an imaginary boundary, and vice versa.

As defined by Wimmer (2008: 975), a boundary ‘displays both a categorical and a social or behavioural dimension’. The former refers to acts of social classification and collective representation; the latter to ‘everyday networks of relationships that result from individual acts of connecting and distancing’. Boundaries ‘draw the line that delimits an imagined community of “people like me”’ (Lamont 2000: 3), which always implies the acknowledgment of those who do not belong to that community: ‘each identification (“I am Swiss”) … implies a categorical boundary (the non-Swiss)’ (Wimmer 2013: 3). Boundaries are not necessarily barriers: they do not always ‘imply closure and clarity’ (Wimmer 2013: 10) leading to cleavages. In the first place, boundaries allow an ‘us’, an in-group, to distinguish itself from a ‘them’, an out-group. This responds to the human need for identification and belonging and does not automatically imply the existence of tensions or conflict between ‘us’ and ‘them’. In fact, boundaries can be more or less thin or thick, and present different levels of rigidity. The thicker the boundary, the bigger the perceived distance between ‘us’ and ‘them’. The thinner the boundary, the closer people result to be.

This paper aims to empirically examine this process, and especially its unexplored dimensions, in the case of Burundi, where Hutu, Tutsi and Twa, the three social categories usually referred to as ethnic groups, share the same language, culture and territory (Chrétien 1985: 129).¹ Ethnicity in Burundi is a social construct that results from specific social processes and is adopted by individuals as an identity reference. The sense of ethnic belonging varies among individuals and in time: the ‘content and grip’ of ethnic categorisations ‘on individuals’ imaginations are a function of social and historical conditions’ (Fearon & Laitin 2000: 858). Burundi thus illustrates well how ‘ethnicity is a matter of social organisation above and beyond questions of empirical cultural differences’ (Barth 1998: 6), therefore requiring a social constructivist approach in the study of ethnicity.

A scholarly consensus holds that political violence with an ethnic character creates or reconfigures the experience of the bounded nature of ethnicity by increasing the sense of belonging to a distinct group and facilitating the differentiation from outgroup members (Smith 1981; Appadurai 1998; Brubaker & Laitin 1998; Wood 2008: 547–50; Wimmer 2013: 70). A similar consensus exists regarding the link between political violence (physical acts and discourses), and the memory thereof, and ethno-genesis in Burundi (Chrétien et al. 1989: 51; Malkki 1995; Lemarchand 1999: 10; Uvin 1999: 265; Arusha Peace and Reconciliation Agreement for Burundi 2000; Chrétien 2008: 41).

Political violence in Burundi is related in an important way to the policies implemented by the Belgian colonial power holders during their indirect rule (1919–1962), which privileged the Tutsi in the access to power and education to the detriment of the Hutu. This situation was maintained and reinforced after independence. In the following decades, every attempt of Hutu revolt was harshly repressed by the Tutsi ruling class. In 1972, Hutu insurgents started a rebellion in the South of the country. The brutal repression by the army led to hundreds of thousands of deaths.² Educated, wealthy and power-holding Hutu were targeted. Some authors do not hesitate to classify what happened in 1972 as genocide (Lemarchand 2002: Chrétien 2008: 59). In 1988, further violence took place in the northern communes of Ntega and Marangara, following the killing of local Tutsi administrators by some Hutu. The government's reaction was once again brutal, targeting Hutu (Lemarchand 1996: 126). Similar violence followed in other regions in 1991 with over a thousand victims (Thibon 1992: 156).

In 1993, after a particularly ethnicised campaign (Reyntjens 1993), Melchior Ndadaye became the first democratically elected Hutu president of Burundi. Three months after he took office, Ndadaye was killed in a military coup. The event unleashed political violence, once again along ethnic lines, leading to a civil war in which hundreds of thousands of people found their death. In 2000, the signing of the Arusha Peace Agreement, followed by two ceasefires in 2003 and 2008 between the government and rebel groups, represented the beginning of the end of the 1993 civil war and decades of large-scale violence in the country. In 2015, renewed large-scale violence took place in the context of presidential elections. If some underlined the fact that among protesters were Tutsi as well as Hutu (Van Acker 2015: 8; Purdeková 2019: 31), others claimed that it was an attempt at regime change in Burundi, supported by ‘the Tutsi’ not wanting to accept the same Hutu president any longer (Kavakure 2016; Ndayicariye 2020).

In sum, since independence, the meaning of ethnicity and the relationships between members of different ethnic categories in Burundi have been shaped by episodes of mass categorical violence, defined as ‘large-scale, group-selective violence’ (Straus 2015: 17) directed towards members of a specific ‘category’. Academic research on Burundi mostly analysed the relation between ethnicity, politics, and conflict,³ while a minority of studies took into consideration bottom-up perspectives or analysed the ways in which people relate to their ethnic identities in daily life (Malkki 1995; Ingelaere 2009; Turner 2010; Berckmoes 2014; Samii 2013; Sommers 2013). Therefore, the focus of this paper on the salience of ethnicity outside the institutional or political sphere is not only a way to recognise that people's daily activities are equally ‘constructive’ of ethnicity (Fearon & Laitin 2000: 855), but it also fills a gap in academic literature on Burundi.

Additionally, in Burundi and beyond, both empirically and theoretically, little is known about the social processes that diminish the feeling of belonging to a distinct ethnic group and of antipathy to ethnic outsiders (Brubaker 2004: 19; Wimmer 2008: 976). Building on this theoretical and empirical consensus regarding the link between political violence and the salience of ethnicity, on the one hand, and on this knowledge gap regarding the processes that drive the decline of the experience of rigid boundaries following the exposure to violence, on the other hand, the aim of our paper is twofold. First, we illustrate and further examine the emergence and salience of ethnicity in the context of political violence. Second, our main aim is to identify the main drivers in the social processes that reduce the salience of ethnic boundaries across space, time and ethnic belonging. Thus, we ask: what makes perceived boundaries scripted by the experience of violence disappear, if anything? Methods and data are presented in the next section. The following two sections then present the findings of our analysis. The penultimate section discusses the notions of multiple temporalities and identities in the study of Burundi and beyond. A final section concludes.

METHODOLOGY

In order to arrive to a bottom-up and in-depth understanding of micro-dynamics, extensive fieldwork was conducted in two rural sites called Bugendana and Mugara in 2008, 2015 and 2018–2020. The field sites were selected based on their relevant and diverging histories of violence and their consequences.

In 1993, when the persecution of Tutsi by Hutu started after the assassination of President Melchior Ndadaye, Tutsi living in Bugendana fled their homes seeking refuge in the neighbouring hills, becoming IDPs (internally displaced persons). This was followed by an attempt to ‘restore order’ by the army, which was mainly composed of Tutsi and perceived as pro-Tutsi. In 1994, in Bugendana, an important IDP camp was built to gather all the IDPs from the neighbouring localities. In 1996, Hutu rebels (the so-called assaillants) attacked the IDP camp, and more than 600 Tutsi IDPs lost their lives in one night. At least until the last period of fieldwork, the IDP camp was still inhabited.

In 1972, Hutu insurgents were said to have started a rebellion against Tutsi domination in Mugara. The brutal repression led by the Tutsi-dominated military targeted the said rebels (called abamenja) and all Hutu that were educated, wealthy and in positions of power throughout the country. To escape death, hundreds of thousands of people, mostly Hutu, took refuge outside Burundi. In Mugara, most if not all Hutu inhabitants fled. In their absence, mostly Tutsi came to take advantage of their land. With the end of the 1993 civil war, Hutu who had taken up refuge abroad returned, with more than 500,000 refugees returning to Burundi between 2002 and 2011 throughout the country (Rema Ministries 2012: 19).

In total, we interviewed 202 individuals across both sites, 117 of them both in 2008 and 2015, and 22 of them only once, either in 2008 or in 2015. Sixty-three individuals were added in 2018–2020 (the variation in the respective years is clarified below). They belonged to different ethnicities (Hutu and Tutsi), gender and age categories. Different experiences of violence allowed them to be identified as ‘IDP’ (internally displaced person), ‘former IDP’, ‘never displaced’, ‘returnee’, ‘former prisoner’, ‘demobilised’ (see Table I). The fact that data were collected in different moments in time, some of which coincided with increasing political tensions in the country, did not have a significant influence on the type of life stories collected, as the latter concerned events of a relatively distant past. Our iterative research approach aimed to reinforce interviewees’ trust in the researchers, for the interviewees to feel more comfortable when talking about their own lives.⁴

Table I. Overview of the interviewees (Bugendana and Mugara, 2008–2020).

Fieldwork was conducted by three different teams of researchers.⁵ During each period of fieldwork, we strove to have a balanced composition of the research teams in terms of gender and ethnicity. The identity of each member of the research team – revealed (gender, skin colour) or guessed (ethnicity) – can have an influence on the interaction between interlocutors. These influences can never be excluded but the mixed composition in each fieldwork period and across respondents rules out systematic biases.

In 2008, respondents were selected through a random sampling scheme stratified by a number of identity markers including ethnicity. The research design and approach is discussed in detail in a number of publications (Ingelaere 2009: 30–8; Guariso et al. 2018: 1367–73; Ingelaere & Verpoorten 2020: 525–7).⁶ In 2018–2020, starting from the initial sample, additional interviewees were selected through snowball sampling. During all periods of fieldwork and in all cases, we followed the self-identification of the respondent and excluded people for which we could not establish the ethnicity for further analysis.

During the three research periods, data were collected through semi-structured interviews, life histories in particular. A life history is ‘a purely subjective account – a detailed perspective on the world’ that functions as a social commentary, communicating meaning (Plummer 2001: 20). It is thus an exquisite tool to approach ‘ethnicity as cognition’ since they allow to focus on ‘perceptions, interpretations, representations, classifications, categorizations, and identifications’ (Brubaker et al. 2006: 79). All interviews were conducted at the house of the interviewee, to allow him or her to speak (more) freely. Life histories collected in 2008 and 2015 first inquired about basic demographic information and migratory movements and were subsequently structured around six themes: socio-economic situation, feelings of security, trust in members of the ‘other’ ethnicity,⁷ trust in members of their own ethnicity, political representation and expectations for the future. When interviewees pointed out changes of the situation for each of these themes, they were asked to explain the reasons for these changes. A variation of this interview format was adopted in 2018–2020. Following questions on places of residence, we asked about life experiences: education, work, travels and experiences of violence. Then we enquired about contemporary daily activities. The aim was to observe who emerged as ‘other’ in the interviewee's narration, to detect when and how the ‘others’ were described in terms of ethnicity. For this purpose, we also asked for definitions of ‘friend’ and ‘enemy’, to understand what figures were perceived by the interviewees as close (‘friends’), and who was placed among the ‘others’ (‘enemies’). This allowed us to observe when and why ‘friends’ were members of the same ethnicity.

The empirical material – interviews and field notes – were analysed with the help of Nvivo. Analysis was conducted in an inductive way with the help of few ‘sensitising concepts’ borrowed from the literature on boundary-making and -unmaking (Lamont 2000; Brubaker 2004; Wimmer 2013).⁸ Sensitising concepts are not ‘definitive concepts [that] provide prescriptions of what to see’ but they ‘suggest directions along which to look’ (Blumer 1954: 7). Evidently, people do not speak in terms of, or act along, visible boundaries in everyday life. Therefore, we aimed to identify boundaries and their changing nature (emergence, dissolution, salience, rigidity, thickness) by looking for experiences, perceptions and behaviours that revealed distance and proximity, and distancing and rapprochement, between categories of people. In doing so, we looked for (changing) indications of ‘the emotionally laden sense of belonging to a distinctive, bounded group, involving both a felt solidarity or oneness with fellow group members and a felt difference from or even antipathy to specified outsiders’ (Brubaker 2004: 19). The research methodology allowed us to probe and analyse these processes by examining (changing) patterns of interactions observed, and answers to questions regarding the (changing) nature of interpersonal trust, feelings of fear and security, or identifications of friends and enemies. It is important to note, however, that we have not adopted a linear (time) or systematic comparative (space) approach in the analysis and presentation of the findings; we made the waxing and waning central to the analysis, irrespective of time and place, and across ethnic groups, our main concern being the ‘how’ and ‘why’ questions. Processes of boundary waxing and waning evidently present different characteristics in different contexts. The analysis of data from different sites in Burundi allowed us to detect common features of these processes and thus identify underlying processes of boundary-making and -unmaking, irrespective of their development through time or in a specific context.

In Kirundi, the main language spoken in Burundi, Hutu, Tutsi and Twa are referred to as ubwoko (plural: amoko). Ubwoko is translated as ‘category’, ‘sort’, ‘variety’ and can be applied to tree and mineral species as well as to other types of classifications (Mworoha 1987: 96). During colonial times, the term ‘ethnicity’ was applied to Burundian amoko by Western observers during their study of the Burundian population, when the term ‘ethnicity’ was taking hold in their own scientific milieus.⁹ In light of this colonial influence, we prefer to use the term ubwoko (sing.) and amoko (plural) when presenting, discussing and analysing the narratives of respondents. Ultimately, we aim to understand what the ubwoko represents for Burundians, therefore leaning upon the ‘semantic plasticity’ (Saur 2014: 138) of the term, the translation of which oscillates between ‘category’, ‘type’ and ‘ethnicity’. Only when the use of the term ubwoko would lead to awkward wording, we employ ‘ethnicity’ or ‘ethnic group’, keeping these terms in quotation marks.

BOUNDARIES (RE-)EMERGING AND PERSISTING: ACTS, MEMORIES AND DISCOURSES OF/ON VIOLENCE

Direct experience of violence is an important cause of inter-group distancing: the perception of members of the ‘other’ ubwoko is directly affected by it. Violent experiences represent traumatic events that come abruptly into someone's life and provoke a shock, often with lasting consequences.

I lost all the trust that I had in the Hutu, considering the nastiness with which they chased us away [in 1993]. (Tutsi woman, 52, IDP, Bugendana, April 2008 Int.)

The Tutsi always wanted to threaten the Hutu, and there have been many victims during the killings. They exterminated my family, my husband and my friends. (Hutu woman, 54, returnee, Mugara, 2008 Int.)

How could I trust [the Hutu] when it is them who caused my fall? … when it is them who are at the origin of my pain, who have killed my husband, and looted all my belongings? (Tutsi woman, 54, former IDP, Mugara, 2015 Int.)

The memory of these experiences reinforces the experience of boundaries when it is recalled to interpret later episodes of violence, or to anticipate future (imagined) violence. When possible outbreaks of violence are foreseen, some rely on experiences of past violence to make sense of it. Then, an ‘ethnic’ reading of the situation is (re-)activated. This underlines the nature of the ubwoko as ‘intermittent phenomenon’, which ‘happens at particular moments, and in particular contexts, when [people] interpret their experience or diagnose situations … in ethnic terms’ (Brubaker et al. 2006: 208).

In this way, violence becomes an ‘absent presence: a recollection of past violence and an imagination of future violence … constantly present as a structuring force in social life’ (Hermez 2012: 330). Anticipation of violence refreshes the past experience of violence, which leads to inter-group distancing, and in this way it contributes to maintain the experience of a rigid boundary between what is experienced as ‘groups’. Assumptions about the other ubwoko are based on the experience of past violence, thus personal memories. In daily situations, people are thus required ‘to “fill in” unspecified information from their stocks of tacit background knowledge’ (Brubaker 2004: 76).

When probing for reasons why intergroup distancing happens, our interlocutors also evoked violent discourses, which reminds us that ‘threats of violence are also violence’ (Galtung 1990: 292), verbal violence thus being another reason for boundaries thickening. Verbal violence seemed to be mainly motivated by, or linked with, political competition: ‘The electoral campaign was conducted in a suspicious way, the Hutu were throwing threatening slogans against the Tutsi and the members of the UPRONA’ (Tutsi woman, 48, IDP, Bugendana, 2008 Int.); ‘Trust decreased sharply following the beginning of the havoc [created by] political parties. Bad words were pronounced by the Tutsi’ (Hutu man, 50, never displaced, Bugendana, April 2008 Int.).

Violent discourses are closely linked to discourses on distant violence, sometimes transmitted to the next generations. Both contribute to increased experiences of rigid boundaries between amoko. Memory of the past gives scripts for the interpretation of more recent episodes of violence. What it means to be Hutu or Tutsi is often based on narratives, oftentimes around the conflictual relationship between amoko (Gatugu 2018). Because past violence caused inter-group distancing, the interpretation of more recent violence based on the memory of the past contributes to reinforce the distance between amoko.

[In 1992] Hutu and Tutsi were throwing insults to each other, related to politics and to the 1972 events. (Hutu man, 48, never displaced, Bugendana, March 2008 Int.)

The conflict was assuming an ethnic nature because they said that the Tutsi killed the president and we thought that we would have returned to the situation of 1972. (Hutu man, 68, former IDP, Mugara, April 2008 Int.)

Throughout the country, as these quotations illustrate, the ‘events’ of 1972 created ‘sufficient fear to suppress Hutu unrest for two decades’ and ‘crystallized Hutu and Tutsi identities’ through ‘a climate of permanent mutual fear’ (Uvin 1999: 258). Through the threat of genocide, ‘the state constituted the Hutu as a singular community’, of which they had been hardly aware previously (Russell 2015: 441–2). In fact, the nature of violence ‘is not intrinsic to the act itself; it emerges through after-the-fact interpretive claims’ (Brubaker & Laitin 1998: 444). Thus ‘violence generates mythmaking, which itself becomes a constitutive element of further violence’ (Lemarchand 1996: xi). This shows how belonging to an ubwoko is shaped by both human actions and speech (Fearon & Laitin 2000: 848), speech being actually a type of human action. It is after-the-fact interpretative historical claims, intended as discourses elaborated on violence, either directly experienced or not, that provoke inter-group distancing. When asked to clarify why interpersonal trust toward people of the other ubwoko declined or did not increase, some interviewees explained:

They said that it was the Tutsi who did not want the Hutu in power. (Hutu woman, 48, never displaced, Bugendana, 2008 Int.)

I was afraid that the Hutu could attack me because they considered all the Tutsi as enemies who had killed the president. (Tutsi woman, 52, former IDP, Mugara, April 2008 Int.)

The Tutsi are like snakes. They can show you that they trust you, while it is not true. (Hutu man, 50, former IDP, Mugara, 2008 Int.)

These types of narratives closely resemble what Malkki called mythico-histories (1995: 54), which provide visions of the world and moral evaluations of past events and contemporary ‘others’. It is from the transmission of these types of narratives that persons of different ubwoko arrive at diverging accounts on the same violent events (Lemarchand 1996: 33; Mukuri 2004: 427; Ingelaere 2009: 123–36; Manirakiza 2011: 39; Taylor 2013: 456). Diverging narratives of violence, each specific to either the Hutu or the Tutsi ubwoko, show how boundaries can be made through discourses and emerge in a discursive way.¹⁰ As Brubaker (2004: 17) explains, ‘ethnicity, race and nationhood are fundamentally ways of perceiving, interpreting, and representing the social world. They are not things in the world, but perspectives on the world.’ It is in the discourse on violence and through violent discourses that an antipathy towards marked others emerges and boundaries are established and maintained.

BOUNDARIES FADING, BLURRING AND CONTESTED

According to our analysis, boundaries between amoko can become thinner, resulting in a rapprochement between people, in three main ways: when non-violent interactions take place, when ‘non-ethnic’ categorisations and identifications occur, and through people that occupy positions ‘in between’ what is experienced as demarcated collectives of people.

Non-violent interactions: spontaneous or organised

Violence left an important mark on the territory of Burundi, often leading to the creation of ‘ethno-geographic’ settings mainly inhabited by members of the same ubwoko, such as IDP camps (as in Bugendana), or specific neighbourhoods in Bujumbura.¹¹ In addition, violence led to massive exoduses of people at different moments in time, where they lived and some still live as refugees, mostly in neighbouring countries. In these relatively closed settings, either inside or outside the country, memories of the past can be told and circulate more freely than in a setting inhabited by members of different ubwoko, and thus reinforce group identity, as evoked above. Post-war socio-spatial experiments entail large-scale policy interventions that move or replace people ‘in the name of coexistence, integration and sharing after war’ (Purdeková 2017: 534), which does not happen without friction or outright resistance. In any case, returning to the home country or leaving the site of displacement inside the country often only happen in the absence of open violence, allowing returnees and former IDPs to gradually resume previous life routines without being afraid of losing their life.

In Bugendana, violence led to clear-cut physical demarcations of Hutu and Tutsi areas. To this day, the IDP camp remains almost entirely populated by Tutsi, while the hills surrounding the camp are populated by Hutu, as very few Tutsi returned to their homes on the hills. In Bugendana, since 1996, a general trend is that both Hutu and Tutsi look at members of the ‘other’ ubwoko with suspicion or fear.

As a Burundian, can you say that there is peace in Burundi? There cannot be any peace in a country like Burundi where one ubwoko lives separately from the other. … Go to Mutaho, you will understand. When you see it, if you are not afraid, it is because you have not seen what happened in Burundi. (Hutu man, 57, never displaced, Bugendana, October 2019 Int.)

Seriously, if it were you, what would you do? After all, a tree cannot hit your eyes two times. And you can always protect yourself from an enemy by doing all that you can do. … why even the younger ones, when it rains on the hill, accept to be hit by the rain and come here to the camp[?] They see that it is not safe there. (Tutsi woman, 57, IDP, Bugendana, September 2019 Int.)

By contrast, the ‘ethno-geographical’ boundary between Hutu and Tutsi in Mugara is not as rigid as it was when the first Hutu refugees returned from exile. Over the years, some Tutsi have sold or rented out the land to other people (Van Leeuwen 2010: 756; Rema Ministries 2012: 29), including Hutu. Moreover, increasingly more Hutu have returned since 1993 and especially since the 2000s. In Mugara, Hutu and Tutsi do not live in separate, distant areas between which some spaces of interaction exist, like in Bugendana. Mugara resembles more a village developed around a central area, presenting a centre, different neighbourhoods and outskirts. In this setting, the mixing of Hutu and Tutsi is more accentuated than in Bugendana, a pre-condition for the experience of ‘ethnic’ boundaries to fade and blur.

Moving beyond the ethno-geographic dimension of boundary fading, it is clear that in the absence of open violence, the possibility to return to cultivating the fields is particularly appreciated by those for whom agriculture represents the main if not the only means of livelihood. Under these circumstances, the behaviour of members of the ‘other’ ubwoko is under scrutiny. On the one hand, abstention from violence in itself shows the will to have peace: the absence of threats from people who used to be threatening in the past allows interpersonal trust to gradually increase. On the other hand, visible actions and behaviours coming from the ‘others’ that are judged as positive (greetings; warm welcomes; offers of beer, food, jobs) are interpreted as an attempt at rapprochement, thus reinforcing trust in the ‘others’.

After [the 2005] elections, I saw that some Tutsi too wanted peace. (Hutu woman, 51, former IDP, Mugara, April 2008 Int.)

Among them, there were some who were coming here and we shared everything, and they started to greet us in the same way they were greeting the others. (Tutsi man, 55, IDP, Bugendana, 2015 Int.)

We returned to our houses and the Tutsi called us to go work for them. (Hutu man, 40, never displaced, Bugendana, April 2008 Int.)

Next to these more or less spontaneous occasions of inter-group contact, some interviewees mentioned sensitisations and instructions received from political and religious figures in positions of power that facilitated inter-group rapprochement. This is in line with Allport's (1954) ‘contact hypothesis’, according to which inter-group contact can help reduce prejudice, most effectively ‘in a societal context marked by supportive institutional structures [and] the agreement of relevant authorities’ (Aiken 2013: 36).

There was an administrator … who put us together and told us that there are no Hutu or Tutsi and he asked us to forget what happened and to love each other. … It is him who brought peace back here. He could put together Hutu and Tutsi. … He told us to meet and discuss, forgetting what had happened. (Hutu man, 60, never displaced, Bugendana, 2015 Int.)

They have taught us forgiveness at the radio, and we have forgiven already. Even if we come across those who committed the crimes, we greet each other, then each one continues his path, without any threats. (Tutsi man, 56, IDP, Bugendana, 2008 Int.)

It means that whereas mass (categorical) violence makes collectives experienced as bounded ‘groups’ out of mere categories, like the amoko, it is positive contact that unmakes experiences of bounded ‘groups’, making real people re-emerge out of abstract categories. To be of greatest effect, this should be facilitated by ‘a broader social and normative climate conducive to improved intergroup relations’; ‘positive intergroup contact must be of a non-adversarial quality, must take place between groups afforded equal status in society, must ideally be conducted over an extended period of time, and must be undertaken in the pursuit of cooperative or superordinate goals which actively aim to transform group divides’ (Aiken 2013: 36).

Other categorisations: collective or individual

The second important way through which the salience of boundaries declines is a process of blurring. Boundaries can be of different types: territorial, political, economic, social, cultural, religious, linguistic. The type of boundary reflects the lines along which the socio-political differentiation between what is experienced as ‘groups’ is made. Because different types of boundaries with different positions and characteristics coexist at the same time, individuals have ‘multiple belongings’, which ‘have not the same importance, in any case not at the same moment’ (Maalouf 1998: 19). Thus, in Burundi, depending on different factors and circumstances, boundaries are relevant in different ways for different members of the same ubwoko. The boundary between amoko is sometimes blurred with either a territorial or an economic dimension. Some interviewees explained how one's territory of origin sometimes prevails over the ubwoko in the vision and division of the social world: ‘When I see a group with a Tutsi who comes from here, even if he lives in Bujumbura, I feel confident because I see a Tutsi from my place of origin. The same applies to a Tutsi who sees a group with Hutu who come from here, he feels confident and goes to ask them what they are doing there’ (Hutu man, 56, never displaced, Bugendana, 2015 Int.).

In line with the relevance of the territory of origin in the identification of the ‘others’, the experience of cross-border migration (which happened abundantly in Mugara) contributes to blurring the boundary between amoko. Most times, this is related to the presence of land conflicts after returning to Burundi.

What I noticed is that if you have a plot of land here in Mugara, your first potential enemies are the locals, the natives from here, and especially the former refugees. Why? They think that you are richer than them. … Life is still not good for them, so if they hear that you have one hectare, or half, where you can harvest palm nuts, [for them] you become one of those people who received land from the state, those who came from the countryside. The same people with whom you were together can be even more harmful than the others. Why? They see that you become rich, and they start asking questions to themselves about you. … As far as I can see, there is no Tutsi who constitutes a problem for me, and no Hutu [either]. (Hutu man, 26, son of returnee, Mugara, November 2019 Int.)

In some cases, the political reasons that pushed some to flee the country and experience cross-border migration, while other people from the same ubwoko remained in Burundi, represent the element that blurred the boundary between amoko. In this way, ‘return migration to Burundi … created a new set of group categories based on where individuals were during the war’ (Schwartz 2019: 128). For many people in Mugara, after their return to the country, an ‘ethnic’ reading of the situation is not applicable because the parties in land conflict often belong to the same ubwoko. This blurs the boundary between amoko with a boundary based on the experience of cross-border migration: the distinction between returnees and those who never fled Burundi becomes more relevant than the distinction between Hutu and Tutsi, as land conflicts are between returnees and non-returnees (Schwartz 2019: 135), and among the latter are Tutsi as well as Hutu, sometimes even family members (Kamungi et al. 2005: 217; Van Leeuwen 2010: 756). This is one of the ways in which boundaries between amoko are blurred and become thinner.

If we consider [the Hutu] who remained here in the country, I do not know if we are going to consider them in the same way as those who were in Tanzania. Those who remained here were like the Tutsi, because they did not speak the same language as us, who were in exile.

(Interviewer) Why did you flee, when others remained here?

If you understand well, we were not on the same page with those who remained here. (Hutu man, 43, returnee, Bugendana, 2015 Int.)

Boundaries can exist between different collectives experienced as ‘groups’ (inter-group) or inside the ‘group’ itself (intra-group). For this reason, boundaries never separate one ‘us’ from one ‘them’: there are always several ‘us’ and several ‘them’. Economic status also permitted to revise the relevance of belonging to a specific ubwoko in the identification of the ‘others’, opting out of an ‘ethnic’ reading of the main cleavage structuring the social world.

I would not say that there are differences [between Hutu and Tutsi]. Everything is mostly the result of poverty. We have seen this. Even if the fact of being Tutsi derives often from wealth, if a person does not represent any danger, an enemy is someone who is a source of trouble for you. Even the latter can help you with something. When you have problems, you can be helped by him or by someone of your same [ubwoko] …. You see that the ubwoko is not a problem [in itself]. (Hutu man, 57, never displaced, Bugendana, October 2019 Int.)

The previous quotation also underlines the importance of individual characteristics and actions, not of the ubwoko as a bounded category, in the identification of the ‘others’. This represents another way in which boundaries between amoko are blurred. By refusing to make generalisations and descending to the individual level, the interviewees reduced the relevance of the ubwoko of any other type of boundary-making between groups ‘as a principle of categorization and social organization’ (Wimmer 2013: 61), relying on the judgement of single actions to position themselves and the others. Those actions could be positive, such as supportive behaviours, or negative, such as conflictual relationships, often related to land disputes, a thorny issue in post-war Burundi (Wittig 2017).

There are no groups, I trust people individually because there are honest people and dishonest people. (Hutu man, 48, former IDP, Mugara, 2015 Int.)

Myself, I am a Hutu, and to trust another Hutu, one needs to see what good things he does; as for a Tutsi too, I can trust him considering what he is doing for me or for the country. (Hutu man, 69, never displaced, Bugendana, 2015 Int.)

By putting emphasis on the individual instead of what is experienced as a demarcated ‘group’, this ‘emotionally laden sense of belonging to a distinctive, bounded group’ (Brubaker & Cooper 2000: 19) decreases.

Contesting the boundary as such: interstitiality

The turn to the individual away from dominant types of identifications and categorisations in society is also at work through people situated in-between ‘us’ and ‘them’, who contest, by their very existence, boundaries between groups. This turn shifts the focus from collectives experienced as bounded ‘groups’ to individuals, from the main social ‘groups’ of reference to those ‘lost to the group’ (Wimmer 2013: 42). By doing so, it complements analyses of boundary-making that focus on the varying degrees of thickness of the boundaries analysed above. It brings in what Bhabha (1994: 4) has called an ‘interstitial passage between fixed identifications’. These are individuals who – in the reciprocal perceptions and interactions between ‘us’ and ‘them’ – are not only perceived as ‘other’ by (most) members of the amoko between which they are situated, but they are also aware of this, and this is what makes their position interstitial, in the social landscape of boundaries. During fieldwork, we encountered several of these situations and persons. For lack of space, we only discuss one in some detail.

In Bugendana, a Hutu man fled to live in the predominantly Tutsi IDP camp in 2010, after living in the DR Congo and in other places in Burundi between 1993 and 2009. In 1993, this man suffered from violence perpetrated by both Hutu and the military, the latter perceived as close to the Tutsi. Hutu allegedly persecuted him because he had married a Tutsi woman, while the military represented a threat for him because it targeted the Hutu during operations aimed at restoring order (after Hutu attacked their Tutsi neighbours following the assassination of President Ndadaye).

I am married to a Tutsi woman, and in 1993 I was beaten by Hutu and threatened because I did not want to participate in the killings. They even broke my wrist [he shows his wrist]. For this reason, I could not ask any of the Hutu on the hill [to give me a plot of land where I could live]. This is also why I went to my brother-in-law. And it is thanks to the Tutsi that I have sheet metal for my house in the camp. In 1993, they stole three goats and burned down my house. … Before 1993 and during the crisis, I used to work at …. Then, the military wanted to kill me and I gave up. I gave back the keys … and fled. I never went back there. (Hutu man, 56, IDP, Bugendana, November 2018 Int.)

An important aspect of his story and identity is related to the fact that he had married a member of the ‘other’ ubwoko. His interstitiality is thus also related to, though not necessarily fully based on, what is commonly referred to as a ‘mixed marriage’ in the social landscape of Burundi. People married to a member of a different ubwoko, as well as people whose parents belong to different amoko (‘mixed’ descent), may also be considered ‘in-between’. However, this depends on their self-identification, the identification by others, and the relationship between these two sources of identification.¹² This can happen along or beyond the prevailing societal norms prescribing how to classify these situations. According to these norms, people who decide to marry a member of another ubwoko still belong to their own ubwoko, and children of ‘mixed marriages’, socially speaking, inherit the ubwoko of the father (inheritance of the ubwoko is patrilineal in Burundian society). Nevertheless, their individual identity can also be seen as situated in the overlap between their own and their partner's ubwoko, in the case of ‘mixed’ couples, or between the parents’ different amoko, in the case of ‘mixed’ descent. Despite societal norms imposing one categorisation, the position in the social landscape actually depends on the (self-)identifications, which can vary over time and according to the circumstances.

The interstitiality of this man might have originated in his marriage with someone from the other ubwoko but was shaped by actions in the context of violence. In 1993, when faced with the obligation to take sides, he refused to make a choice, to mark ‘“them” off from “us”’ (Wimmer 2013: 71) through violence in a situation of categorical uncertainty (Appadurai 1998). If the overlap of categories was not problematic until a period of tensions, it became dangerous when violence broke out. Eventually, this person had to make a choice; he then turned to those who allowed him to continue to live, also by providing him with sheet metal for his house, as he mentioned during the interview.

However, this Hutu IDP did not become Tutsi. He might appear to have crossed the boundary between amoko, but is not perceived by the new group as ‘one of them’. This is what makes his situation interstitial: he seems to have left the categorisation as Hutu but is not fully integrated into the Tutsi IDP group. The very stereotypical description of him made by a Tutsi IDP reveals the perception of him as belonging to the other ubwoko.

There is one … also known by the name Murundi [‘Burundian’]. There is another one who lives next to the road. They are considered as IDPs when aid arrives.

(Interviewer) Why is [he] called Murundi?

He is a pure Hutu. He is a brick maker, he likes working. He himself called himself that way. When he has been drinking and he comes back home he says “here is the Hutu who comes back home!”. He also calls himself umukozi [‘worker’]. (Tutsi man, IDP, Bugendana, October 2019 Int.)

This quote is testimony to the fact that this Hutu IDP does not identify himself with the Tutsi ubwoko when behaving within the context where these identifications are evident or dominant. But neither does he seem to identify with the Hutu ubwoko, as exemplified by felt distance expressed through feelings of fear and distrust towards the people and places where these identifications dominate: ‘I cannot go back to [the hill] because I do not have any land there. But I cannot go there, not even for a short time, I have to manage my time there. If I go there, I need to come back before the dark. You never know what might happen during the night’ (Hutu man, 56, IDP, Bugendana, November 2018 Int.).

From the boundary-making perspective, analysis of interstitial identities provides an important contribution to the study of boundaries because it emphasises that individuals are not always situated on one side of a boundary, and that interstitiality contests boundaries without any identification and categorisation processes decreasing the boundaries’ thickness. Interstitial identities show that individuals can remain on the very boundary, stuck in an in-between position. From these interstices, they ‘provide the terrain for elaborating strategies of selfhood … that initiate new signs of identity’, and it is in these ‘interstices – the overlap and displacement of domains of difference – that the intersubjective and collective experiences of nationness, community interest, or cultural value are negotiated’ (Bhabha 1994: 1–2). One could easily add ‘ethnicity’ or ubwoko here. It is here, in this interstice, that the very meaning and relevance of the dominant boundary is contested.

DISCUSSION

In this article, we analysed how boundaries between amoko wax and wane in contemporary Burundi. Through the analysis of empirical data and the adoption of a constructivist perspective, we could confirm the fairly established insight that interactions, either violent and thus leading to distancing, or non-violent and experienced as positive, thus facilitating rapprochement, are drivers in the emergence or decline of the experience of rigid boundaries, or in their waxing and waning. Less established is the importance of the notions of ‘multiple temporalities’ and ‘multiple identities’, which we discuss in this session.

The analysis of the empirical material clearly shows that in the aftermath of violence, time plays a role in the decrease of rigidly experienced boundaries. Generally speaking, the fading and blurring happens gradually over time, and more so in periods more distant from intense violence. However, just as time does not heal all the wounds (Ingelaere & Verpoorten 2020), our analysis of the making and unmaking of boundaries underscores that it is not simply a question of a linear progression in time but of ‘multiple temporalities’. ‘“Temporality” refers to the orientation of experience to a temporal frame, through remembering the past, experiencing the present, or imagining the future’ (Otake & Tamming 2021: 547). Boundaries created by intense processes such as violence tend to persist over longer periods or to periodically or contextually re-emerge. The potential for tensions, conflict and ‘crisis’ is always present (Turner 2010: 125; Berckmoes 2017); even more so when the memory of the past continues to inform in an important way people's daily perceptions and interactions. More specifically, multiple temporalities in people's experience might provide a fertile ground for political discourse emphasising differences between amoko (oftentimes referred to as ‘ethnist’ discourse), which could resonate well with the experience and perceptions of people for whom the boundary between amoko, even several years after the last episodes of open violence, remains thick. When this happens, the boundary is likely to be strengthened. Therefore, boundaries do not simply fade with the passing of time.

This means that there is hardly ever a simple ‘before’ and an ‘after’, as a linear conception of time suggests, to the experience of acts of violence due to the social processes it generates. A linear sequencing of events does not do justice to the everyday experiences of boundary-making and -unmaking. The memory of past violence, in personal or collective recollection, continues to inform in an important way people's daily perceptions, interactions and expectations about the future regarding the type and nature of the boundaries experienced. This reinforces the ‘intermittent’ nature of the ubwoko (Brubaker et al. 2006: 208) and points to the need to question the linear timeframe leading from violence to peace and embrace the notion of multiple temporalities. Policy interventions such as peacebuilding or ‘transitional justice [appear] mechanistically conceived, suggesting a past of violence and a present for justice and closure’ (Igreja 2012: 408). On the contrary, dealing with past violence implies dealing with its presence in the present, such as the experience of rigid boundaries. Transitional justice initiatives in Burundi do not easily adopt such an approach. The working of the Truth and Reconciliation Commission, for instance, is mainly geared towards establishing the forensic truth of acts of violence that happened in a distant past, contrary to what the population often expects from such a process (Ingelaere 2009; Taylor 2013). More fruitful would be to explore instead, or also, the narrative and social truths about the nature and relevance of boundaries between amoko in the everyday. Neither is there a lot of attention, and thus money, going into initiatives that can more easily address these temporalities in formal or informal healing processes or cultural practices (Ingelaere & Kohlhagen 2012; Lambourne & Niyonzima 2016; Van Leeuwen et al. 2020).

In parallel to the persistence or re-emergence of rigid boundaries along amoko through ‘multiple temporalities’, the decrease of its rigidity or even dissolution resides in the multiplicity of belongings. Individuals have ‘multiple belongings’, which ‘have not the same importance, in any case not at the same moment’, and which represent the ‘constituents of personality’ (Maalouf 1998: 19). In fact, the ‘very capacities in which one is defined on various levels, or within various circles such as family structure, local life, the workplace, and the nation, make one necessarily multiple and not fully congruent with only one identity definition’ (Král 2009: 26). Other identifications of self and other besides those related to amoko, individual assessments of people instead of collective ones, and the existence of people in interstitial positions show that the social landscape is actually not divided into Manichean ‘us’ and ‘them’, into Tutsi and Hutu. These elements show that alternative categorisations of and narratives around ‘others’ are possible, and that ‘the others’ are not necessarily the same as those who represent ‘the others’ for other groups of people.

In a society where the boundary between Hutu and Tutsi is overall still very relevant, the existence of people in interstitial positions, for instance, questions the relevance of that boundary. Other categorisations, based on economic status, place of origin or migration history, can also contribute to raising awareness that living ‘beyond’ amoko is actually possible, although relatively difficult. This is not to suggest that people abandon their identity as members of specific amoko, if that is ever possible, but that they be open to different narratives around ‘the others’ that may decrease the thickness of the boundary between Hutu and Tutsi in contemporary Burundi. However, just as with transitional justice efforts, peacebuilding initiatives do not fully tap into the potential of these processes. This is for instance the case with the consociational power-sharing arrangement considered as one of the hallmarks of dealing with past violence in Burundi. Although this approach is generally considered instrumental in reducing the role of ethnicity in politics (Lemarchand 2007; Curtis 2013; Vandeginste 2015; Reyntjens 2016; Raffoul 2020), in light of this analysis, it might also paradoxically entrench the salience of ‘ethnic’ identity in wider society in an essentialist way, thus failing to avoid ‘groupism’, the ‘tendency to treat … groups as substantial entities’ (Brubaker 2004: 64), ‘chief protagonists of social conflicts’ (Brubaker 2004: 8), which haunts policymaking on and the scholarly study of violence and peace in Burundi.

CONCLUSION

The analysis presented in this article aims to address, through an empirical contribution, important blind spots in the knowledge on the reproduction and transformation of the amoko as bounded social forms in Burundi following mass categorical violence. Our analysis of processes of inter-group distancing and rapprochement does not consider Burundi's amoko as monolithic, homogeneous entities. Instead, we aimed to identify the main characteristics of the processes of waxing and waning of boundaries between amoko across time, space and ‘ethnic’ affiliation. Our analysis underlines the importance of a constructivist understanding of ‘ethnicity’ in the scholarly study as well as policymaking in Burundi, which often tacitly work with an essentialist understanding. Contrary to such a point of view, the bottom-up perspective developed in this article clearly shows how the meaning and nature of the amoko is not immutable or ‘frozen’ in time. Instead, the experience of boundaries has a ‘variable and contingent, waxing and waning nature’ (Brubaker 2004: 19), the intensity and meaning of which are always situated and determined by a series of factors. Violence, both physical and in discourse, leads to the emergence of rigid imaginary boundaries between the Hutu and Tutsi amoko. The rigidity of these boundaries decreases through non-violent interactions, the emergence of non-ethnic categorisations, and the awareness of interstitiality, which correspond to processes of boundary blurring, fading and contestation. Through the analysis of empirical material, this study allows us to better understand how the waxing and waning of ‘ethnic’ boundaries take place in contemporary Burundi. Ultimately, it underlines the importance of embracing notions such as ‘multiple temporalities’ and ‘multiple identities’ in peacebuilding and transitional justice processes, to avoid essentialising groups and paradoxically entrenching the salience of ‘ethnic’ identity in the wider society.