Remembering and silencing complexity in post-genocide memorialisation: Cambodia’s Tuol Sleng Genocide Museum

In genocide, complex political actors can take on changing roles of perpetrator, victim or hero at different points in time. In post-genocide societies, political actors seek to shape memory of the violent past to forward their own interests, often undermining this complexity and painting a more black-and-white picture that ties in with Transitional Justice practitioners’ dichotomous assumptions about perpetrators and victims. This article looks at how complexity is remembered and silenced in a post-genocide memorial space that included many complex political actors during its tenure as a security centre: Tuol Sleng Genocide Museum in Cambodia. Here, the audio guide and permanent and temporary exhibitions (as well as changes to these) allow for a co-existence of competing memories, demonising the Khmer Rouge regime for its immense cruelty and simultaneously constructing victimhood for former Khmer Rouge cadres. This could serve as a starting point for discussing complexity, but instead silences in the exhibitions and audio guide create an ambivalence in attributing these roles that masks this complexity.


Introduction
Individuals can take on a wide range of roles during violence, including perpetrating, bystanding, rescuing or even becoming victims (Fujii, 2009;Luft, 2015;Williams, 2018a). To account for this diversity and complexity, concepts of complex political actors highlight the synchronicity of these actions for some people (Baines, 2009;Bouris, 2007;Jessee, 2017). Furthermore, after the violence ends societies vary in how they remember this complexity of actors who both engage in victimising actions, but also attempt to rescue others from it or become victimised themselves.
In post-genocide societies political actors will sometimes choose to render memory of the past black and white, emphasising the guilt of some, the victimhood of others and establishing the heroism of a select few. This is particularly likely when it serves to legitimise these political actors in the post-genocide order or delegitimise adversaries. Moreover, within the field of human rights and transitional justice much of both the practical and theoretical work on post-conflict settings exacerbates these claims, forwarding dichotomous readings of the past that allow only for clear constructions of perpetrators and victims as separate and distinct. Memory studies' more nuanced perspective on the construction of the past helps to recognise and theorise this complexity of responsibility and disentangle the conflicts surrounding it better than more dichotomous approaches.
What is significant about an acknowledgment of complexity and a nuanced reading of this? Embracing complexity allows actors in post-genocide societies to engage in more nuanced discussions on agency, responsibility, culpability, heroism and victimhood, highlighting the situational nature of violence and the different roles people can assume at different times. This assertion of complexity allows for individuals in post-genocide contexts to develop a more active awareness of roles within the system of violence and the possibilities that existed for agency, possibly undermining some of the more simply structured black-and-white narratives. Not only does this allow a more accurate memory of the past but it also has the potential to re-structure relationships in the post-genocide society. Without acknowledging this complexity, dichotomous allocations of the responsibility for past violence risk creating resentment or alienating groups; at the same time, a nuanced reading of complexity that highlights agency and individual responsibility can give useful perspectives in terms of prevention.
In Cambodia, the totalitarian nature of the Khmer Rouge regime created complex political actors, too, as many Khmer Rouge cadres themselves were also victimised, falling prey to the internal purges of the regime and rendering themselves more likely to be targeted for arrest and execution. Since the fall of Democratic Kampuchea, as the Khmer Rouge regime was called, the role of former cadres has been represented in distinct ways in the Cambodian memoryscape (Williams, 2019). Two 'mnemonic role attributions' are particularly influential: First, the 'generalised demonisation mnemonic' entails a clear attribution of guilt that demonises the Khmer Rouge as the group which wreaked this immense cruelty; second, the 'universal victimhood mnemonic' allows anyone except the absolute highest leaders to claim some form of victimhood regarding the totalitarian rule of the Khmer Rouge (Williams, 2018b(Williams, , 2019. While complex political actors have been discussed in transitional justice processes (Bernath, 2016;Manning, 2015;Williams, 2019), less attention is paid to this complexity's manifestation in memorials. Memorials are an important forum for negotiating interpretations about the past and what they mean today as they are 'physical representations of imagined communities' (Naidu, 2014: 38) that fix and store memories about the past (Buckley-Zistel and Schäfer, 2014: 4). Memorials are thus 'political tools, often created and utilized with specific political agendas [. . . and] it appears that the concerns of the present loom much larger in these museums than the difficult memories of the painful past' (Sodaro, 2018: 5). This political role rests on a memorial's ability to frame the past in specific ways that shape the political perceptions and attitudes of visitors (Hamber et al., 2010;Naidu, 2014;Sodaro, 2018).
In post-genocide Cambodia, Tuol Sleng Genocide Museum (TSGM) has become the central memorial space for remembering the Khmer Rouge past, along with the killing fields at Choeung Ek, and is thus dominant in interpretations about the meaning of the past violence in the country, particularly for its many international visitors (Sion, 2014). And yet, most people who were imprisoned and killed at S-21, the security centre at TSGM's location, were former Khmer Rouge cadres, making them complex political actors. As such, this article poses the question: How does TSGM deal with the complex nature of the space's victims?
This article's main argument is that at TSGM there is an ambivalence around the meaning of victimhood and perpetration in the exhibitions and audio guide drawing on both the generalised demonisation mnemonic and the universal victimhood mnemonic. These two somewhat contradictory readings of former Khmer Rouge cadres' roles co-exist, but without being discussed together in a complex reading of these roles. The two mnemonic role attributions are made compatible by silencing any mention that most people tortured here and then killed were purged cadres of the Khmer Rouge. As such, this silencing allows the two mnemonic role attributions to co-exist in an ambivalent fashion but reduces the portrayal of complexity. While a less complex story is told by omitting a detailed discussion of the possible complicity of the victims of this space, this curation allows for the site to be embedded usefully in broader memory politics and is in line with other under-complex forms of dealing with the past in Cambodia (Bernath, 2016).
This speaks to broader debates in the memory literature about complex political actors and how they can be remembered in post-genocide societies. The insights generated by this close reading of one memorial site in Cambodia demonstrate how competing types of mnemonic role attributions are rendered compatible through silences. These silences, however, avoid discussions of how individuals can engage in different types of roles during violence and by ignoring complexity undermine perspectives for prevention.
In this article I draw on insights gathered during multiple visits to TSGM since 2014 that allow an in-depth perspective on the memorial space and its exhibitions. Besides an analysis of the main exhibition, I also analyse the audio guide that a commercial vendor, Narrowcasters, provides at the site, temporary exhibitions curated by staff members, temporary exhibitions curated by other organisations but displayed at TSGM, as well as semi-structured interviews with museum staff, exhibition curators, audio guide creators and others. Different actors are important for overall effect of the site today, although visitors usually cannot differentiate between who is responsible for which contribution; for example, visitors are less interested in who created the audio guide or what exhibitions are curated by outside actors or museum staff. This data is complemented by participant observation between January and March 2018. 1 Over the years, I have visited many of the temporary exhibitions, others I know only from the exhibition catalogues. I analysed the collated data using qualitative content analysis to understand the various themes discussed at the museum as well as staff perceptions of these. As such, my analysis marks an appraisal of the site at a specific moment in time with no allusion that this cannot or will not change; quite to the contrary, much has started to change and adapt in the past couple of years at TSGM, and it will be interesting to see, whether and when the silence I have identified here is discussed and complicated within the permanent exhibition. 2 This article begins with a conceptual discussion of the politics of post-genocide memory, discussing ideas around complexity, silences and mnemonic role attributions. This is followed by a brief introduction to the Cambodian case, before introducing TSGM and analysing the two main mnemonic role attributions at the site. The article then discusses how silences in the two narratives allow them to co-exist and how this interacts with conceptions of complexity and victimhood, before concluding with a brief discussion.

Complexity and silences in the politics of post-genocide memory
In countries that have experienced war, atrocity or genocide, it is common for various different and competing memories about the past to co-exist, propagated by different actors or at different points of time. Politics, narrative and memory mutually constitute one another. Memory is vocalised through collective narratives about the past that in turn shape interpretations of the present political landscape (Wertsch and Billingsley, 2011). Some memories are reproduced while others are lost over time (Roudometof, 2002: 7); some memories become more important, while others become simplified (Williams, 2007: 166). As such, these memories are political and there is competition around which of these different versions of the past asserts itself, as this will have impact on political and societal relations today (Barahona de Brito et al., 2001;McDowell and Braniff, 2014;Wertsch and Billingsley, 2011). Importantly, which memories are forgotten or silenced is thus a deeply political issue. Collective memory in post-genocide settings can be particularly fraught with controversy and narratives structuring memory of this past are especially contested (Björkdahl et al., 2017;Buckley-Zistel and Schäfer, 2014: 4) due to the legitimacy that they can provide for political power in the present (Druliolle and Brett, 2018;Sodaro, 2018).
One way to analyse the politics of memory is through the concept of 'mnemonic role attributions' defined as 'the sum of how actors, their roles, their responsibility and their suffering are categorised as they are remembered regarding a certain period of time' (Williams, 2019: 163). Memories compete with each other. Who can claim to be a victim of the past violence? Who is deemed to be a perpetrator and thus responsible for this violence? Should bystanders' actions be understood as culpable or passive? The concept of mnemonic role attributions is particularly useful as it draws attention explicitly to the legitimacy claims that are implicit in the attribution of roles and the moral connotations they have, with attributions of victimhood affording moral legitimacy (Bonacker, 2013;Druliolle and Brett, 2018;McEvoy and McConnachie, 2013) and attributions of perpetration leading to a demonisation and de-legitimisation of an individual or group.
In remembering violence, individuals' pasts can be attributed to multiple roles, rendering our understandings of this violence complex. Complex political victims are victims who are no longer innocent and pure, but can become complicit in the system, playing their part in perpetuating and supporting the discourses and practices that ultimately also victimise them (Bouris, 2007: 84). Complex political perpetrators are perpetrators who are responsible for their actions but see their culpability mitigated by also being victims, for example as in the case of child soldiers (Baines, 2009). This kind of complexity abounded during the Khmer Rouge regime as cadres were naturally perpetrators, but at the same time were often victimised in internal purges, being arrested and killed as suspected enemies.
This complexity does not always translate into recognition in the memory of past violence, as transitional justice processes or post-conflict government policies often strive for politically useful, clear-cut and dichotomous attributions that mask the complexity. Approaching the topic from the perspective of memory studies allows us to gauge the degree of complexity that is attributed to individuals' roles of perpetrator, victim, bystander or rescuer.
This ties into recent work that highlights the importance of not only what is remembered, but also what is forgotten or silenced (Buckley-Zistel, 2009;Connerton, 2008;Eastmond and Mannergren Selimovic, 2012;Mannergren Selimovic, 2020;Mason and Sayner, 2019). Silences go beyond a mere absence of speech and can communicate as much about memory of the past as any spoken narrative, allowing positive and negative attributions, as well as ambiguity to be communicated through the silence (Eastmond and Mannergren Selimovic, 2012: 506). In the curation of memorials silences are not to be understood passively but as an 'intentional encoding of place' (Meierhenrich, 2011: 286). As such, silences are a 'socially constructed space entailing both remembering and forgetting' (Winter, 2018: 27) and can be used as a political strategy that is served by selective memory that legitimates certain perspectives (Eastmond and Mannergren Selimovic, 2012: 507).
Bringing this together, it is clear that the way the past is remembered is significant for political power in the present and that legitimacy is won or contested through attributions of roles regarding the past violence. What roles are attributed and how complex memories are allowed to be, is part of a political process that highlights or silences certain aspects of past roles to serve political interests today. It becomes more difficult when -as in the case of this article -these memories may seek to pick out certain parts of the complexity that appear to be contradictory or at least ambivalent in their construction of the past and yet both versions are politically useful in some way. Sometimes, political actors can have significant interest in rendering ambivalent narratives compatible in order to avoid cognitive dissonance or a de-legitimisation of one or even both of the narratives. One strategy to render ambivalent mnemonic role formations compatible is to silence certain elements and focus more strongly on other parts of the narrative that are not contradictory. As such, silence surrounding complexity can serve a political function within the broader memoryscape. These silences within mnemonic role attributions allow ambivalence to be maintained, while foregrounding more complex readings of these memories.

The politics of memory in post-genocide Cambodia
To explore the empirical dimensions of this argument, we now turn to Cambodia. I begin with a brief overview of the violence under the Khmer Rouge and the politics of memory since so that TSGM's place within the broader context becomes clearer (for excellent introductions see Chandler, 2008a;Kiernan, 1996). Democratic Kampuchea existed from 17 April 1975 until 7 January 1979, although it was embedded in a longer continuum of violence that included a previous civil war in 1970 and then continued with a second civil war that lasted until the late 1990s. Under the totalitarian reign of the Khmer Rouge, between 1.7 and 2.2 million people died, about half from hunger, overwork and sickness, the other half by execution (Tabeau and Kheam, 2009: 19). Living conditions were horrific, all members of the population, including Khmer Rouge cadres, experienced hunger and tiredness and feared for their lives and all lost family members to the violence. While the killing disproportionately targeted ethnic minorities, in particular ethnic Vietnamese and Cham (Tabeau and Kheam, 2009: 19), any divergence from the expectations of the revolution could also be seen as treachery and the person declared an 'internal enemy' or 'enemy of the revolution', which lead to 're-education' or execution (Chandler, 2000: 45-76).
There was a wide system of security centres that were organised hierarchically and interrogated people were passed up the chain if deemed dangerous (Ea, 2005). At the top of this system was S-21 in Phnom Penh (for an overview see Chandler, 2000), the highest security centre. S-21 'processed' those traitors most dangerous to the revolution. From the perspective of the Khmer Rouge these most dangerous people were often Khmer Rouge cadres themselves who were suspected of being CIA, KGB or Vietnamese agents, meaning that at S-21 the majority of interned individuals were former Khmer Rouge who had fallen victim to one of the many internal purges. S-21's central position in the system is reflected as the location for TSGM.
On 7 January 1979 Cambodian Khmer Rouge defectors, supported by the Vietnamese military, liberated Phnom Penh and toppled the regime. The Khmer Rouge retreated to the Thai border, and the next civil war ensued, continuing into the 1990s (Bultmann, 2015;Chandler, 2008a: 277-295). In this context the government saw its authority disputed within the country and remained isolated internationally given complex Cold War politics surrounding Vietnam. As such, the government sought to create legitimacy for the Vietnamese intervention and liberation, highlighting the terrible past from which the country was saved, as well as emphasising the imminent threat of violence should the Khmer Rouge regain power (Brown and Millington, 2015: 32;Hinton, 2018: 47;Hughes, 2006: 272;Tyner et al., 2014: 286). In this vein, it drew heavily on the generalised demonisation mnemonic to construct and vilify this enemy. Demonisation in the 1980s focused on the so-called 'Pol Pot-Ieng Sary' clique in an attempt to demonstrate these as traitors to the ideals of the revolution, given that the new regime was also communist, but tarred the Khmer Rouge as a whole in order to undermine their contemporary legitimacy. The generalised contempt for the Khmer Rouge remains an important background for the legitimacy of the government today.
Government memory politics shifted in the early 1990s with the Paris Peace Agreement, the UN peacekeeping mission UNTAC and Prime Minister Hun Sen's 'Win Win Policy' that promised defecting units of the Khmer Rouge amnesties and economic benefits in exchange for laying down arms and reintegrating. Successive rounds of defections ultimately decimated the Khmer Rouge, the civil war ended and peace returned to Cambodia. This return to peace through his policy is key to Hun Sen's political legitimation strategy today and the victimhood of former Khmer Rouge is accepted by many victims in Cambodian society today (Williams et al., 2018: 56).
With the end of the civil war, the necessity for mobilising hatred evaporated, and it became more important to successfully integrate former Khmer Rouge. Hun Sen called on the population to 'dig a hole and bury the past' (quoted in Chandler, 2008b: 356). This 'induced amnesia' (Chandler, 2008b) uneasily coexists with subsequent efforts by the government to pursue transitional justice, although the government is careful to focus these on only the highest leadership. This new perspective of government memory politics can be understood as the universal victimhood mnemonic that allows most Khmer Rouge beyond the very small circle of leaders to deny responsibility for any wrongdoing during Democratic Kampuchea and even enables them to emphasise elements of victimhood they have experienced.
Thus, at different points in time former Khmer Rouge have been represented variously with an emphasis on varying facets of their more complex roles as perpetrators and victims. In this context, it is clear that it matters how responsibility for past violence is allocated and how roles are represented as it has manifest consequences for political legitimacy and authority in the present. Next, we turn to the specific case study of TSGM for a deeper understanding of the memorial site and the way complexity is negotiated and silenced within this space.

Remembering and attributing roles at Tuol Sleng Genocide Museum
TSGM started receiving visitors in 1979 on the site of S-21 as evidence of the horrors of Democratic Kampuchea (Brown and Millington, 2015;Tyner et al., 2014: 285). In line with the government's broader politics described above, this can be read as a bid to legitimise the toppling of the previous regime not as an invasion but as a liberation from a totalitarian and genocidal regime (Brown and Millington, 2015: 32;Hinton, 2018: 47;Hughes, 2006: 272;Williams, 2004). The memorial museum's exhibition was finished in time for visits by a legal commission and foreign guests at the People's Revolutionary Tribunal, a trial in absentia of Pol Pot and Ieng Sary (Tuol Sleng Genocide Museum, 2019: 112), rendering the space a crime site and evidence in the trial. 3 While the literature often implies that there is 'evidence that the museum was designed primarily for foreign consumption' (Ledgerwood, 1997: 89), there was clear interest amongst the local population to visit: an estimate by TSGM puts the number of visitors to TSGM at around 324,921 Cambodian visitors in the first 2 years, as well as 21,444 foreigners (Tuol Sleng Genocide Museum, 2019: 67). As such, the museum's curation was clearly targeted at both domestic and international audiences. Today, international tourists dominate among the visitors to TSGM, 4 as the memorial museum is seen as a 'must-see' destination (Buckley-Zistel and Williams, 2020;Isaac and Çakmak, 2016), but in recent years TSGM has prioritised making the site more accessible for Cambodians as a memorial and education space. 5 The exhibition draws strongly on shocking aesthetics that do less to inform and more to provoke emotion (Hughes, 2008;Isaac and Çakmak, 2016;Violi, 2012), providing an 'impressionistic, immediate access to the exhibits, based mainly on sensations and impressions rather than on cognitive content' (Violi, 2012: 48). While the recently introduced audio guide provides considerably more information than was previously available, it is also key to creating an emotional response through graphic descriptions and the use of emotive language. The audio guide is also key to visitor experiences as most visitors listen to at least parts of the narrative during their visit.
Despite the changing political landscape described above, much of the permanent exhibition is as it was when TSGM opened. Even though some parts have been re-arranged, some elements removed and some things added (see photo overview in Tuol Sleng Genocide Museum, 2019: 84-89), the character of the exhibition has been maintained (for analyses of the space, see Brown and Millington, 2015;Hinton, 2016;Hughes, 2003Hughes, , 2008. The main exhibition has also been complemented by temporary exhibitions, some of which are displayed over several years (curated by outside organisations such as DC-Cam), and more recently also with in-house curated exhibitions on an annual rotation. The biggest change to the visitor experience was arguably the introduction of an audio guide in 2014 that gives visitors considerably more information, context and details about the space. The space has been curated in this way as it is politically useful to the government to frame memory of the violent past in this way. Like all public institutions in the country, the memorial museum can only act within the confines of government approval. Two mnemonic role attributions are prominent at TSGM that derive from broader memory politics: the generalised demonisation mnemonic and the universal victimhood mnemonic (Williams, 2019).

The generalised demonisation mnemonic
The generalised demonisation mnemonic 6 stylises the Khmer Rouge as a whole into an 'evil other' and was prevalent in the 1980s as the new government sought to foment hatred for the Khmer Rouge in the context of the civil war (Chandler, 2008b). The attribution of blame and guilt is generalised in such a way that the Khmer Rouge is seen as an undifferentiated whole without any reflection of individual actions or culpability. As such, it is not interested in the culpability of individual cadres but solely in the guilt of the entire group.
First and foremost, this mnemonic becomes visible at TSGM by drawing on the idea of evidence, both of the actual site and of the exhibited items, pictures and spaces (see also Elander, 2014). As such, there are also many references to 'evidence' throughout the audio guide 7 and this is also a key theme for staff members at TSGM. 8 Victims of the Khmer Rouge amongst the broader population perceive the memorial museum as an important piece of evidence for the existence of the Khmer Rouge regime 9 and the space was framed as evidence for the People's Revolutionary Tribunal in 1979(Tuol Sleng Genocide Museum, 2019, as well as in recent years for the hybrid tribunal, the Extraordinary Chambers in the Courts of Cambodia (ECCC). This evidentiary basis at TSGM demonstrates strongly the evil of the past regime and is thus a key component to the above national and international legitimation strategies that the government has pursued.
The generalised demonisation of the Khmer Rouge is created at TSGM by discursively constructing the Khmer Rouge as a generalised actor who is responsible for the horrors of the space as well as for the broader regime. Similarly to Choeung Ek Killing Fields (Bickford, 2009), there are few references to individual perpetrators in the exhibition, with the exception of S-21 director, Kaing Guek Eav, alias Duch, and the highest leaders of the regime such as Pol Pot, busts of whom are displayed in cages on the floor of the exhibition. 10 The broader context is discussed at length in the audio guide, but prior to this visitors were confronted with an 'almost total absence of information material' (Violi, 2012: 47), even though TSGM is seen as representative of other spaces around the country.
In the end, the space clearly shows the terrible violence wreaked upon the country by the Khmer Rouge. Various elements, from torture equipment and uncremated bones to purportedly authentic spaces, coalesce to create a space that is shocking and horrific to visitors (Buckley-Zistel and Williams, 2020;Isaac and Çakmak, 2016). A generalised sense of cruelty inflicted on victims, evidence and authenticity attributes responsibility for the violence and perpetration abstractly. Given the sparse information before the introduction of the audio guide, for many years the demonisation of the Khmer Rouge was strongly generalised and any mention of culpability points only to the highest leaders. The introduction of the audio guide does provide more information but does not change the character of the exhibition's effect.

The universal victimhood mnemonic
The ideal-typical counterpart to the demonised Khmer Rouge perpetrator is the innocent victim, which became politically important in the 1990s in the context of Prime Minister Hun Sen's 'Win Win Policy.' One of the most prevalent ways of remembering the past is in terms of universal victimhood, in which almost everyone who lived under Democratic Kampuchea can be seen as victims, including people who were low-level cadres of the Khmer Rouge (Williams, 2018b(Williams, , 2019. Culpability is attributed purely to the leaders in the highest echelons of power while all others are perceived to have been coerced to participate. Levels of coercion were indeed very high; however, this mnemonic does not account for any agency of lower-level Khmer Rouge cadres (Williams, 2018a).
At TSGM, victims are constructed in both a personalised and simultaneously anonymised manner through the presentation of countless portrait photographs that were originally taken as part of the registration process at S-21 (Tyner and Devadoss, 2014) and take on significant 'iconic power' (Carrabine, 2017). The exhibition of these photos allows a degree of personalisation as you can see the people and their faces, although they become part of an anonymous mass, rendering an identification with the individual more difficult (Caswell, 2014: 11-12;Edkins, 2013: 141). Furthermore, there is no personalisation of the photos by attaching names or life histories, although there are discussions within TSGM today about changing the approach to these pictures and inscribing names to them, 11 which would constitute a significant change after 40 years of complete anonymity. The photos are now published on the archive website that was launched in 2021 and include some of the victims' names, although as yet no biographic information. 12 Currently, though, a homogenous victimhood is suggested and without personalising information one receives no opportunity to determine anything about the victims. This leads to a general construction by some visitors of all victims being innocent (Buckley-Zistel and Williams, 2020), demonstrated in one quote of S-21 being 'a place of captivity and torture for innocent, ordinary people at the hands of the Khmer Rouge'. 13 With the introduction of the audio guide, the personal stories of several victims of S-21 are told, 14 albeit not demonstrating any complexity in their life stories that might complicate their image as innocent victims. Also, while in the audio guide the experiences of the presented victims are personalised, this is not the case for the presented interrogator. 15 This victimhood and innocence is implicit and cumulative through the various sections of the museum.

Silences that allow for the construction of non-complex victimhood
These mnemonic role attributions have different foci as the generalised demonisation mnemonic focusses more strongly on culpability and blame, while the universal victimhood mnemonic is more interested in victimisation and innocence. There is no contradiction between the two regarding the roles attributed to the highest leaders of the regime, nor for the broader population, but there is a contradiction in how low-level cadres are remembered. However, in TSGM both mnemonics seem to effortlessly co-exist without presenting visitors with any contradictions and I argue here that this becomes possible due to a silence in the larger narrative at the memorial museum: the exhibitions at TSGM do not discuss that the majority of people interned, tortured and subsequently killed were purged cadres of the Khmer Rouge suspected to be internal enemies (Elander, 2014: 59;Manning, 2012: 170). Due to S-21's central position in the security centre network, many of the people killed here were former Khmer Rouge suspected of being internal enemies: of the estimated 18,133 prisoners at S-21, 16 over 60% were purged Khmer Rouge. 17 These people were high-, mid-and also low-ranking cadres who, until their arrest, were part of the oppressive system themselves, developing or implementing the totalitarian policies of the state, propagating the ideologies that justified Khmer Rouge violence or participating actively in that violence themselves. However, they were then arrested, imprisoned and interrogated, just as other people deemed counter-revolutionary, with torture as a regular feature in interrogations. Ultimately, almost all prisoners were executed when they had 'confessed to their crimes' and named other people they had supposedly collaborated with. As such, the majority of people killed at S-21 can be seen as complex political actors who are both culpable in their actions as perpetrators, and yet also victimised. While it is not new to think about complex political actors in the Khmer Rouge (Bernath, 2016;Elander, 2014: 59;Manning, 2012: 170), here I focus on how complexity is eradicated through this silence about the former perpetrator status of the victims of this memorial space.
Throughout the exhibition there is no mention of the high proportion of victims of the space being former Khmer Rouge. 18 Further, a good example of this silence is to be found in the audio guide when it discusses who the people were that were arrested and brought to S-21: 'At the beginning, many of the people arrested were supporters of General Lon Nol. [. . .] Detainees were mostly what the Angkar called "New People" -anyone living in a city, including monks, professionals such as doctors or lawyers. And their families. There were students and teachers, and foreigners too. Over 150 victims had worked on the staff here. But even they were arrested as traitors. And there were at least 89 children '. 19 Given the detailed enumeration of various groups who were killed at this place, the lack of mention of the high amount of Khmer Rouge cadres who were imprisoned and killed at S-21 is quite striking. Another example can be found in the most recent exhibition to celebrate 40 years of the memorial museum's existence. In one part of the exhibition and its catalogue are pictures of a flyer that was published by the Ministry of Propaganda and Culture before 1985 which lists the 1513 names of prisoners known at that time according to groups (Tuol Sleng Genocide Museum, 2019: 42-43). Notably, the first category is indeed 'cadres', listing 187 names, while the remaining 1326 names are assigned as workers, citizens, students, teachers, health sector, engineers, state officials of former regime, former regime soldiers and students and state officials from abroad. It is interesting that at this early point in time, the existence of some Khmer Rouge cadres as prisoners was acknowledged, but nonetheless still significantly downplayed their proportion, and did not find entrance into the permanently communicated narratives of the exhibition.
The way victims of S-21 are portrayed at TSGM is partially in line with ideas on an 'ideal victim', an idea of those victims who '-when hit by crime -most readily are given the complete and legitimate status of being a victim' (Christie, 1986: 18). Christie (1986) delineates five categories of an ideal victim: vulnerability of the victim; respectability of the project the victim was undertaking; blamelessness of the victim for where they were; the 'big and bad' nature of the perpetrator; and the absence of a prior relationship to the perpetrator. 20 This emphasis on innocence and blamelessness is inherently problematic as it rests on a dichotomous understanding of victims and perpetrators (McEvoy and McConnachie, 2012: 531) that precludes the complexity at issue here.
Applying these dimensions of the ideal victim to TSGM, how the 'big and bad' nature of the perpetrator is attributed, was discussed above as demonisation. Given the abstract portrayal of the Khmer Rouge, the category of an absence of a prior relationship is also assumed. At the same time, the victims are clearly perceived by visitors to TSGM as vulnerable and blameless for their presence, as is evidenced in many references to their innocence in visitor comments about the site (Buckley-Zistel and Williams, 2020). Given the totalitarian environment they were living in and the suffering they endured at S-21, it is certainly understandable to see these purged cadres of the Khmer Rouge as victims, but it is striking nonetheless that no mention is made of the fact that the majority of people killed had more complex biographies than one would expect from ideal victims. While there are references to individual Khmer Rouge functionaries who were imprisoned and killed here in the audio guide, these are not located within the larger context of a victim identity. 21 However, this is not surprising, given that this is congruent with broader memory politics in Cambodia. Furthermore, nowhere is Khmer Rouge cadres' innocence attributed to the fact that many were recruited as children. This is surprising given that child soldiers are easily classified as victims despite their actions (Drumbl, 2012;McEvoy and McConnachie, 2012: 533). In 2017-2018, a temporary exhibition called 'Children of Angkar' was curated at TSGM about children under the Khmer Rouge regime in which there was also information about children as soldiers and other roles in the Khmer Rouge; however, there was no problematisation of this, nor did it feed into any discussion of what this could mean for complexity.
It is striking, however, that the final criterion of the ideal victim -respectability of the project that the victim was undertaking -is not evident for victims of the Khmer Rouge purged at S-21. These 'complex political victims' (Bouris, 2007; on Cambodia see Bernath, 2016) thus break with the notion of ideal victims through their loss of innocence conceptually, although -and this is key to my argument -the respectability of the project is just not discussed, thus not precluding victimhood for former Khmer Rouge cadres, but foregoing the opportunity to discuss this complexity. As such, the victims of this space are not differentiated according to their culpability (and with this their complexity). They are not portrayed as forgivable perpetrators, but instead it is the silence regarding their status as former Khmer Rouge that allows them to be seen as innocent.
While discussions about the Khmer Rouge background of most prisoners do not feature in the exhibition or in tours, these topics seem to be reflected upon by staff members. One senior staff member said regarding the Khmer Rouge status of the prisoners that 'we didn't tell them [visitors] the truth. We still hide it'. 22 Another statement qualifies this a little by saying that 'we don't hide that, but we need the right time to talk to them [visitors]' 23 and that this occurs only occasionally with very motivated student groups. However, one senior staff member mentioned that when they conduct outreach programmes, the staff have to be very careful regarding this topic. When students pose challenging questions, they avoid these because of the possibility that the students could have connections to the government and that, if they were to report these topics, the staff of TSMG would no longer be able to conduct outreach. 24 These silences are thus enforced through external pressures by the government (see typology in Mason and Sayner, 2019), as any changes to the exhibition have to be approved by the responsible government departments, creating certain 'red lines' that cannot be crossed. 25 Besides this external silencing, staff members also self-censor as they are uncertain how the topic will be understood by visitors. For example, a staff member explained the lack of discussion of this topic as follows: 'to be honest, we dare not to reveal that [. . .] a majority of the victims here were former Khmer Rouge, because we see that, to us, these people still are victims -even though they were part of the Khmer Rouge.
[. . .] But we [are] afraid that when we [. . .] reveal this to the students they may have different thoughts.
[. . .] what we are afraid [of]: They might think that 'oh, those were former Khmer Rouge, so we don't need to really pay respect, because they had also done something wrong.' [. . .] But some day in the future we will try to reveal this '. 26 When the topic of talking about Khmer Rouge cadres as prisoners came up in other interviews, reactions were along these lines, demonstrating a high degree of reflexivity among staff members about the topic, but a perception that it cannot be discussed with the broader public in most situations. Another staff member explained: 'I never told the students that these photos are Khmer Rouge portraits. [. . .] I said all are/were victims. All victims [laughs]. It's not important to tell them about the Khmer Rouge portraits or the innocent victims, because all the photos here, we assume that they are all victims'. 27 The logic underlying this was a fear that visitors could react to knowledge about the more complex personal biographies in a way that they would no longer want to pay respect to these former Khmer Rouge victims. 28 In Khmer, 'Peal Sie Peal' (evil killing evil) is a common concept that basically exculpates the commission of bad acts and reduces ideas of victimhood of the people being targeted as they are perceived as evil, too, for example in gang violence. 29 As such, staff at Tuol Sleng are worried that if the complex backgrounds of the prisoners were known, a moral relativisation of the heinous crimes could occur and the victims of this space would no longer be afforded the expected respect. It is interesting that the staff members -as well as broader civil society actors I interviewed -silence the former Khmer Rouge status of many of TGSM's victims, but often have a keen understanding of the complexity anyway.
While there is a silence on the status of most victims of S-21 having been Khmer Rouge cadres, complexity is introduced in some smaller ways. The audio guide acknowledges and discusses the fate of some S-21 staff members and shows not only that many were recruited as children but also the indoctrination processes they were subjected to, reducing their agency and attributing victimhood. 30 This theme of Khmer Rouge cadres as victims is also developed in two temporary exhibitions 31 at TSGM that explicitly frame Khmer Rouge cadres as victims. The exhibition 'Victims and Perpetrators? Testimony of young Khmer Rouge Comrades' 32 presents former Khmer Rouge and discusses their memories of the time and why they joined. The individuals are mostly presented as victims of the regime, portraying the difficult working conditions, the suffering they experienced and their constant fear of being arrested and killed. Similarly, the exhibition 'Stilled Lives. Photographs from the Cambodian Genocide' 33 presents portraits of people who joined the Khmer Rouge revolution with their biographies. The exhibition tells us that of the 51 people presented in the accompanying book, all were recruited to the Khmer Rouge and most were fated to end up in prison, with 42 of them dying, mostly by execution. As such, the narrative of the exhibition very much asserts that young men and women who were recruited to the Khmer Rouge are destined for death and should thus be seen as victims. While complexity is introduced in these examples, there are two key limitations. Firstly, they do attribute victimhood, but fail to discuss any agency that these people may have had in their roles as Khmer Rouge. Second, it neglects the systematic nature of this complexity and the prevalence of these complex political actors at S-21. By highlighting only individual examples without referring to the vast amount of imprisoned Khmer Rouge cadres, the scale of the phenomenon is undermined.
This silence about the Khmer Rouge background of most of TSGM's victims allows the memorial to function as a space that clearly portrays the crimes of the Khmer Rouge and the construction of the Khmer Rouge as the brutal and violent regime that it was, while at the same time creating empathy for the victims, particularly through the iconic photographs. These photographs mask and silence the complexity as they preclude the possibility of someone being seen here as both perpetrator and victim (Elander, 2014: 59). Pathways of some S-21 staff are portrayed in a way that allows for some Khmer Rouge to be discussed as victims. As such, these both tie into broader societal discussions of culpability and victimhood. The silence avoids the fact that in this specific place the majority of the victims cannot be seen as 'ideal victims' with the ideas of innocence tied to this. Instead, their more complex biographies complicate easier black-and-white attributions. As such, it is rendered impossible to talk about complex biographies or go into the motivations of why people participated in any depth, creating deep frustration for staff members 34 and precluding in particular an important educational perspective.
With this silence, a part of the complexity of political action cannot be understood properly, and the grey zones of victimhood and perpetration that are so important in understanding mass violence are rendered invisible. As such, while silences at the individual level are portrayed as being potentially empowering and providing the possibility of agency (Mannergren Selimovic, 2020;Porter, 2016), here the silences function more to disable and facilitate forgetting, and as such the silences become a 'tool used to uphold hegemonic discourses and erase dissonant aspects of the past' (Mannergren Selimovic, 2020: 2). This ties into the type of memory politics that Connerton (2008) labels 'prescriptive forgetting' (p. 61). While this silence around the background of most victims may seem to only constitute a small detail of the site, it is key to broader understandings of complexity in violence and the memory of it.

Conclusion
In this article, I have shown how various facets of the government's political interest in remembering the violent past are represented at TSGM, allowing both for a generalised demonisation of the Khmer Rouge in the sense of broad culpability for the violent past and universal victimhood for anyone but the highest leaders. These two mnemonic role attributions serve important political purposes independently of each other, meaning that the government has no interest in remembering the past in complex terms. Instead, the two mnemonic role attributions are rendered compatible through the silence that hides that the majority of people imprisoned at S-21 were in fact Khmer Rouge cadres themselves. As such, S-21 takes on a 'representative role' for both mnemonics despite its very specific position within the broader network of security centres (Manning, 2012: 170). This silencing, however, precludes a discussion on the complexity of roles that ties into a lack of complex memory regarding roles, responsibility and agency in Cambodia more broadly (Bernath, 2016).
These insights are important not only for understanding how complexity is negotiated in postgenocide Cambodia, but also for the broader literature on memory studies. The article's key argument is that competing narratives can co-exist in various ways that render complexity more or less visible and comprehendible. This contributes to our understanding of how silences can be politically useful and how they can be implemented even when there would not necessarily even be a contradiction between complex attributions and other narratives. The consequences of this complexity are important for understanding violence, as well as for what it means for remembering and preventing violence. Without a more complex reading of the past, topics such as agency, responsibility, opportunities for resistance and the grey zones of suffering while being a perpetrator are rendered invisible. This has manifest consequences: firstly in how communities can talk about their past, marginalising memories beyond the dichotomous ascriptions; secondly, it precludes effective prevention efforts as it renders the past too simple and avoids the nuances in which agency can be developed.
Over the past few years much has changed at TSGM with the introduction of an audio guide, a reading room, a reflection space and testimony programmes, as well as an increased engagement with arts and music. Furthermore, staff have expanded educational outreach programmes and introduced the curation of annual temporary exhibitions. Educational programmes are being implemented in cooperation with schools, internal research projects are being launched and the archives are being digitised and professionalised. While much is changing at TSGM, the main exhibition has not been adapted and remains essentially very similar to the original curation. Importantly, despite all the implemented changes to the site, the silence and the ambivalence it creates around mnemonic role attributions as well as the complexity that this silence masks remain the same across these introduced programmes.
However, with the temporary exhibition on the 40-year anniversary of the opening of TSGM that opened in August 2019, a wealth of additional information is being presented to visitors in an unprecedented way. Furthermore, there are increased discussions on adapting the main exhibition at some point. If such reforms to the space become politically viable, it will be interesting to observe what changes to the curation are made. While the temporary exhibitions held annually by the museum in the last few years would suggest a considerably more modern approach to its curation, it will be interesting to see whether the silence remains regarding the identity of most people killed in this place or whether the curators decide to engage with the complexity of the place and the individual biographies of the people who died there. This could certainly present an opportunity for a deeper discussion of what agency, victimhood and culpability meant under the Khmer Rouge. from 1975 to 1979'. A further example, among many others, is at stop 30 of the audio guide that posits that 'The individual skulls here provide key forensic evidence'. The analysis of the audio guide is founded in the English-language version of February 2018 that I received from the commercial rental company of the audio guide, Narrowcasters. 8. Interviews with various staff members of TSGM between January and March 2018. 9. Interviews with various civil parties to the ECCC, as well as other victims of the Khmer Rouge between April and June 2018. 10. I thank Keo Duong for pointing this out to me. 11. In a survey conducted by TSGM in early 2019 among national and international visitors, guides and civil society actors an overwhelming majority of 91% (of 121 respondents) supported the inclusion of names and biographies. The staff generously shared the summary of their results with me. 12. Informal communication with a staff member. 13. Review on Tripadvisor from the year 2016 (drawn from data used in Buckley-Zistel and Williams, 2020 Figure 1.9 of Co-Prosecutors' Closing Brief in Case 002/2, Annex F.2. Older calculations put this number even higher at around 80%. 18. As noted in the second footnote, things are changing and a new inscription has been added since this paper was accepted. 19. Audio guide Stop 5. 20. Recent psychological work shows that these conceptual categories are empirically also constitutive in laypeople perceptions of victimhood and thus a useful metric for thinking about attribution of victimhood (Lewis et al., 2019). 21. For example, audio guide Stop 8 refers to Koy Thuon, former Minister of Commerce as one of the executed; also, the most recent exhibition catalogue dedicates two pages to the famed poet and alumnus of Tuol Svay Prey High School, the former school at which S-21 was located, Khun Srun who became a functionary of the Khmer Rouge and was purged.