How does policy learning occur? The case of Belgian mental health care reforms

This article asks how policy learning is achieved and whether and how it impacts on policy change. By drawing on the empirical case of Belgian mental health reforms, it shows that policy learning occurs through the very practice of policy-making. In-depth analyses of the process of preparing and devising, a current reform of mental health care delivery, called Reform 107, evidence that the transformation of policy learning – through verbal expression, inscription in documents or enactment in social situations such as meetings – is crucial to its impact on policy change. A phenomenological approach to knowledge in policy helps to perceive and describe the transformation of policy learning through practical actions and interactions involved in devising policy change. Analytically, looking at this transformation entails shifting the focus from big and visible changes in policy objectives and instruments to micro policy practices such as meeting and writing documents. Placing the focus on micro policy practices should not lead, however, to a disregard for the social context in which they develop. The interactionist concept of linked ecologies provides the means to consider social regulations influencing policy learning without underestimating their very ephemeral and contingent nature.


Introduction
In this article, we ask how policy learning is achieved and whether and how it impacts on policy change. We address this question by examining past and current policy reforms in the Belgian mental health sector. Since the World Health Organization (WHO), together with the European Commission, stressed the burden of mental health disorders, several European countries have made important steps towards a complete deinstitutionalization KEYWORDS Policy learning; policy change; mental health policies; policy practices; knowledge in policy of mental health care delivery. The notion of deinstitutionalization involves a shift from hospital-based to community-based mental health care.
In Belgium, from the early 1970s onwards, several reforms were directed towards deinstitutionalization and improving the integration of residential and community care. However, due to strategies of resistance and appropriation enacted by both types of services, these reforms instead resulted in an increase in sector specialization, differentiation and segmentation. At the turn of the twenty-first century, taking advantage of development of the WHO mental health strategy for Europe, Belgian mental health policy-makers stimulated further changes in the sector, through successive waves of pilot projects supporting the development of integrated care networks delivering mobile and community mental health care. They took a step further in 2010 and launched a big reform of mental health care delivery called Reform 107.
Through this article, we show how policy learning and policy change articulate together through policy practices (Freeman, Griggs, & Boaz, 2011) involved in preparing and devising Reform 107. It focuses particularly on the practice of meeting and writing a policy document (Freeman 2016), such as performed by two ad hoc committees, namely the task force and the think tank. It argues that the transformation of policy learning, through enactment in social situations or inscription in documents, is crucial to its influence on policy change.
We first present a framework for studying policy learning in practice. Based on constructionist approaches to policy learning (Freeman, 2006), this framework implies a conception of policy learning as practical, social and processual achievement. It places the focus on policy practices, which are viewed as concrete actions and interactions leading to policy change. However, given the decisive influence of social regulations specific to the Belgian mental health sector on past mental health care reforms (Thunus, 2015;Thunus & Schoenaers, 2012), we assume that policy practices have to be placed in the social context wherein they happen. The concept of linked ecologies (Abbott, 2005) is used to outline the professional, social and political actors involved in this context and the emergent and contingent coalitions in which they regularly engage. Analytically, the framework mobilizes the phenomenology of Knowledge in Policy -Embodied, Inscribed, Enacted (Freeman and Sturdy, 2014). A phenomenological approach to policy learning helps, indeed, in perceiving and describing the moment when policy learning passes from one form (embodied, inscribed, enacted) to another, and then influences policy change.
Second, we provide a brief presentation and sociological conceptualization of the Belgian mental health sector. This conceptualization includes the definition of three ecologies representing professional and specialized approaches to mental health problems and two inter-ecological alliances or coalitions, enacting conflicting strategies in relation to change in mental health policy.
Third, we start the empirical section by introducing the policy guide outlining Reform 107. In 2010, the presentation of this small blue-covered policy guide to local actors raised many questions. Local actors were still focused on previous pilot projects and did not expect a new reform at that time. Moreover, they felt particularly uneasy with key aspects of the mental health care model outlined in this document. Thus, we ask the question: where does this guide come from? By taking a close look at concrete policy practices involved in preparing and devising Reform 107, we highlight progressive changes in the involved actors' conception of the Reform. These changes or policy learning, explain most aspects of Reform 107. In this sense, we can say that policy-makers learned how to change Belgian mental health policy by acting together towards that objective.
Fourth, we discuss the particular interest of the empirical case of Reform 107 for research on policy learning and change. We argue that its main interest lies in the puzzling nature of Reform 107 and the ambiguity surrounding its conception. Ambiguity requires a great deal of sense-making. Sense-making, in turn, possibly leads to innovation (Weick, 2015). In this particular case, we show that sense-making leads to a skillful combination of innovation and path dependency.
To conclude, we summarize the main aspects of the practice-based and sociological approach to policy learning and change developed through this paper, before pointing to its strengths and limitations.

A framework for studying policy learning practice
Our framework for studying policy learning in practice has two main aspects: theoretically, it relies on practical, processual and social conceptions of policy learning and change; analytically, it implies a phenomenological approach to the transformation of policy learning through concrete policy practices.

Policy learning as social achievements
Our framework draws on sociological approaches to policy processes, particularly the interactionist sociology and the French sociology of organisations (Crozier & Friedberg, 1980;Friedberg, 1992Friedberg, , 1997. As noticed by Radaelli (1995), by introducing 'a perspective which views policy in terms of learning and network of interactions' (Radaelli, 1995, p. 160), French organizational sociologists opened the way for considering the 'dynamic aspects of the policy process' (idem).
On the other hand, by stressing that the meaning of things, for instance professions, work and public policies, changes through social processes, interactionist sociologists drew attention to the micro-events through which such changes are achieved (Bucher & Strauss, 1961;Corbin & Strauss, 2008, 2014Schatzman & Strauss, 2012).
By relying on those perspectives, we define policy learning as a practical, processual and social achievement deriving from collective actions and interactions and entailing changes in the meaning of mental health policies. Mental health policies are, in turn, viewed as emergent: they do 'not exist somewhere else in finished form, ready to be looked at and learned from, but [are] finished and produced in the act of looking and learning' (Freeman, 2006, p. 379).
First, by describing policy learning as practical, we mean that it results from concrete actions and interactions, for instance meeting and writing documents (Freeman, 2016), through which different types and forms of knowledge are brought together and related to the problem at stake. Second, by denoting the processual character of policy learning, we imply that it consists of successive and interrelated moves (Corbin & Strauss, 2008) in the dominant framework of ideas and standards that specifies not only the goals of policy and the kind of instruments that can be used to attain them, but also the very nature of the problems they are meant to be addressing. (Hall, 1993) In our framework, policy learning and change are thus closely interwoven: learning is achieved through action and expressed through/reflected in mental health policies.
Third, we assume that policy learning is social in a general and more sociological sense. On the one hand, policy learning is not limited to the political sphere (Hall, 1993). Instead, it concerns multiple actors, particularly professional, social and political actors involved in enacting mental health policies. On the other hand, policy learning does not happen in a social vacuum. It takes place in a particular 'context of action' (Crozier & Friedberg, 1980) involving interdependent actors with different and often conflicting strategies in relation to policy change (Thoenig & Crozier, 1975).
The interactionist concept of 'linked ecology' (Abbott, 2005) helps in mapping such context of action. Considered at the level of the mental health system, an ecology can be described as a collective actor. Abbott distinguishes three kinds of ecology: the professional, the political and the academic ecology. Here, we will concentrate on the professional ecology and its relationship to the political one. Such as defined by Abbott (2005), the ecology of professions denotes 'a set of controlled tasks, and the links between professions and tasks' (p. 248). Tasks become professional tasks following the efforts of various professional groups to define 'certain sets of social or psychological and biological phenomena in the process of fighting over the vast array of potential expert work in the society' (idem). The professional ecology (Abbott, 1988) is not independent from the political ecology: professionals are not able to stimulate change in their field without building alliances with actors from the political ecology, whose policy agenda creates or limits opportunities for change.
In the empirical section, the term coalition is used to denote alliances between two or more ecologies trying to influence policy change. Such interecology alliance might be conceived as 'advocacy coalitions' (Sabatier, 1988, p. 139) composed of 'people from a variety of positions […] who share particular belief systems' (idem). However, by contrast with advocacy coalitions, 'interecology linkages' (Abbott, 2005, p. 247) are not empirically verifiable and pre-existing belief systems. Instead, they are emergent coalitions, produced and reproduced through interactions involving actors from different ecologies. They become visible only by enacting competing strategies (Friedberg, 1997) designed to influence policy change.

A phenomenological approach to policy learning
The phenomenology of Knowledge in Policy -Embodied, Inscribed, Enacted (Freeman and Sturdy, 2014) is helpful in perceiving and describing the transformation of knowledge, as it occurs through the very practice of policy-making. 1 The phenomenology draws on a shift from restricted conceptions of (scientific) knowledge to broad conceptions of (social) knowledge (Radaelli, 1995). Following this shift, the issue of knowledge in policy concerns … not only expert opinions and social research; but also the transformation of expert ideas into the kinds of knowledge actually used by political actors; a knowledge in which research, information held by administration bureaux and even opinions expressed by the mass media are all intertwined. (idem, p. 164) Moreover, it borrows from research on organizational knowledge stressing that 'static and compartmentalized' approaches to knowledge prevented scholars from analysing 'knowledge (or, more appropriately, knowing) […] as an active process that is mediated, situated, provisional, pragmatic and contested' (Blackler, 1995(Blackler, , p. 1040. Following Blackler's invitation to shift from knowledge to knowing, Freeman and Sturdy provided 'a new phenomenology of knowledge based not on who knows what, how and why, but on the form that knowledge may take' (2014, p. 1): Drawing on a simple analogy with the three phases of matter -solid, liquid and gas -we argue that knowledge too exists in three phases, which we characterize as embodied, inscribed and enacted. Furthermore, just as matter may pass from one phase to another, so too can knowledge be transformed, through various kinds of action, between phases. (Idem) First, 'Embodied knowledge is the knowledge held by human actors and employed and expressed by them as they go about their activities in the world' (Freeman and Sturdy, 2014, p. 8). It refers to past experiences informing actions and interactions. It includes verbal and non-verbal, procedural ('know how') and factual ('know that') knowledge. Since embodied knowledge is not separable from human actors, it 'is only mobile to the extent that living human bodies are mobile' (p. 10). By helping human actors to make sense of complex situations, it nevertheless influences the course of collective action (Smith-Merry, 2014).
Second, inscribed knowledge is knowledge inscribed on material supports or in documents. Documents are the most frequent type of inscription in policy-making and organizations (Freeman & Maybin, 2011;Riles, 1998;Rot & Demazière, 2014). They are used to support collective reflection on policy problems, to bring different pieces of knowledge together and to circulate knowledge in meaningful and purposeful ways. The two properties of inscribed knowledge are, indeed, its stability and mobility. They relate to its ability to fix meaning and to carry that meaning away from the situated actions/interactions from which it results (Harris, 1998;Law, 1986).
Third, enacted knowledge is inseparable from action. Inscribed knowledge involves a predetermined script (Akrich, 2010). However, without the intervention of actors prepared to enact it, the ideas inscribed in documents remain latent: 'In the absence of action, knowledge remains latent: thoughts unspoken, skills not exercised, texts unread and instruments unused are indistinguishable from ignorance or nonsense' (Freeman and Sturdy, 2014, p. 12). Assuming that enacted knowledge is not separable from action means recognizing its uncertain and social character. As suggested by Freeman and Sturdy (2014), 'embodied and inscribed knowledge provide resources for action, and may constrain what action is possible, but they do not determine unequivocally what form that action will take' (p. 13).
Taken together, the concepts of inscribed, embodied and enacted knowledge help to capture the practical, processual and social nature of policy learning. At the same time, they provide the means to observe how important the transformation of knowledge is to its impact on policy change.

Methods
This paper is based on qualitative material collected through documentary analyses, semi-structured interviews and direct observations. Documentary analyses focused on policy and organizational documents. They provided data on past and current reforms of Downloaded by [207.241.229.243] at 07:44 22 October 2017 mental health care delivery. The documentary analyses were combined with direct observations of meetings between policy-makers and local mental health professionals (n = 62).
The observed meetings include policy meetings between government representatives and national and international experts; local meetings between front-line mental health practitioners and strategic meetings between managers of local mental health care institutions and community services. This article focuses on policy meetings. Observing local and strategic committee meetings nevertheless helped us to understand how local actors perceived Reform 107 and which aspects of this reform were at odds with their conception of policy change.
Semi-structured interviews (n = 65) were with policy-makers, mental health experts, professionals and managers of psychiatric and mental health services. In this article, we rely mainly on interviews with policy-makers and experts (n = 12). They enabled us to go deeper into their individual trajectories, professional experience, perceptions and strategies. On the whole, policy and expert discourses collected through interviews complemented data collected through document analyses and direct observation, thereby providing a comprehensive understanding of Reform 107.

The case of Belgian mental health care reforms
Considered as both a professional domain and a public policy field, the Belgian mental health sector started growing following the definition of mental illness as a policy issue in the second half of the nineteenth century. Until 1948, despite some attempts to medicalize psychiatric institutions (Wouters & Poll, 1938), they fell under the responsibility of the Ministry of Justice. After WW2, the transfer of responsibilities from the Ministry of Justice to Public Health heralded a period of increasing growth in this sector.
Following the institutionalization of psychiatric hospitals in the 1950s and 60s, two reforms successively led to the creation of community mental health care services in 1975 and psychosocial rehabilitation centres in 1989. These three types of services -residential, community and rehabilitation -gradually formed three specific but interdependent professional ecologies.
These three ecologies developed interdependently: each ecology not only developed in response to the failure of another in addressing some aspects of mental health problems, but Three ecologies 'Residential ecology' started developing in the 1950s, following a will to transform old asylums into medical institutions. Its continuing development is characterized by an increased professionalization of medical and paramedical staff and the specialization of psychiatric care. 'Community ecology' started growing in the 1960s following social criticism of psychiatric hospitals, which were viewed as isolating psychiatric patients from the rest of the world and neglecting the psychological and social aspects of mental health. It constituted fertile ground for the development of the professions of social worker and psychologist. In the 1980s, the sector's diversification was further improved following the institutionalization of 'rehabilitation ecology' , which shifted the focus from treating mental health problems towards developing functional abilities and social roles. Professionally, it has been developed by occupational therapists and mental health professionals from the behaviourist school. changes in one ecology implied changes in the next, through modifications in the sector's institutional organization.
The relationships between the three ecologies or their interdependencies, were regulated by conflicting strategies enacted by two coalitions. These coalitions are defined emergent alliances between actors from the professional, political and social ecologies. They formed and reformed over time by advocating the development of particular approaches to mental health and enacting protective/progressive strategies in relation to change in the sector.
By jointly enacting protective and progressive strategies, the two coalitions significantly influenced policy change in the sector. Such 'cross-regulations' (Thoenig & Crozier, 1975), i.e. regulations arising from the encounter of conflicting strategies, resulted in increasing the sector's diversification and specialization, instead of improving its overall integration: 'without a clear objective of collaboration in the interest of the whole system and not part of it, changes in the services offered would have turned the two sides into competitors' (Orenbuch, 1981, p. 117).
At the turn of the twenty-first century, faced with this situation and the development of the WHO mental health strategy for Europe, Belgian public health ministers took further steps towards deinstitutionalizing the system and developing integrated care networks. Indeed, by recalling the global burden of mental health disorders -'in 2004, mental health disorders accounted for 13% of the global burden of disease' (WHO, 2013) -and urging The 'traditional coalition' grew from an alliance between psychiatric hospitals, the professional group of psychiatrists, federal agencies such as the National Institute for Health and Disability Insurance and advisory bodies such as the National Council for Hospital Facilities (NACH). 2 It defends the autonomy of the medical profession and the self-determination of psychiatric institutions (which conflict with the underlying principles of community psychiatry). The NACH played the role of knowledge broker (Schön, 1971), translating international standards into loose concepts making room for appropriation by psychiatric hospitals (Thunus & Schoenaers, 2012). Thanks to the support provided by the NACH and powerful lobbies, the traditional coalition usually acted as an epistemic community (Haas, 1992), directly contributing to the devising of mental health reforms (Dunlop & Radaelli, 2013). By enacting its 'strategy of protection' (Thunus, 2015), consisting of leading the definition and implementation of policy change, the coalition succeeded in protecting its dominant position in the system. The 'reformist coalition' supported the growth of community psychiatry. It includes non-profit associations such as the Belgian Mental Health League; psychiatrists trained in community treatment; and progressive Belgian public health ministers, who are involved in international organizations such as the WHO. The reformist coalition's weak organizational integration is compensated for by the sharing of common values and knowledge. It constitutes 'a network of practitioners and policy makers united by a set of general values and an interest in mutual learning about best practice in mental health' (Sturdy, Freeman, & Smith-Merry, 2013). Since the coalition's ability to influence policy change depends on the presence of proponents of community psychiatry in federal administrations, its political resources are much more uncertain than those of the traditional coalition.
2 the nacH is a federal advisory body composed of three working groups: university hospitals, rest and care homes and psychiatry. the 'psychiatry working group' has played a leading role in defining Belgian mental health policies.
the member states to move towards comprehensive and coordinated mental health care, the WHO provides an appropriate context for launching new reforms. This was particularly true in the case of Belgium, where the number of psychiatric beds per 100,000 inhabitants was the highest in the European Union and was four times higher than the European average (Vrancken, Schoenaers, & Melotte, 2009). Two waves of pilot projects took place in that context. In 2002, the first wave intended to develop local mobile teams providing psychiatric home care and, in 2005, the second wave intended to test working conditions in mental health care networks and circuits through the development of therapeutic consultation. The pilot projects reflected a shift from centralized mental health policies to decentralized and loosely defined pilot projects making room for adaptation to local needs. They gave rise to very particular and local innovations but did not enable change at the system level, which remained centred on hospital care (Schoenaers, Cerfontaine, & Thunus, 2011;Thunus & Schoenaers, 2012).
By contrast, the new Belgian mental health policy, launched by Belgian federal and regional public health ministers in 2010, was directed to trigger big changes in the system's institutional organization. In the following section, by taking a close look at concrete policy practices involved in preparing and devising this reform, we will indicate different ways of learning by making mental health policies. The empirical analyses provided in this section will revolve around the process of preparing the legal and political context for starting the reform and writing a guide 3 defining the corresponding policy programme.

Reform 107: a small policy guide for big policy change
Reform 107 is defined in a small blue-covered policy guide signed by the seven federal, regional and community ministers competent for mental health and psychiatry. This 35-page book, entitled Towards better mental health care through the realization of care circuits and network, exposes a new mental health care model that implies the development of five care functions supplied through local integrated care networks. These functions expand from mental health promotion and prevention (function 1), to mobile teams providing psychiatric home care for acute (2) and chronic problems (3), to alternative residential facilities (4) and intensive hospital treatments (5).
The guide also sets out the implementation plan, which involves, first and foremost, the development of local exploratory projects designed to adapt this model to empirical reality. It insists on the strategic role played by the 'project promoter' (a psychiatric hospital freezing a percentage of psychiatric beds to make the necessary budget available for the local project) and the 'network coordinator' , who is responsible for setting five 'committees of function' and a 'strategic committee' . These committees are composed of mental health professionals representing all kinds of existing psychiatric and mental health institutions. They are responsible for the strategic planning of the local project's development. This part of the guide also details training programmes, including immersion courses taking place in partner countries (England, France and Switzerland), which are proposed to mental health professionals involved in the reform.
This policy guide was presented to local actors in May 2010. Following the speeches given by public health authorities, local actors looked astonished. On the one hand, local actors were still focused on the previous pilots -the therapeutic projects -and did not expect the announcement of a big reform at that time: suddenly, it was the 107! The implementation of the therapeutic projects had started three years before … In the meantime, they [policymakers] had seen how those projects worked out and realised that it would be impossible to evaluate their output quickly. Then, they changed their mind and moved on! When I look back, I have the feeling that when we were all involved in implementing the therapeutic projects, they were already thinking about a new reform. (interview with a therapeutic project coordinator, June 2010) On the other hand, the new policy programme was puzzling. Explicit references to existing mental health and psychiatric services reassured them that policy-makers did not intend to put the past behind them. In the meantime, they felt uneasy about key concepts that seemed very unfamiliar to them. All of them were wondering: where does this Guide come from?

Preparing change in Belgian mental health policies
Considered over the two decades during which Reform 107 was prepared, the change process included four turning points (Abbott, 2001) of very different natures. These turning points are key moments where policy learning induced significant changes in the reform's scope and objectives. They emphasize that policy learning is ongoing, is contingent on the occurrence of expected and unexpected events and is fundamentally social or dependent on other people's actions and interactions. As Karl Weick stressed by reflecting on sensemaking, it is only retrospectively that one can say 'ok, this is how we came to the decision of starting Reform 107' (Figure 1).
The first turning point followed the 2005, WHO Inter-Ministerial Conference on Mental Health in Europe (Helsinki). Following the conference, public health ministers of half of the member countries endorsed the Declaration and Action Plan (WHO Regional Office for Europe, 2015). This document, signed by the Belgian Minister, explicitly supported a shift from residential to community psychiatry. The WHO's objectives were then translated into the Belgian context:  This translation work was conducted by the NACH, in a way that was consistent with the traditional coalition's strategy of protection.
The second turning point was the mental health national experts' and policy-makers' visit to Birmingham: It is not trivial that our ministers had visited Birmingham together. I mean that several ministers … travelled abroad to see how other people were working. In my view, that indicated that we were going to start an ambitious reform, at last! (interview with the Federal Coordinator [FC], 2012/02) The third turning point happened in 2009, just after the shooting of children by a psychotic person in a day nursery. Following the so-called 'Termonde incident' , the Federal Minister for Public Health had been asked by service user relatives' associations to create psychiatric mobile teams. The Minister had released the necessary budget to support this project, but 'that attempt did not result in the development of mobile teams, but the budget and the idea remained unchanged' (interview with a psychiatrist, 2011/12). This idea was then integrated into a reform project, by defining functions 2 and 3 as consisting of mobile teams providing psychiatric home care.
The fourth turning point was the achievement of an Inter-Ministerial Draft Agreement on a new mental health policy. Given the sharing of competencies between seven ministers responsible for different aspects of mental health policies, starting a complete reform of the sector required, indeed, that the Inter-Ministerial Conference for Public Health conclude a draft agreement: The reform was supposed to start in 1997, but we had to wait until 2002 for a first draft agreement, which was changed to 2004, and finally we had to wait again until 2009, when the IMC said, 'Ok! We are going to start now' . (interview, Dir. FPS PH, 2013/02) On the 14th December 2009, the IMC took a step further through a series of decisions, namely, to launch a call for projects and an information campaign on the creation of a new mental health network; to set the schedule of the reform; and to conceive of a strategy that would engage the interest of the residential, community and rehabilitation ecology in the reform.
This set of decisions suddenly accelerated the process, injecting an urgency to translate the IMC declaration into an action plan. That task was allocated to a think tank, which started working in collaboration with a task force from January 2010.

Devising change in Belgian mental health policies
The task force and the think tank were responsible for the drawing up of the policy guide. The task force was composed of representatives of federal, regional and community public health ministers. It regularly met to discuss working papers prepared by the think tank. The think tank was composed of the director of the federal public health department's psychosocial service, its counterpart at the national health insurance service, the mental health policy adviser of the Federal Minister for Public Health and the project leader, namely the federal coordinator.
The work performed by the think tank, together with the task force, can be described as consisting of three main steps: (1) making sense of complex situations by assembling embodied and inscribed pieces of knowledge, (2) meeting to enact a common knowledge of Reform 107, (3) inscribing knowledge in documents. These three steps correspond to three types of policy practice -(a) sense-making, (b) meeting, (c) talking and writing documents -through which policy learning is achieved. They occurred over a very short period of time, in a very ambiguous context: There was something special, the deadline for the Guide was the 26th April; it was just before the collapse of the government. Thus, on the 21st May, we had a government which was only conducting day-to-day matters, and we were confronting an important political crisis. Under such circumstances it was not bad to show that we had been able to work together in a project that involved every level of power. (Interview with the federal coordinator 2012/3) In what follows, by taking a close look at the concrete policy practices involved in learning what Reform 107 exactly was and how to implement it, we will highlight the very processual and social nature of policy learning and how it informs change in the new Belgian mental health policy.

Sense-making
The first step in drawing up the policy guide consisted of making sense of the reform project by assembling knowledge inscribed in existing documents with knowledge embodied by the think tank's members. This step presents most properties of sense-making identified by Karl Weick: it was (first and fourth) ongoing and social, (second) retrospective, (third and fifth) grounded in identity construction, (sixth) enactive of sensible environment and (seventh) directed towards plausibility rather than accuracy.
First, the think tank's members agreed on the objective of shifting from residential towards community psychiatry, but embodied different kinds of knowledge of the mental health systems: 'We had a clear view of where we wanted to go, but we had to refine the concept … to be able to bring the different pieces of our expertise together' (interview, FC, 2012/01).
Second, the first three members of the think tank (a policy adviser and two managers of federal departments and agencies) embodied knowledge beyond mental health reforms. Their knowledge of the protective strategy usually enacted by the traditional coalition helps in understanding their decision to set only a minimal framework making room for 'the creativity of the actors' (Policy Guide, 2010): 'It is not a reform where the government stated: 'you have to do that' . Rather, we built a model and local actors are going to test that model' (interview, Dir. FPS PH, 2013/02).
Third, the federal coordinator was a newcomer to mental health policy-making. His appointment as federal coordinator reflected a political will to move towards a wider conception of mental health reforms, which should include not only the transformation of psychiatric hospital but the empowerment of the 'community fabric' (interview, FC, 2012/02): 'I think that my practical experience and the proposal I made to enlarge the project drafted in existing documents explained their decision to appoint me as federal coordinator' (idem).
Fourth, when it started considering the content of the guide, the think tank first drew on existing documents: The federal coordinator was in charge of writing the Guide, but it was not his own idea! At the very start, there had been the WHO Declaration …, then the NACH advisory notes and the draft agreement of the IMC. (interview, Dir. FPS PH, 2013/02) Explicit references to those documents not only legitimized Reform 107 but also connected it to past changes in the system, thus reassuring local actors that they would be able to cope with the new policy programme. The inscription, in the title of the guide, of a key concept previously defined by the NACH (the concept 'mental health care circuits and networks'), clearly fulfilled this function: it told local care providers, particularly psychiatric hospitals, that Reform 107 arose from knowledge accumulated through past policy processes.
Fifth, by contrast, the implementation plan defined in the guide was significantly influenced by the federal coordinator (reformist coalition), who acquired most of his experience of community psychiatry by participating in European projects: my generation has taken advantage of pioneering participation in the European Horizon, and Leonardo Da Vinci Programmes. Those projects have reinforced our belief in the efficacy of community psychiatry […] It was the very beginning of those kind of projects and, at that time, they included exchange between practitioners and long immersion courses; they were intended to train trainers! (interview, FC 2012/01) Sixth, the coordinator's own method of learning about community psychiatry thus influenced the training programme inscribed in the guide. When asked about this programme, he explained: we cannot afford to train every member of every mobile team. Thus, our idea is also to train trainers; we hope that those who are going to do immersion courses are the right persons and that they will transfer their knowledge to the other members of their teams. (interview, FC, 2012/03) Seventh, given the difficulty of starting a complete reform of mental health care, the policy guide appeared an acceptable and satisfactory result: 'even if it did not perfectly fit with what we were expecting, we were particularly pleased that the guide came into existence!' (interview, Dir. FPS PH, 2013/02).
The resulting knowledge of Reform 107 thus consisted of a satisfactory combination of various experiences, evidence and perceptions of the context in which the reform had to be implemented. This combination enabled some innovations in mental health policies, particularly the involvement of the community fabric and the training programme, but also the reproduction of past policy practices, as the recourse to project making for local creativity.

Meeting (or interacting)
In meeting, different worlds come together. Meetings are thus clearly helpful in (first) knowing different parts of a complex system; (second) confronting one's knowledge of policy change with others' perceptions and expectations; (third) enacting new knowledge and, by this very fact, generating unprecedented commitment to policy change.
First, the think tank 'made considerable groundwork [by] hearing the positions of psychiatrists, GPs [and] meeting trade unions, hospitals and psycho-social institutions' federations' (idem). In doing so, the think tank 'wanted to tell them where we were going and to hear their reactions' (idem).
Second, the think tank also learned about the concerns of regional mental health authorities, before translating them into its global framework: the think tank made decisions and then the document was submitted to the task force. From there, we came to know a series of unexpected effects and requests to adapt the content of the guide to the regional particularities. The matter at hand was to imagine how those different things would come to fit with each other into a complementary totality. (interview, FC, 2012/01) Third, the think tank and task force members not only expressed their respective knowledge and concerns in relation to the reform. By meeting together they enacted something new, and this very fact generated an unprecedented commitment to policy change: We, the federate entities, and I think that the federal also became aware of the importance of the reform as the process was unfolding. The scope of the project had significantly enlarged, and that brought about a deeper interest of each of us in the reform. (interview with a regional policy adviser, 2013/03) Learning enacted through meeting then appears socially constrained and uncertain, but also very powerful in changing the course of action. As emphasized by the last quotation, the very fact of being involved in enacting knowledge of the reform creates commitment to the reform. Moreover, enacting knowledge not only means expressing, and then reproducing, existing ways of formulating mental health policies; it also involves the creation of something new, or policy innovation. In this case, it led to a significant enlargement of the reform's scope.

Writing (or inscribing knowledge in documents)
Inscription enables the stabilization of policy learning. The following instance shows that inscribing knowledge in documents first implies success in representing policy learning enacted through meetings. In turn, representing collective learning means selecting, ordering and simplifying knowledge enacted through meeting. The meeting participants were thus fully engaged and struggled to control this ordering process: Our discussion concerned every word. There are no simple words, there are concepts that express something … It was matter of nuance, a painstaking task; every word was transformed according to each other's concerns and there was a constant suspicion about what the others wanted … (interview with the Director of the Regional Public Health Department, 2014/02) Second, inscription precedes communication. Accordingly, it supposes explaining the expected policy reform in a way that is acceptable and understandable by mental health professionals: [The] challenge was: how to write down a well-balanced project that would be both sufficiently comprehensive to allow for the participation of different kinds of services and precise enough to ensure that the global philosophy would resist attempts to turn the reform away from its objectives. (interview, FC, 2012/02) Third, inscription entails attempts at controlling the social environment in which the guide should circulate: local coordinators are in a challenging position […] We have written down very detailed functions and we are expecting that they succeed in attaining a series of objectives … We will try to cope with this situation through the coaching … But I don't know to what extent we are able to influence them, in comparison with the influence of promoter hospitals. is the opportunity to make it explicit and to circulate it widely in the mental health sector. However, without being enacted, inscribed knowledge remains latent (Freeman and Sturdy, 2014) -and, in the absence of actors prepared to enact it properly (Akrich, 2010), it can be ignored or diverted from its initial purpose. This explained the efforts made by the think tank to control the environment in which the guide had to be enacted.

Discussion
Why is the case of Reform 107 of particular interest for research on policy learning and change? The answer lies in the puzzling nature of the reform and the great ambiguity surrounding its conception. When Reform 107 was launched, it constituted a disruption, a rupture in the normal course of events. Following the national information meeting, both the excitement of policy-makers and the astonishment of local actors drew our attention to the policy guide. It led us to ask the question: how did this document come into existence? How did it take form?
Asking these questions supposed engaging with policy-makers, to retrace the process leading to the reform. This process appeared, in turn, surrounded with ambiguity; many conceptions of the reform coexisted, the sharing of responsibilities between several ministers made decision-making more complex and, when the Inter-Ministerial Conference finally decided to launch the reform, the political context was unstable, resources were lacking, and very short time remained to write the policy guide. All these characteristics made this situation a very ambiguous one (Weick, 1995).
This ambiguity created an opportunity for reformist actors to take leadership of the process. In this respect, the creation of the think tank resulted from an inter-ecological alliance between the mental health policy adviser of the federal socialist minister of public health, the director of the psychosocial department of the Federal Public Service for Public Health and the federal coordinator, who represented the professional ecology, particularly the psychosocial rehabilitation ecology. On the other hand, dealing with this ambiguity required a great deal of sense-making from policy-makers. Since the only thing they knew with certainty was their common will, following the visit to Birmingham, to realize a big reform, they had to learn what kind of reform was possible within such a context and how that reform could be implemented.
Reform 107 thus presented an opportunity to examine how policy learning was achieved in ambiguous situations and, particularly, to wonder whether it leads to policy change. More precisely, it offers the possibility to ask whether such a situation, by affording repeated enactment of the reformist strategy, is liable to lead to significant policy change, or paradigm shift.
Our empirical analysis first evidences that policy-makers almost totally ignored what Reform 107 was before meeting to produce a document describing it. Meeting and writing documents are not simple instruments used to convey ready-made policy programmes and decisions. Instead, they are policy practices through which policy-makers learn about their intentions and how to translate them in words understandable by those responsible for putting it into practice.
The think tank first learned about Reform 107 by assembling various pieces of knowledge together. These included evidence inscribed in different types of documents issued by national (e.g. the NACH) and international actors (e.g. the WHO); experiences embodied by the think tank's members, together (the visit to Birmingham) or separately (the coordinator's participation in European projects); and their perception of their context of action. Since the think tank members had different trajectories, meaning that they previously learned different things in different contexts, the final assemblage was not totally consistent. For instance, the federal coordinator's idea to involve all the 'community fabric' was not consistent with the concept of 'mental health care networks and circuits' , which had been devised by the NACH in the perspective of reorganizing specialized care. Both concepts were nevertheless inscribed in the guide. As suggested by Weiss, 'familiar material juxtaposed in an ambiguous assemblage are recombined in ways that are plausible given the context but unlikely to resolve the ensuing ambiguity' (Weick, 2015, p. 122). Thus, it might be argued that, instead of decreasing the original ambiguity, the think tank reproduced and transferred it to local actors in charge of implementing the reform.
Second, the think tank learned about Reform 107 by enacting knowledge of mental health policy through different kinds of meetings. By meeting, the think tank and task force members confronted their knowledge and expectations of Reform 107 with one another and discovered, finally, that the reform's scope had considerably enlarged. As Freeman remarked, participants in a dialogue are not only learning from each other, but also learning something new. In this way, the meeting is generative, constitutive not merely constative. Something is being made in the process of meeting which is more than what was brought to it. (Freeman, 2008, p. 12 ) The very fact of learning what Reform 107 was together generated, in turn, a great commitment to policy change.
Third, the think tank learned about Reform 107 by expressing it in words. Writing the policy guide meant providing a formal representation of collectively enacted knowledge of Reform 107. It supposed selecting and ordering ideas, negotiating their very meaning and finding words to express them. The resulting document embodied power struggle and disappointment induced by this ordering and formalizing process.
On the whole, in its most explicit and standardized form, that is, as inscribed in the policy guide, Reform 107 was characterized by both path dependency and innovation. Path dependency was mainly reflected by the recourse to exploratory projects making room for local adaptation and the appointment of psychiatric hospitals as the project's promoters. Innovation rather lies in the training programme, the function of coordinator and the global philosophy of the reform. This combination of path dependency and innovation was reflected though the fundamental and conceptual ambiguity inscribed at the core of the policy document.
This combination emphasizes that, though it took advantage of the ambiguity characterizing the start of Reform 107 to take leadership of the devising process, the think tank did not succeed in controlling policy learning and the ensuing policy change. The very practical, social and processual character of policy learning explains this absence of control.
The think tank nevertheless succeeded in producing a real commitment to change in the sector; but what change? Local actors will learn about it throughout the implementation process, by enacting the policy guide locally and arguing with the think tank's members each time they question the accuracy of local translations. For this reason, we might conceive the implementation as a process of learning -by enacting mental health policies, rather than executing a predefined and non-problematic policy plan.

Conclusion
The case of Reform 107 emphasized that policy practices such as meeting, talking and writing documents are not simple instruments used to achieve or to implement predefined policy decisions. Instead, they are social processes through which policy ideas are gradually transformed and new groups are progressively created.
Theoretically, the linked ecologies argument enabled us to account for the constant formation and reformation of social groups involved in framing policy change. In our view, recognizing the social embeddedness of policy learning and then the significance of coalitions of actors and the corresponding belief systems (Sabatier, 1988) should also lead to recognize its very dynamic, practical and processual nature. As we have seen through the case of Reform 107, coalitions of actors are unable to control policy learning and ensuing policy changes. Instead, coalitions of actors are forming and reforming through the very process of policy learning, by appointing spokespersons and gaining commitment from new allies. According to the linked ecologies argument, coalitions are but in process. They result from changes in their environment and social interactions through which that environment is in turn enacted.
Analytically, the phenomenological approach to knowledge in policy provides the means to attend to the transformation of policy-relevant knowledge through practical actions and interaction such as meeting and writing documents. The shift operated by this phenomenological approach, from the types of knowledge and knowledge holders to the forms that knowledge takes, is crucial for apprehending the great diversity of experiences, evidence, information and perceptions involved in policy learning. Not only does this phenomenological approach a priori not exclude any type of knowledge from the picture, but it helps in perceiving its transformation. As we have seen through the case studied in this article, this transformation, through verbal expression, discussion and inscription, is crucial to the impact of knowledge on policy change. Collective and socially mediated enactments of knowledge are at the core of policy learning and account for much of its impact on policy change.
Concerning the analysis of policy learning through practical actions and interactions, a second point that might be made at the end of this paper is the great significance of 'the meeting' (Thunus, 2016). As we have seen, the meeting is a place where different worlds come together (Schwartzman, 1989). In this respect, it offers great opportunity to observe inter-ecological encounters and the formation and reformation of strategic alliances. But meetings are not only places where different worlds connect, they equally constitute temporary links between different levels of action, from the macro to the micro level. They stimulate micro-interactions through which wider organizational, political and professional systems are enacted. Finally, the generative character of meeting must retain our attention. Indeed, whilst most studies evoking meetings stress the poor quality of decisions and knowledge they produce -a result that our empirical analysis did not really contradict -the case of Reform 107 clearly emphasized the power to gain commitment from various actors to the process.
The limitation of our practice-based, social and processual approach to policy learning and change is its very context and case-specific character and the long and detailed fieldwork it requires. The comprehensive understanding of policy learning and change put forward through this article requires, indeed, various kinds of data (texts, discourses and observations) to be collected over a certain period of time. Such deep sociological enquiry is not easily replicable. Empirical analyses such as those provided in this paper nevertheless afford more general statements to be formulated relating, for instance, to the importance of meeting in enacting policy learning. Such statements might stimulate innovative research programmes in the fields of organization and policy studies.

Disclosure statement
No potential conflict of interest was reported by the authors.