Doing policy: enacting a policy assemblage about domestic violence

ABSTRACT This article explores how public policies regarding domestic violence aim at assembling a fragmented domain of views, attitudes and practices in a coherent manner. We propose to approach policy from an object-oriented anthropology, which makes it necessary to understand how objects come into being and to explore their ontology. We argue that policy objects, such as domestic violence, become real and multiply in practice when they associate in an assemblage: their ontology is relational. This implies an emphasis on motion, non-coherence and multiplicity in the study of how policy objects are enacted. We illustrate this approach by sharing a narrative of the development of a policy instrument, the reporting code on domestic violence and child abuse in Rotterdam (the Netherlands). The case study is focused on the interaction between the reporting code and its normative inscription, which is contested by a number of professionals. In order to demonstrate that policy instruments aim at constructing a new social order, we draw special attention to various modes of syncretism that are applied to create coherence.


Introduction
This article explores how public policies regarding domestic violence aim at assembling a fragmented domain of views, attitudes and practices by representing a range of different enactments of domestic violence as seemingly coherent. We view policies as meaningful but temporary, if not ephemeral social actors that constitute a powerful network, and therefore, too, we define policy, following Latour (2005), as an object. Accordingly, we propose to approach policy from an object-oriented anthropology, which makes it necessary to know how objects come into being and to explore their ontology. We define this as an objectoriented approach, because we would like to bring objects and the practices that sustain those objects to the foreground of analysis and emphasize the need for a shift of attention from social categories, such as ethnic groups or economic classes, to a different range of objects such as assemblages (Rabinow 2003, 76-77). We argue that policy objects, such as domestic violence, become real and multiply in practice when they associate in an assemblage: their ontology is relational. This implies an emphasis on motion, non-coherence and multiplicity in the study of how policy objects are enacted. We illustrate this approach by sharing a narrative of the development of a policy instrument, the reporting code on domestic violence and child abuse in Rotterdam, a cosmopolitan city in the Netherlands. As an element constructed in a policy assemblage, the code may be considered a nonhuman actor which makes the non-coherent practices in the field of domestic violence to cohere in an assemblage. The code, however, is contested by a number of professionals, which reveals the normative implication that policy instruments may be introduced in order to construct a new social order. In the analysis of the case study, we draw special attention to various modes of syncretism that are applied to unite the multiple linkages between each and every actor involved in the policy assemblage. It will confirm, to paraphrase Mol (2002), a policy multiple.

The relational ontology of policy assemblages
An anthropological approach of policy has much in common with interpretative analyses of policy as they share the same interest in meaning and how those meanings are produced. The analysis of our case study in this paper is also grounded in such an approach. At the same time, we seek to contribute to the development of the field of critical policy studies through the integration of an anthropological or interpretative perspective in the approach of actor-network theorists, such as Latour, Law and Mol. Accordingly, we propose to view policy no longer as a unified concrete thing, but as a fluid object in the sense of an emerging assemblage that simultaneously intermediates and transforms relations between actors in a certain problem space, and which is enacted in a myriad of everyday practices. The enactment of policy assemblage in practice, in turn, changes the assemblage itself again as it invariably involves multiple exchanges and interactions between all actors involved. The activities that occur in those practices, however multiple and different they may be, are closely associated or tied together to reach a sense of 'unity' in a bewildering world of multiplicities. Following Latour, we argue that multiplicity is the property of the object itself, rather than humans interpreting that object from different point of views. In other words, a policy viewed as an assembled object allows 'to be deployed as multiple and thus allowed to be grasped through different viewpoints, before being possibly unified in some later stage depending on the abilities of the collective to unify them' (Latour 2005, 116). We will demonstrate that this view of policy opens up and contributes to the field of critical policy studies, because it stresses the importance of 'the gathering of humans and nonhumans itself' as the site of the production of meaning and subjectivities.
The view of policy as an assemblage was introduced by  in the introduction of their seminal volume 1 in which anthropological perspectives on policy were reconsidered and further developed. Their view of policy as assemblages was a response to criticisms on earlier work (Shore and Wright 1997), in which they proposed a so-called governmentality approach that soon was rendered as 'too Foucauldian', since it did not leave sufficient space for individual agency (Shore and Wright 2011, 17). In response to these criticisms,  countered the implication of seeing subjects as 'docile bodies' who are disciplined without resistance or contestation. Instead, they proposed to view policies as 'actants that have agency and that change as they enter into relations with actors, objects and institutions in new domains ' (2011, 20). They claimed that if we wish to understand policy as a fundamental organizing concept in contemporary societies, we need to shift our attention to how policies develop and how they are enacted in practice. Shore and Wright (2011, 20) see this as a key contribution from an anthropological perspective to the interdisciplinary study of policy. To make sense of those enactments, we need to study objects in practice and pay as much attention to the material or the nonhuman part of policy as to the immaterial or human part of policy. Domestic violence policies, for example, have countless human and nonhuman elements that play a role in the practice of reducing domestic violence. There are different definitions and ideologies of domestic violence, different instructions for police, epidemiological findings, shelters for battered women, perpetrators, accounting reports and so on. Some of these elements make stronger connections than others (Latour 2005;Callon 1986). The political process of defining what domestic violence is, for example, has a strong connection with criminological data, and the methods used to describe how many people suffer from domestic violence and what characterizes the nature of, say, victimhood. Therefore, we need to assign ourselves the task of tracing the connections between those heterogeneous elements and trying to figure out how those different networks and the practices that sustain them coexist in a policy assemblage. In other words, we need to explore its ontology, not only by describing the distinctive features of policy objects, but also what they are doing in practice when the object is interacting and interfering with other objects and perhaps also in other practices. 2 We will not so much focus on the historical ontology as in many governmentality studies (see for example Hacking 2002;or Walters 2012), but rather refer to relational ontology stemming from anthropological perspectives on science and technology and from actor-network theory. This ontological approach refines key concepts such as multiplicity and non-coherence. Furthermore, it makes it possible to use a more adequate time frame by situating our stories in the contemporary.

Some characteristics of policy assemblages
An assemblage is 'the product of multiple determinations that are not reducible to a single logic' (Collier and Ong 2005, 12). Following this approach, we cannot view policy or policy instruments as concrete objects, but instead, we should view policy as a network with different modes of ordering that seemingly hang together. As we will highlight in our case study, it is the interaction within these assemblages that enables or disables the course of action of other agents (Latour 2005, 71). As we already mentioned, a policy assemblage is made up of a network of heterogeneous elements, at once constituting, multiplying and transforming it. A multitude of actors and their agencies will inevitably lead to controversy, self-reflexivity and subsequently motivate the emergence of new policies and politics, which is indeed a distinctive feature of policy assemblages. It captures the radical 'uncertainty, non-linearity and contingency of change ' (McFarlane and Anderson 2011, 162) that accompanies policy and the conceived problems it seeks to address.
An advantage, as we see it, of conceptualizing policies as assemblages is that it allows the policy analyst to be critical of seemingly inevitable and sometimes timeless depictions of solutions of contemporary problems. It raises questions of alterity because different practices will show different logics and, thus, potentially different solutions. In our case study of the reporting code presented below, we will unravel not only the process of its emergence, but also show how this instrument was assembled from different logics and domains in a seemingly coherent instrument. Other alternatives for the perceived lack of reporting and addressing violence and abuse in time could also have been materialized. This is another characteristic of assemblages: they 'draw attention to the ways in which meanings are materialised in practices, in settings, in the ordering of things and in how such conditions shape the possibilities of thinking and acting' (Clarke et al. 2015, 49-50).
Assemblages, through their persisting emphasis on a relational ontology emerging from practices, also provide new perspectives and a new vocabulary for policy analysts to avoid a singular view of reality. Practices disturb singularity and raise questions with regard to the comprehensive and occasionally decontextualized use of analytical concepts such as 'globalisation', 'neo-liberalism' (Clarke et al. 2015, 51-52) or, for that matter, 'authoritarianism'. We are not arguing that those tendencies do not exist or do not have a profound effect on the lives of people. Rather, we argue that 'thinking with assemblages' draws attention to the process, messiness and hard work of establishing relationships in and between different sites, actors, ideologies and technologies leading up to a topology of different practices in which multiple versions of the same policy objects coexist. Economic, political or symbolic power structures make up 'just' a few of the many elements in a policy assemblage. The way a policy assemblage, as we shall show in the case of the reporting code, configures itself through time and space will nonetheless affect professional ways and the lives of people on the receiving end of policies. At the same time, however, the agency of professionals and target groups and the connections they make are as much part of the production of the assemblage as more powerful actors who have more access to economic, discursive or symbolic means to establish stronger associations (Kingfisher 2013).
This approach of policy as an assemblage allows for a specific contribution of anthropology to the study of policy: it will draw attention to the fact that policies invariably attempt to capture in a singular and linear perspective a reality that is inherently noncoherent since it is made up of multiple layers that are frequently contradictory. In an anthropological analysis of policy, it is therefore important, as  have suggested, to take into account the actors' own metaphysical reality, to describe their view of the world and how those views open up new possibilities and engage in different associations making up a policy assemblage. This is important if we aim at understanding why politics and policies rarely succeed in engineering society. Unexpected events always take place. As a consequence, assemblages are inherently unstable and may be regarded as 'a distinctive type of experimental matrix of heterogeneous elements, techniques and concepts' (Rabinow 2003, 56). For that reason, too, it can be enlightening to reconstruct policy assemblages from an ethnographic description of actual objects, actions and practices in order to gain better insights into the multiple, nonlinear and ostensibly obscure relations.

Multiplicity and non-coherence
Multiplicity is a complex aspect of policy assemblages and is the product of heterogeneity and the diversity of associations in established networks (Law and Singleton 2014, 384). Multiplicity refers to the different but simultaneously existing realities that are produced in practice. Domestic violence is not just one, concrete and stable policy object that represents one reality, as is often assumed. What domestic violence is, what it means and how it is approached in policy terms and actions differ from practice to practice and, of course, from time to time. We wish to illustrate this with a small example that relates to and provides contextual information about the case study that we shall introduce in the next section.
One strategy to reduce or even exterminate domestic violence is to focus on early warning signs. In political discourses, this strategy is normatively motivated and based on a singular view of reality. Its underlying rationality is based on the assumption that all forms of intimate violence are intrinsically bad and harm the victims in serious ways, whose emotional or economic dependency on perpetrators is taken for granted. For that reason, early warning signs should be noticed and acted upon, either by bystanders or professionals. It is assumed that if this is done effectively, then incidents of domestic violence will diminish. This rationality is rather straightforward and makes up the policy perspective, which assembles logics from, among others, feminist discourse, health prevention studies and ideas about accountability into a normative framework. This perspective, however, also assumes domestic violence to be a singular reality, whereas in practice, the reality of domestic violence is inherently multiple. After all, it assembles many different kinds of abuse and mistreatment (such as physical, mental and emotional abuse) and also different categories of people suffering from it (including children, partners and parents) into a single discursive category of domestic violence. In practice, domestic violence is even multiplied in different knowledge practices (stretching from pedagogy to pediatrics, from sociology to forensic medicine and from police stations to social counseling) that built upon an immense 'hinterland' of methods, techniques and epistemologies, resulting in different versions of domestic violence.
Notwithstanding the multiplicity of domestic violence in practice, singularity is underlying the lack of problematization of early warning signs. Somehow, the policy perspective suggests that it is relatively easy to observe indications of abuse or mistreatment, but that is hardly the case. A Dutch multidisciplinary guideline for medical professionals that is based on all available research and expert knowledge about recognizing, diagnosing and treating domestic violence, for example, states unequivocally that there are no specific early warning signs to recognize abused children or adults. There may be circumstances in combination with injuries or peculiar behavior that might point to domestic violence (Trimbos instituut 2009), but it is generally genuinely difficult to recognize early warning signs of domestic violence. Signs of domestic violence are never self-evident. On the contrary, a thorough assessment of a complex situation is almost always necessary. Furthermore, it requires significant experience to perform such an assessment and to deal with the situation. Thus, in leading policy perspectives, signs of domestic violence are 'out there' waiting to be seen, but in practice, those alleged signs are more fluid and multiple, and therefore require interpretation before they may be read as signs of abuse. The interpretation and assessment is enabled and constrained by concrete practices (and their methods, techniques and circumstances) in which this work takes place in multiple manners once again.
These insights are derived from actor-network theorists such as Annemarie Mol (2002) and John Law (1994). Mol, for example, argues that in a hospital, different disciplines operate on different logics and use different technology to establish a diagnosis and decide on the best course of treatment. Thus, a disease like atherosclerosis is enacted in slightly different ways. In a similar way, we argue that what domestic violence 'is' and 'how' it is dealt with depend on the revelation when and where this is the case (Mol 2002, 53). This is not a matter of different meanings or perspectives, but rather there are different yet simultaneously coexisting versions of the same object. Those different realities are non-coherent. We introduce the concept of non-coherence to point to the fact that there are multiple versions of an object, but that this does not mean that those versions cannot be assembled into a configuration, so that they collectively prevent the object from fragmentation. Although we argued that the leading policy perspective behind this specific domestic violence strategy is articulated in a singular reality, this does not mean that the policy practice in which this perspective is enacted is also singular. On the contrary, domestic violence policies are invariably concerned with differences and multiplicity, and the question how to address these in practice. To deal with multiplicity and non-coherence, those policies require coordination, distribution and inclusion, processes to which we shall refer as modes of syncretism. We will introduce those modes of syncretism later on as a tool to analyze the practice of assemblages.
So, how can we relate relational ontologies to a perspective on policy as an assemblage that focuses on practices and relations between actors and entities? First, in an approach of policies as assemblages, the materiality of policies is considered by accounting for practically every-thing. Policies, as tools of politics and administration, make up the glue that hold together, or aim at holding together, a patchwork of humans and nonhumans in a certain problem space.
Furthermore, a policy process is a matter of assembling coexisting practices and materials, including the numbers, protocols and scripts that they generate. These do not add up to one order or one practice. On the contrary, there are many: they are multiple yet they associate. 'Multiplicity is […] about coexistences at a single moment. To make sense of multiplicity, we need to think and write in topological ways' (Mol and Law 2002, 8). It is therefore relevant to follow the movement of objects in practice and see how they relate to each other and how they are 'enacted'.
The question, then, is how different practices as part of a heterogeneous patchwork of relations act together in a policy assemblage? How does policy operate? What kind of associations are being collected and connected in such an assemblage? To address these questions, we will first present a case study of policy making with regard to domestic violence in Rotterdam. Subsequently, we will make an analysis of this extensive case by employing modes of syncretism and show our interpretation of how policy is conducted.

A narrative of domestic violence policy in Rotterdam
Rotterdam is a city of 610,000 inhabitants in the west of the Netherlands. We present the story of a policy conflict in this city around the introduction of a new policy instrument aimed at increasing the reporting of domestic violence in order to achieve more effective prevention.
The narrative is reconstructed on the basis of an analysis of, among other things, policy documents, research reports, newspaper articles, political statements, televisions and radio programs. In the months of June through October 2013, we traced those materials from the Internet and made observations during a number of policy events. We followed the leads of a dispute between the city's administration and the community of social and medical professionals in Rotterdam, the CEO of a mental health organization in particular. The materials documenting this conflict collectively formed a network of sociomaterials representing smaller and larger events concerning the introduction of the reporting code as a new policy standard. We made use of search engines from different websites, such as the council's information search system of the city of Rotterdam, newspapers and Google. Our search efforts resulted in a number of 95 materials that we have used to reconstruct the case study. Finally, we made use of the tools offered by actor-network theory, such as 'modes of syncretism', to analyze the case and to understand how 'unity' is produced. This story, then, constitutes an anthropological representation and analysis of what happened during the policy process leading to the reporting code.

Contextualizing the policy infrastructure of domestic violence
In Rotterdam, domestic violence is perceived as a serious social problem. In 2008, for example, 8.2% of all citizens above 16 years of age reported in a survey that they had been a victim of domestic violence at some point in their live (Christiaanse et al. 2010, 59). At the beginning of this century, the local government launched a plan to tackle domestic violence more seriously. Around 2005, the approach to stop domestic violence shifted even radically. Until then, the policy of the local government focused mainly on subsidizing local initiatives of welfare and health-care organizations to stop domestic violence. Although some of those organizations collaborated in realizing a citywide approach, it was acknowledged that action protocols for police officers and professionals, amongst others, were lacking and that leadership and management from the local government was necessary to facilitate the cooperation of the myriad of organizations concerned with domestic violence (Scala Rotterdam 2003, 12).
In 2002, it was first recognized that the policy to stop domestic violence required a broader and more effective approach. In 2005, this was realized by replacing an advisory center managed by a local emancipation organization by a new advisory center and the start of so-called networks on domestic violence that covered all districts in the city. The new advisory center against domestic violence and the management of the networks became part of the public health department of the city, which ensured direct involvement of the local administration.

Marking professional responsibility
In 2005, the social democrats (PvdA) also published a controversial report about domestic violence concerning teenage girls of different ethnic groups in Rotterdam. One of the results of the inquiry was the estimation of counselors in the Cape Verdean community that more than half of all Cape Verdean girls in the city were victim of sexual intimidation or sexual violence committed by their fathers or friends of their mothers (PvdA-gemeentefractie Rotterdam 2005, 16). One of the recommendations concerned the introduction of mandatory reporting by professionals in case child abuse or domestic violence was suspected.
Interestingly, this recommendation of the social democrats was supported by their political opponents in the city council, viz. the main local, populist party called Leefbaar Rotterdam. Ms Anker, at the time alderman 3 for this party, made a similar statement on national radio in January 2006. She argued that physicians, imams, reverends and teachers, amongst others, should be obligated to report domestic violence and child abuse, even despite (medical) confidentiality. In case of neglect, she argued, professionals should be prosecuted under criminal law.
This statement was in accordance with the national political discourse on abuse, but medical professionals raised serious concerns since it deviated fundamentally from medical discourse. The association for general practitioners in Rotterdam responded immediately after Anker had made her statement. It argued that such a policy would hamper the relationships based on mutual trust between practitioners and patients. General practitioners expressed their concern that patients would avoid their assistance, entailing serious risks for the well-being and security of children and youngsters.
When Ms Anker resigned from the local government, Ms Kriens of the social democrats became the responsible alderman for domestic violence in the city's administration. She also intended to introduce the reporting obligation. Again, however, the professional community of Rotterdam resisted the planned policy. The resistance motivated her to note their objections and engage into a discussion with professionals. According to Kriens, professionals responded that signaling domestic violence and acting on it was already part of their daily routine. They made clear that they always evaluated identified cases on the basis of their medical training and professional rules of conduct. Kriens could partly accept this argument, but still concluded that vital information about victims in tragic cases was not shared and thus not acted upon adequately. For that reason, too, she argued that the local government had no option but to set a new standard. 4 In May 2007, the city's administration organized a debate that was attended by approximately 90 professionals from different sectors. This policy event may be regarded as a transition from voluntary aid to compulsory aid when domestic violence is suspected. It was assumed that professionals did not always raise direct questions in meetings with clients who may have been subjected to domestic abuse. The transition that was set in motion therefore aimed at changing the communication strategy of professionals in relation to clients. A hallmark of counseling is to ask 'open' questions and listen actively to answers in order to explore presumptions, but in the debate in Rotterdam, it was argued that this approach of clients must change. It was considered to be more fruitful to pose 'closed' questions, leaving no escape for people who would then have to reply to professional presumptions with a simple yes or no. 5 The shift from exploring to establishing a client's condition, however, leaves less space for selfdetermination by clients and perhaps, ultimately, also for the knowledge and skills of professionals.
On the materiality of the reporting code Toward the end of 2007, the debate led to the introduction of a new instrument called the Rotterdam Reporting Code on Domestic Violence and Child Abuse (Rotterdamse meldcode huiselijk geweld en kindermishandeling). This instrument materializes the transition called for by political leaders and a number of influential professionals in Rotterdam. We propose to analyze this as an alignment of political, professional and organizational practices into an assemblage that transforms social reality in different ways, although it remains non-coherent nonetheless. How is this accomplished?
The reporting code is introduced as a roadmap, guiding professionals when they are confronted with or in case they presume domestic violence has taken place. So, the code is functional about finding the right way for professionals. This was argued to be necessary in order to compensate for the supposed inadequacy of past professional practices. A new Dutch policy word was coined to describe this situation: handelingsverlegenheid, to be translated as 'shy to act'. Three underlying rationalities seem to inscribe shyness into professionals: the issue of domestic violence is considered to be taboo, professionals are lacking relevant education and, finally, the important issue of confidentiality.
Taboo refers to cultural inappropriateness to talk about domestic violence, let alone to address the issue when speaking with clients. Furthermore, it is argued that lack of knowledge about domestic violence is making professionals also less attentive to notice it: one cannot see what one does not know. So, this is one way of transforming the social: to see, speak and act on a hidden and painful issue by means of a policy instrument that guides professionals and contributes to a different ordering that is supposed to lead to a safer society.
But this is not all. The code is also about collaboration, about working together in the professional domain. It aims at preventing that organizations or professionals do not inform each other when they suspect domestic violence. Collaboration as part of the code is conceptualized as having two sides. The first is about working together with colleagues in the professional's own organization; to ask colleagues with, perhaps, more knowledge for advice, especially when one decides not to report a case but to offer or organize aid themselves. Employees are asked to be self-reflexive when it comes to domestic violence. In consultation with coworkers, they have to evaluate whether they act appropriately. If not, they have to make arrangements for other solutions and call in the support of other specialists, either inside or outside the organization. This, then, is asking questions about the inscribed knowledge and skills of professionals that seek to extend their personal boundaries to interpersonal inscription. Information, knowledge and skills are multiplied between two or more professionals.
The other side concerns collaboration with professionals from other organizations. Sharing information between those organizations and subsequent actions are viewed as an absolute necessity to tie up fragments of the social into some sort of ordering. This is what the code, as a mediator, and being a social thing in itself, is doing: connecting practices and transforming the social in such a way that it may cohere in a policy assemblage of domestic violence. The code's agency is pressing organizations and professionals to report cases to advisory centers, from where coordination will be initiated. Collaboration, then, is about evaluating personal and institutional boundaries, distributing knowledge and information about cases across those boundaries and coordinating an approach between different practices all aiming at stopping domestic violence.
Finally, the code is also about accountability, being responsible for actions taken or not taken, and to be able to answer for it, preferably in written records. Here, we should add a colon and extend the sentence with a few words: when things go wrong. This is important. It is precisely this addition that relates accountability to the construction of the code as a road map and its associated virtues. Professionals have to account for the road taken, whom they have consulted along the way and what actions they have initiated with or without other organizations. In this context, confidentiality is a major issue. How to consult and collaborate with other people when you are not permitted to exchange information about clients or act against their wishes? However, the law on confidentiality is not that rigid, as many professionals seem to think. At least, this interpretation of the law is clarified in the code at some length. It seems legally possible to inform other agencies about a case of domestic violence without the consent of the people concerned. But to do so requires some precautionary measures.
Following the road map, all steps taken should be written down, so the professional can justify his protocol when he or she has to account for her or his actions in case of a disciplinary dispute. There are at least two kinds of faux pas for which a professional can be held accountable. The first is when a professional has violated against confidentiality and is being summoned by a client. The second mistake is when a professional did not respond adequately to a case of domestic violence of which s/he has some knowledge. The steps provided by the code are rendered as the most adequate way to address domestic violence. In a sense, the code provides clarity for a range of different practices through its normative purity and its allegedly unambiguous implications for enactment in practice. After mapping any signs of abuse, the code subsequently leads professionals through a series of steps, such as consulting experts, engaging into a conversation with clients, assessing the collected information and, finally, to make a decision about the course of action required.
Guidance, collaboration and accountability are the three main ingredients of the reporting code that was composed by professionals and civil servants in such a way that it could be introduced in practice. Surprisingly, however, it is not about the victims and perpetrators of domestic violence, who are, what Law calls, manifestly absent (Law 2004, 84). The materiality and agency of the reporting code is about professionals and organizations dealing with victims, leaving out the victims themselves, let alone the perpetrators and bystanders. Still, it aims at achieving an improved situation and a different social order in a future in which domestic violence is absent.

Controversy about the nature of the new social order
The controversy about the reporting code began, as noted above, with the political desire for a reporting obligation. In view of the resistance from professionals, however, the city council proposed a reporting code instead of a reporting obligation. Nevertheless, one organization declined to cooperate after the final presentation of the reporting code. Lamé, CEO of a mental health organization (RIAGG), refused to sign an agreement with the city that would commit his organization to the implementation of the code. Why did he oppose the reporting code that aims at reducing or even eliminating domestic violence?
Lamé argued that he is not opposed to developing new policies regarding domestic violence. In fact, his organization is dedicated to this issue and offers counseling to victims and perpetrators. In the opinion of Lamé, however, the reporting code (and other instruments monitoring youth, such as the Digital Child File) does injustice to the work of professionals and, even worse, creates an 'illusion of safety' because of its emphasis on allegedly comprehensive collaboration between, for example, schools, police, social work, health care, social housing estates and community workers. As such, it suggests that severe forms of abuse and even death resulting from violence could become something of the past. Widespread surveying, monitoring, signaling and counseling is presented as a guarantee for everyone's safety. Lamé is highly critical of such political promises of paradise. In his view, such a paradise will never be realized, since there will always be 'crazy' and unpredictable people.
Lamé argued that professionals and (health)-care organizations were scapegoated by politicians. On national radio, 6 Lamé asserted that ultimately this will result in a government that forces (health)care-organisations to think and act alike. This will lead to a disqualification of the ones who have to make arrangements for solutions. Those professionals are now imagined as retards and capable of nothing but failure. This will derail the care system, as becomes clear from the treatment of professional ethics, such as confidentiality and the careless treatment of personal data. Everybody mails files into the world for just about nothing.
This opinion entailed a further escalation between him and the local government. The latter added special criteria to its subsidizing policy with which organizations had to comply. Obviously, these criteria aimed at the implementation of the reporting code and a warning systemcalled SISA 7for professionals. After more than a year of debating, RIAGG clearly had no intention to comply with this rule. Subsequently, the local government decided that it no longer wanted to finance RIAGG's prevention activities with regard to domestic violence. Some opposition parties in the city council claimed that these rules had been 'invented' to abolish RIAGG. Accordingly, they expressed their concern that this policy would in fact be detrimental to the victims of domestic violence. Alderman Kriens was urged to normalize the situation and grant subsidy to RIAGG, but she was unyielding. RIAGG was refused subsidies when it became clear that its board of governors supported the organization's CEO. This decision, in turn, caused RIAGG to go to court to challenge the decision of the local government, which event received significant exposure in local and national media.
Notwithstanding the political pressure, RIAGG won its case in court. It ruled that the grant had been refused on grounds considered incorrect. The judge also ruled that close reading of the reporting code revealed that the second step (of the original code) obligated organizations to report severe cases of abuse. This step in the code was unlawful and for that reason, too, the additional criteria of the city's subsidizing policy had to be repealed. This legal victory, however, did not automatically imply that Lamé also won the political battle. In fact, he lost when the Dutch parliament voted for the national implementation of a reporting code that was based on the example of Rotterdam and which had to be implemented nationwide as of 1 July 2013. At that moment, the administration of Rotterdam announced that 500 organizations had already signed the agreement to implement the code. Coherence at long last.

Doing policy: enactments and modes of syncretism in and between practices
How should we analyze this narrative in terms of relational ontology? Mol (2002) and Law (1994) described processes of non-coherence and multiplicity in practices that belong to a single place, such as an academic hospital and a laboratory. In our narrative, however, we are confronted with different practices in different places. The topology is expanded beyond the boundaries of a single organization. As a corollary, we have to take into account that policy practices are not confined within an organizational boundary. Following Law (2004, 143; but see also ;Strathern, 2004;Latour, 2005;Mol, 2002), we argue that practices or orderings, and the materials and realities they produce, have a 'hinterland' far beyond the scope of one practice and its enactments of reality. This is what the conception of an assemblage of practices implies. Our point is that public policies, as shown by our case study, are not confined to practices in one place. They are distributed across different places and require to be enacted publically to come to life, to generate action. Accordingly, we witness different kinds of enactment. The first kind of enactment, not necessarily more important than the second, is about the public politics of purification: the messy process of reordering boundaries between practices and making objects to coexist in an assemblage. A second enactment takes place when the novel ways of new policies reach the streets, become public and begin interfering in particular practices and lives, not just in discursive ways, but reworking the object by producing new instruments and technologies that are enacted recursively in those practices. Although politics and policy as its main instrument is encompassing all practices in an assemblage, a slightly different version of the object may emerge. How then, is this enactment accomplished? How is policy made?
Policy instruments, such as the Rotterdam reporting code on domestic violence and child abuse, try to cohere different yet coexisting practices in an assemblage, at least to some extent. Usually, this is rather difficult. In a recent paper,  observed, following Latour, that modern practices are both pure and impure: they perform unity and appear coherent, but underneath lies a heterogeneous patchwork of logics and knowledge that is messy and non-coherent. We argue that the same point can be made about policies, which are enacted for the purpose of purifying irregularities between practices, in order to construct the image of a different social order. At the same time, however, particular practices and the people involved inevitably enact their own version of the policy object, making the new image of a social order multiple and non-coherent. Our narrative about the reporting code in Rotterdam is a perfect example of this observation. It raises the question how policy assemblages hold modern practices together and how they achieve coherence in and between them?  suggest that a process of syncretism is in order. Syncretism may be defined as a continuous process or effort to make the non-coherent cohere. Each assemblage or policy practice is syncretic by nature to the extent that it involves a perpetual association and integration of heterogeneous elements. The specific composition of a policy assemblage may, as we saw with the introduction of the reporting code, change over time; new relations between elements from different domains may or may not be established. In such a setting, the concept of syncretism is a powerful analytical tool to trace those relations and to understand how they are turned into a seemingly coherent policy assemblage.
However, the concept of syncretism evokes, at least in anthropology, functionalist connotations of cultural fusion and the existence of pure forms, but that is not how Law et al. conceive of syncretism. On the contrary, they emphasize the different and messy processes and practices that are presented and performed as a unifying whole. Law and colleagues distinguish between the normative and descriptive genre of syncretism. In a normative genre, the focus is on defending the boundaries of a practice and thereby its purity. Juxtaposed to this is a descriptive genre, in which messy processes may be highlighted, thus allowing for the simultaneous coexistence of different, possibly conflicting factors or features. To enable these different genres of syncretism to operate simultaneously, several styles of syncretism may be deployed. Law et al. identify six styles in the narratives they enquired empirically: denial, domestication, separation, care, conflict and collapse. Mol (2002) identified, as we see it, several other styles of syncretism, such as distribution, coordination and inclusion. This list is not exhaustive as more modes of syncretism may emerge in other narratives (Law et al. 2014, 177). We shall not elaborate all these styles, but some of these may be identified in the narrative we presented, such as the styles of apparent conflict, domestication, distribution, coordination and inclusion, but also the styles of analogy and appropriation.
The concept of 'implementation' used in policy studies generally refers to a rational ordering, in which a policy is transferred to a next phase. This process rarely allows for a transformation of the policy as it is based on an illusion of a stable wholeness. Alternatively, the concept of appropriation allows for transformation each time a policy is transferred to a new domain, thus also allowing, besides the agency of the policy instrument, for the agency of other actors involved. After all, they translate policies that are enacted between practices into their own practice, thus also enacting their own version. In so doing, they make its agency cohere with other relevant or contingent factors or another logic.
The code has been appropriated in different ways in a diversity of practices (see also Verwijs et al. 2012;Inspectie voor de Gezondheidszorg 2013). Police officers, for example, report each signal of domestic violence because they do not aim at building a relationship of trust between themselves and the people they meet. Professionals working in communities or in welfare are normally neither accustomed nor trained to deal with domestic violence, so for them the code may be a useful instrument. For professionals working in health care or youth care, however, a reporting code is of limited value in addition to protocols that are already in place concerning all forms of abuse, or they may have little confidence in the follow-up after they have reported a case (Verwijs et al. 2012, 39-40). Against that background, they often pursue other channels that are more in line with their own professional practices.
Thus, appropriation is what occurs in practice, which is frequently rather messy. As such, professionals may or may not have problems with a reporting code, but other factors and contingencies also play a crucial part. Organizations may endorse a code only symbolically in order to avoid losing subsidies, as the case of RIAGG showed, turning the code into a sheer bureaucratic device. Other issues in the organization may require more attention. Professionals may receive, or they may perceive it as such, poor advice from advisory centers and lose faith in collaboration or, finally, the administrative obligations that are associated with the code are considered as too burdensome. All these contingencies are part of daily practices.
These enactments and the related processes of appropriation in practices are fluid and recur in a policy assemblage, where, in turn, they generate other enactments. We are not talking about macro/micro or local/global divisions, but we fear that in any policy analysis there is a real risk of losing a number of important connections. In a previous section, we mentioned the establishment of an advisory center. It existed already before the code became another policy instrument. Now, the code includes the advisory center. It could not even have existed without the advisory center. But the same is true the other way around: the advisory center is not the same without the code. To enact the object of domestic violence, it needs to provide advice, their version of domestic violence, to other professionals or citizens. The professionals of the advisory center also need the reports of cases by other professionals, mediated through the code, to generate action and perform their version of domestic violence. But asymmetries will arise during the process of appropriation: it will never be the same because enacted differently in different practices. Nevertheless, they are all part of the assemblage. We cannot attribute hierarchy to those different elements, one is not larger or smaller than another. Indeed, the topology of the assemblage is, what Latour calls, flat (2005, 171).
What happens when domestic violence brings up connections between different practices, which necessitates a politics of purification to make everyone speak the same language? The reporting code, for example, constituted an unfair political symbol for the allegedly inadequate conduct of professionals, at least to some medical professionals (e.g. in RIAGG). Some people in the medical domain, however, just like the city's administration, find it unacceptable that RIAGG is unwilling to cooperate. Thus, the CEO of a drug addiction treatment clinic emphasized the possibilities that open up in their organizational network when a case is reported. People receive support in dealing with their addiction and their debts, and victims are offered counseling. For that reason, too, he commended the initiative of the administration to develop, in consultation with professionals, the reporting code. He framed the development of the code not only as a wish from the administration, but also as 'our will', 8 leaving little space for alternative discourses. Lamé continued to resist this discourse of unity and refused to be subjected in this manner, resulting in controversy, marginalization and impurity.
In the narrative of the reporting code in Rotterdam, we distinguish, besides the style of apparent conflict, the style of domestication. This style is characterized by a reduction of the differences to make non-coherence cohere (Law et al. 2014, 180). This occurred when the local administration expressed a desire for a more resolute instrument in the form of an obligation to report abuse, but could not proceed since its goal contravened some criteria of professional conduct. Subsequently, the political repertoire gradually shifted from the viewpoint of domination to the viewpoint of domestication, which eventually resulted in the reporting code. Instead of imposing a singular political reality that did not cohere with professional standards, this mode of syncretism turned 'difference into something that coheres after all' (Law et al. 2014, 180). The policy problem was transferred, so to speak, from an untameable situation to a tamed situation. But what exactly was tamed?
The problem of domestic violence was not resolved. What was resolved involved a perceived disciplinary problem, at least for the time being. For the administration, the core problem was that professionals might recognize problems of domestic violence, but neglect to act on it. Furthermore, they do not share information about their suspicion with colleagues or advisory centers. As a corollary, many people continue to be subjected to domestic violence. The administration argued that when all organizations train their employees to recognize signals of abuse and comply with the reporting code, then the problem of domestic violence will decrease or even be eliminated. Nevertheless, dramatic incidents may always arise and undermine the ostensible coherence of policy. Then a temporary and superficial state of purity will be rendered impure again after which it will once more require to be purified. Along with three other colleagues responsible for domestic violence in Amsterdam, Utrecht and The Hague, the alderman of Rotterdam expressed that the code was seen as a first step in their strategy, while they were not prepared to exclude the possibility of a reporting obligation in the future (Vos et al. 2008, 16). 9 In sum, the non-coherence between the logic of political practice and the logic of professional practice was reduced by a shift in terminology (from an obligation to a code) and an emphasis on professional autonomy and accountability. This resulted in a code that differentiates between practices that are pure at the surface, but impure in practice.
Another genre of syncretism that is deployed in this context concerns the frequent use of analogies. Van Meijl (2013) drew attention to the role of analogies between the present and the past in indigenous accounts of cultural change. He observed that they render cultural change as timeless in analogies between contemporary and historical practices. They make comparisons between traditions in the past and in the present, not to deny changes, but to defy and withstand them (Van Meijl 2013, 16). Differences between the past and the present are neglected by pointing to the similarities, however superficial they may be (Van Meijl 2013, 31). Because of its practical ontology, the concept of analogy is also useful for the analysis of non-coherence in policy practices. In a policy perspective, it is even possible to view analogy as a style of syncretism not only to defy change, but also to promote change. We will offer an example of both forms.
RIAGG accused the city of 'policy violence' because critical, reflexive voices were punished for disagreement with public policies. Lamé and others referred to this method of policymaking as Stalinist, drawing an analogy with the communist era of the Soviet Union in which centralized, undemocratic and unifying policies derailed the lives of millions of citizens. The first association of this analogy evokes anxieties about a possible derailment of the medical sector following the unifying policies of the local government that are compared with the imposition of communist policies. The second association refers to a style of politics that is considered harmful for people who oppose supposedly oppressing policies. This use of syncretism is performed to emphasize the unifying dimension of policy in order to reveal the underlying non-coherence between practices. Thus, political and professional logic are disconnected with the aim to guard the borders defining the logic of professional practices and sustain an illusion of purity. Indeed, Lamé wished the code to be abandoned as an element in the policy assemblage.
The opposite may also occur: analogy is used not to defy but to promote change. Above we mentioned the plea of the four largest cities in the Netherlands to introduce one law to address domestic violence. The authors argued that this would provide a decent solution to several policy problems and also that it would set a definite standard based on the principle that domestic violence is intolerable. An analogy with legislation on the prohibition of the use of physical or psychological violence in the upbringing of children was drawn to promote a new law. Indeed, policies on child abuse have a longer history in the Netherlands than policies on domestic violence. By stressing the similarities between two different yet related domains of violence in private situations, policies on domestic violence could be represented as an expansion of policies on child abuse. And by doing so, both domains were unified and related coherently in a new assemblage, in spite of the multiplicity of underlying practices. 10

Concluding remarks
In this article, we have demonstrated, following , that policies, in our case: policies dealing with domestic violence, can best be viewed as assemblages. We have explored the relational ontology of policy assemblages and demonstrated how policy in all its complexity is practiced through a focus on multiplicity, connectedness and non-coherence. We illustrated this with a case study of the introduction of a reporting code in Rotterdam. The political view that professionals do not adequately deal with signs of abuse or mistreatment materialized in a policy instrument that was assembled from different logics and domains. The reporting code, as a nonhuman agent, aims at assembling different practices to perform a coherent policy, but at the same time, it transforms and even undermines established practices of social counseling and care. The reporting code is introduced as a 'road map', highlighting its essence as an instrumental tool for professionals. However, its instrumentality is not as 'innocent' as it may seem. The reporting code inscribes a pure and normative policy frame into professionals that emphasizes the need for collaboration and accountability. As a consequence, professional spaces are transformed through the specific ways in which the code is enacted and, subsequently, how new ties in and between practices are established.
We have also shown that dissonant voices, such as the voice of the CEO of RIAGG who tried to resist political interference through the use of conflict and analogy as modes of syncretism, did not succeed in redirecting the course of the policy assemblage. However, this did not mean that the reporting code has been implemented in all practices as intended. On the contrary, we have demonstrated that asymmetries arise during the appropriation of the reporting code in different practices resulting in different versions of the reporting code. As such, the reporting code is both a product of an assemblage and a generator of new policy assemblages. For that reason, too, we argue that a perspective on policy as assemblages provides alternative concepts and definitions in the field of critical policy studies.
First, an assemblage approach stays close to actors and draws not only attention to how human actors perform policy in different practices and sites, but also to nonhuman actors who are part of the same scene. This view corrects asymmetry in mainstream policy analyses which tend to treat human and nonhuman actors differently, as if they belong to different worlds. Somehow, nonhumans are absent in many policy analyses or they play a role only in the background. However, codes, as we have discussed, but also numbers, paperwork, time schedules, knives or drugs are as much part of the domestic violence policy assemblage and the messy patchwork it constitutes, as battering perpetrators, victims, interest groups and policy makers who try to introduce new regulations.
Second, we would like to highlight that assemblages perform policy in the actual contemporary. This characterization of policy dynamics is relevant for cases that are still unfolding, such as domestic violence policies. Anthropologists might collect and describe the modes of syncretism that appear in a myriad of narratives and enactments of lived realities and show how policy is practiced. We also propose appropriation and analogy as additional styles of syncretism to analyze the different kinds of enactment of policy and the way it seeks to order social lives in different ways.
This brings us to our main point: through our focus on non-coherence and multiplicity in policy assemblages, we are able to describe different versions of domestic violence that escape from hegemony and rationality. It enables us to describe the enactment of domestic violence policies in and between different practices and assemble these into an ethnography. Such an analysis revolves not around a single unifying policy object, but involves a policy multiple in all its complexity and demonstrates why policy can be so unruly and unpredictable in practice. (22 September 2015) that the reporting code should be modified.

Notes
Step 5 of the reporting code should be transformed into a mandatory duty for professionals to report child abuse to the authorities. This would unburden the professional with the difficult decision whether or not a report must be made. 10. In this article, we cannot do justice to the complex discussion about different definitions of child abuse and domestic violence. For convenience sake, we propose to understand child abuse as a diverse range of violence and neglect that parents or other caregivers impose on their children. Domestic violence, on the other hand, is commonly understood as a diverse range of violence between adults with an intimate, friendly or dependent relationship.