Topography and scale in a community-driven maternal and child health program in Eastern Indonesia

In 2007, the Indonesian government introduced Generasi, a community-driven development program to address village priorities such as reducing poverty, maternal mortality, and child mortality. When describing Generasi’s biggest challenges, program facilitators on the eastern Indonesian island of Flores used a geographic vocabulary of fields (medan, lapangan) and topography (topografi) that evokes the demands of supervising Generasi’s implementation across dozens of mountain villages with poor infrastructure. But their geographic language also extends metaphorically to the enduring problems of scale and governance. I analyze these discourses of topography and field in relation to the changing therapeutic landscape of maternal and child health services in the Manggarai highlands of western Flores, then follow Generasi’s scalar scaffolding from the meetings and clinics in villages, to the technocratic policy work in Jakarta, and to the academic spaces of Auckland and Cambridge. Medicine Anthropology Theory 47


Keywords
Manggarai, Flores, Indonesia, community-driven development, maternal and child health, therapeutic landscapes A multisited therapeutic landscape In 2007, the Indonesian government introduced Generasi Sehat dan Cerdas (Healthy and Smart Generation), an offshoot of its National Community Empowerment Program (Program Nasional Pemberdayaan Masyarakat, or PNPM), a successful community-driven development program. This program, here called 'Generasi', was designed to address key policy priorities at the village level, such as reducing poverty, maternal mortality, and child mortality. Under Generasi, more than 5,400 villages across the archipelago receive an annual block grant and, with the assistance of village facilitators, each village allocates these grants toward activities that support the improvement of health and education indicators. In Indonesia's capital city Jakarta, development technocrats monitor and refine the Generasi program, while multilateral and bilateral aid agencies provide significant technical support, including several sophisticated randomized control trials to evaluate Generasi's impact, subcontracted to social scientists with expertise in fields such as development economics, public health, and medical anthropology.
In 2013 and 2015, on behalf of the Indonesian government and its international donors, I conducted fieldwork in Manggarai Timur district on the eastern Indonesian island of Flores in Nusa Tenggara Timur (NTT, figure 1) province to study Generasi on the ground. When asked to describe the biggest challenges in implementing Generasi, the program staff in Manggarai Timur repeatedly used a geographic vocabulary of fields (lapangan), topography (topografi), and region and territory (daerah) that evokes the demands of supervising a government program across dozens of mountain villages with notoriously poor infrastructure. But their geographic language also extends metaphorically to the enduring problems of scale and governance in a national program subject to the principles of community-driven development. In this article I analyze the use of spatial metaphors of topography and field in relation to the changing governance of maternal and child health services in Indonesian villages, taking the Manggarai highlands as a case study example. Focusing on the language of topography and field also allows me to critically describe how Generasi is scaffolded at multiple scales, from its community participation meetings and monthly maternal and child health clinics in Manggarai villages, to the technocratic policy work in Jakarta, and to the academic spaces of Auckland and Cambridge. In my analysis I first draw upon the work of critical feminist geographer Cindi Katz (2001) and borrow her related topographic metaphor of 'contour lines', which allows for links and comparisons across localities, and in turn, helps us analyze the ways that a heterogeneous group of Generasi actors come together and imagine the 'scale making project' (Tsing 2005) of a national development program. I then introduce several ethnographic examples that illustrate how Generasi actors collect and collate the program's health-reporting requirements. These metrics link data to particular places, but nonetheless appear drained of meaning and value if we take the time to carefully observe the context in which these numbers are counted.
I set the stage here with a brief description of a day's fieldwork studying the Generasi program during my second visit to the Manggarai highlands in mid-2015. I joined a factfinding team charged with investigating the effects of a recent disruption in government funding to Generasi since January and other program outcomes in advance of the next impact evaluation. Our team consisted of three Indonesians from Jakarta -a social development economist and a monitoring and evaluation specialist, both from the World Bank, and an analyst from a development economics research firm based at the University of Indonesia -and two foreigners, an economics doctoral student based in Cambridge, Massachusetts, and myself, a medical anthropologist currently based in Auckland. We chose to visit Manggarai Timur because I had visited this district during my first study of the Generasi program in 2013 (Grayman, Anggraini, and Ruhanawati 2013), and this afforded an opportunity to observe changes over the past two years.
On the morning of our third day (of five) in Manggarai Timur, our team met per usual at the Generasi program office in the subdistrict of Bonde. The program staff in Bonde -a subdistrict facilitator, his assistant, a local facilitator, and three administrators -arrived late, each with good reason, not least being that program funds were still frozen, awaiting the green light from Jakarta to resume activities and salary back-payments that had been suspended since January. In short, there was not much work, staff morale was low, and the office had no funds for electricity. The assistant attributed her tardiness to flu-like symptoms (masuk angin) triggered by the strenuous walk to and from the village of Semang during the prior day's fieldwork. The local counterpart was late because his motorbike broke during the return journey from Semang, and he wondered aloud who would pay for the damages sustained on work duty.
Indeed, to reach Semang requires an effort, but not an atypical one for the mountainous interior villages of Flores. From the Generasi office, located in Bonde's administrative center beside the well-paved provincial highway that traverses Flores, we traveled by car for twentyfive minutes on a pot-holed and twisting road that hugs the side of a steep ravine until the road ends at Dantena, the village I visited in 2013. 1 From Dantena, a wide hilly path built unevenly with rocks continues mostly downward to Semang. We parked our car in Dantena, and continued on foot for another twenty-five minutes to reach Semang's village center. While sturdy motorbikes and their skilled drivers might pass, this is likely where the local counterpart damaged his. Halfway to the village center, Semang's dilapidated clinic, infested with active beehives, stands alone with a commanding view of the valley behind it. Past the clinic, Semang's neighborhood of roadside houses begins in earnest; along the way there are three evenly spaced communal sources of clean piped water on tap, in constant use for bathing, laundry, dishes, and nearby household use. Semang is divided into three hamlets. 1 All subdistrict (kecamatan), village (desa), and hamlet (dusun) names used in this article are pseudonyms to protect the identities of program staff and other stakeholders. To further clarify, Manggarai Timur is an identifiable district (kabupaten) on the map of Flores (figure 1), but kecamatan Bonde, desa Semang, and desa Dantena (all pseudonyms) are not.
With partial support from Generasi, each hamlet hosts their own posyandu, monthly maternal and child health clinics managed by the village midwife and staffed by five hamlet volunteers, typically in one of the volunteer's houses.

Method and approach
The description above provides a sketch of Generasi's therapeutic landscape (Gesler 1992(Gesler , 2005 and some of the 'stakeholders' (pemangku kepentingan) drawn together in it, reflecting the concerns of 'program actors' (pelaku program) that the topografi daerah, or regional topography, poses challenges to issues of scale and governance (a brief discussion of this terminology is provided below). As of this writing, I have spent thirteen weeks conducting field-based research on the Generasi program during three rounds of fieldwork (2013, 2015, and 2016), with three of those weeks spent in the Manggarai highlands and four weeks spent in program, donor, and government offices in Jakarta. 2 Some anthropologists maintain a strict distinction between scholarly and programmatic work, preferring not to take a role within development programs, not least because program staff 'frame problems in terms amenable to technical solutions . . . screen out refractory processes to circumscribe an arena of intervention in which calculations can be applied . . . [and] address some problems and necessarily not others' (Li 2007, 2). Other anthropologists combine the two roles, conducting analysis from within intervention programs, arguing it is an important, if not privileged, site for critical ethnographic insight. Good (2012, 531) explains that such a vantage point illuminates matters in a different way: 'the structure of health services, and the difficulty of initiating change' becomes 'evident to us precisely because we have been trying to build systems of care'. I stand with the latter view, writing as participant-observer in the role of a short-term consultant, drawing on more than ten years of work and research in Indonesia's development, health, and humanitarian sectors. Research from within these sectors and their programs allows for multisited ethnography (Marcus 1995), enables consideration of how an ostensibly community-driven health intervention is simultaneously national in scope, and affords a look into the effects of 'seeing like a state' (Scott 1998), perhaps more accurately stated as 'seeing like a governing agent' (Kipnis 2008, 282), upon therapeutic landscapes, and vice versa. The ethnographic data collected from this type of fieldwork -interview notes and transcripts, program and government documents, participation in project planning and dissemination meetings in Jakarta, travel itineraries, descriptive field notes including photographs, and observations of clinical activities, 2 Other locations include program and donor offices in Jakarta, as well as sites of program implementation (in 2013) in Sukabumi district in Jawa Barat (West Java) province, Pamekasan district on Madura island in Jawa Timur (East Java) province, and (in 2016) Gorontalo district in Gorontalo province on Sulawesi island. For further elaboration on the 2013 sites, see Grayman et al. 2013. particularly the aforementioned posyandu -allows analysis using some theoretical orientations to therapeutic landscapes, but forecloses others. Wilbert Gesler's (1992, 737) foundational text on therapeutic landscapes outlines a range of humanist and structuralist approaches, the former emphasizing 'subjectivity, individuality, creativity, the importance of experience, meaning, and value' but critiqued for failing to acknowledge underlying structures, and the latter emphasizing materialist or political economic ideas but critiqued for discounting the agency of individual actors. In a recent edited volume on the anthropology of landscapes, the editors and contributors also reflect on this persistent dichotomy between phenomenological and structural approaches (Árnason et al. 2012). Rather than seek to resolve such debates, the editors acknowledge that 'the two approaches . . . ask different questions and require different material in order to address' them; instead, they 'invoke the practice of ethnography as a means to move beyond, rather than resolve the dichotomy between experience and structure' (Vergunst et al. 2012, 12). They propose ethnographic interventions in the form of studies of: landscape lines that either facilitate travel (routes) or obstruct it (boundaries), the aesthetics of landscape, and the ways narrative and landscape mutually constitute one another. This article follows their suggestions in the following ways. First, I consider the landscape lines that alternately facilitate, distort, or obstruct Generasi's (and other actors') efforts to improve maternal and child health. This includes navigating the difficult physical topography of the Manggarai highlands, but to address questions of governance and scale requires a metaphorical topography as well. Toward that end I enlist a reading of critical feminist geographer Cindi Katz's (2001) analysis of the signature feature of topographic maps: the contour line. Next, I look at how the Generasi program operates through a projection of scale, in the community ownership of a national program. To do this, I use scholarship on governmentality (Rose 1999;Li 2007Li , 2011 to frame a series of martial metaphors that were invoked repeatedly by my interlocutors in ways that project mastery and inscribe hierarchy upon this therapeutic landscape while also conscripting communities into public health interventions. Throughout the analysis, I return to the narratives of program administrators, government officials, and Generasi stakeholders at the subdistrict and village levels. Indeed, it was the repetition of terms such as 'topografi', 'lapangan', and other spatial metaphors by various informants that inspired this article. An important methodological tension to acknowledge in multisited fieldwork is that ethnographers cannot possibly achieve the same experiential depth across every location. As noted above, I am an anthropologist with more than ten years of experience working with and studying development, humanitarian, and civil society organizations in Indonesia. I use this experience in my ethnography of the national Generasi program, focusing on spatial and scalar discourses, but I have less direct experience with the Manggarai highlands. To account for this shortcoming, I make use of anthropological scholarship from the same region, particularly the recent work of Catherine Allerton (2012Allerton ( , 2013, whose research in Flores fortuitously (for me) concerns the lived experience of the Manggarai social landscape. Her ethnography offers a complementary approach to my own, as she explores phenomenological perspectives drawn from years of living with the residents in a particular Manggarai community. Such perspectives were unavailable to me given the limited time I spent in the region, and because I pursued a different research agenda. Instead, my analysis of Generasi's nationwide maternal and child health intervention in the Manggarai highlands veers almost by default toward the political and structural aspects. One of the first critical insights that Allerton's ethnography offers is a reminder that therapeutic interventions in the Manggarai highlands have a long history.

An archeology of therapeutic interventions into the Manggarai landscape
In her account of Manggarai houses, Allerton (2013, 44-46) cites a colonial report on Manggarai architecture written by the Dutch administrator C. Nooteboom (1939), which describes what might be the first therapeutic intervention into the landscape informed by an emergent biomedical science. The following passage from Allerton's (2013, 222-23) ethnography quotes extensively from Nooteboom's report: Nooteboom outlines two types of large houses that were once found in Manggarai, "often inhabited by several dozens of families" (1939,221): an elongated, oval house and a round house. He noted that in many villages the entire population was housed in one of these large structures, which were "pitch-dark in broad daylight and always stuffy by the smoke of the fires and the presence of so many people." According to Nooteboom, the house-posts of these large structures were relatively short, creating a "hotbed of germs" in the space occupied by "dogs, pigs and children" underneath. The occupants of these houses were consequently "very prone to all kinds of diseases," and the continued existence of these structures was, "from a hygienic point of view," untenable. Therefore, despite the "obviously very important objections" that these houses were a "general sanctuary" central to religious ritual and performances, the colonial administration in Manggarai decided to have them all pulled down, so that by the time of Nooteboom's stay in 1934 they were "lost as a cultural element." In their place, the Department of Public Health ordered the construction of "model houses of strictly controlled maximum measurements and a maximum number of occupants," with taller house-posts and a nearby "model lavatory" (ibid., 222-223).
Prior to Indonesia's independence, Manggarai villages were built in the mountains partly as a security measure, to avoid slave raiders (Allerton 2013, 151). But in the name of 'sanitation' and 'development', and implicitly to make highland communities more legible for governance, the postcolonial state relocated villages to accessible lowland sites, in rectangular, single-family houses, built beside roads (ibid., 45). Although I did not solicit these older village histories during my fieldwork in the Manggarai highlands, it is safe to assume that the design and construction of the houses beside the rocky road are an innovation within the past fifty years. Houses with cement foundations and metal roofs (figure 2) were probably built within the past twenty years, the result of higher incomes, especially remittances from migrant laborers. When we asked one of Semang's posyandu volunteers about improvements in health, it was a telling association that she attributed the massive reduction (in her estimation) in gastrointestinal illnesses during the past decade to better quality housing, paid for by the remittances sent home by men working on construction projects elsewhere. I would also attribute the reduction in gastrointestinal illnesses to access to clean water. Multiple respondents recalled that the faucets we observed with such active use (as illustrated in figure 3) were installed in the early 1990s by an NGO funded by the Australian Agency for International Development (AusAID). Hayes 1998; Koblinsky, Marzoeki, and Harimurti 2010). 4 This brief archeology reveals how the Manggarai therapeutic landscape -conceived as a nexus of shifting and intersecting social relations -connects spatially and temporally with other places (Raffles 2002, 183;Massey 1994, 120). I use a multisited approach not just to track these spatial and temporal connections but also to apprehend how programs like Generasi produce scale, linking and nesting villages within their respective subdistricts through data collection and aggregation techniques, up to the district, provincial, and national levels (Tsing 2005, 58).
Despite policy nods toward 'coordination', donor-and state-driven health interventions are neither coherent nor monolithic, especially during Indonesia's decentralization era (Aspinall 2014;Jung 2016;Rosser and Wilson 2012). During my visits to Manggarai, we met health staff juggling the competing demands of multiple interventions mandated by different sectors and levels of the Indonesian government, in part driven by competing donor priorities. For example, my focus here on the national Generasi program, which is preoccupied with the 'demand side' of health services, cannot fully account for the provincial-level Australian Indonesian Partnership for Maternal and Neonatal Health (AIPMNH), which is preoccupied with the 'supply side' of health services. In theory, and in the words of policy experts in Jakarta, these two programs are complementary, but on the ground where supply and demand meet, they compete and confound, producing unexpected friction (Tsing 2005). The view of this therapeutic landscape changes based on where one stands, leading to 'polysemic' (Bender 2001) interpretations of which sites matter most in determining the causes of and solutions for poor maternal and child health outcomes. Competing policy priorities force the question: are government clinics or communities more accountable for delivering improved health outcomes?

Defining communities in a community-driven development intervention
In contrast with the health interventions that precede or compete with it, Generasi is distinguished by its introduction of a community-driven development (CDD) paradigm into the health sector. Although 'participation' has been an important operational element in health development programs since at least the 1978 Alma-Ata Declaration, which promoted primary health care as a means to ensuring 'health for all' (Brown, Cueto, and Fee 2006), the concept has been subject to wide interpretation (Morgan 2001) and critique (Cooke and Kothari 2001). The CDD paradigm, as it was imagined and implemented on a large scale by World Bank social scientists in Indonesia in the late 1990s, was the first to include insights from social theory, particularly from anthropologists. 5 Their intervention was to operationalize the concept of social capital, focusing on processes of development instead of specific outcomes, and empowering communities to plan and manage development projects of their own design (Bebbington et al. 2004).
CDD programs use the language of 'program actors' (pelaku program) and 'stakeholders' (pemangku kepentingan) instead of the hierarchical language of 'officials' (pejabat or petugas) and 'experts' (ahli) because the ideology of empowerment requires communities to take charge of their own development decisions. Formal Generasi program actors mainly include salaried facilitators and their administrative support staff at district and subdistrict levels, who are employed as consultants to the local government; and 'village empowerment volunteers' (kader pemberdayaan masyarakat desa) who facilitate program activities, conduct outreach, and receive small stipends for their service. Generasi stakeholders include government officials including elected village leaders, service delivery professionals such as midwives and teachers, and of course the 'beneficiaries' (penerima manfaat or sasaran) who participate in decision-making processes and program activities that will benefit them. In my experience observing how Generasi works, the stakeholders with expertise in health, education, and village governance more broadly, play decisive roles in assisting Generasi facilitators to refine community beneficiary suggestions into actionable programs. Recognizing that communities are hardly uniform or egalitarian, CDD projects include 'a strong battery of monitoring instruments' that are designed to ensure inclusive planning, prevent local-level corruption (or elite capture), and improve internal auditing (Guggenheim 2006, 113). In CDD projects, subdistrict facilitators manage a competitive bidding process among villages for project funds. Village facilitators ensure the inclusion of representative voices from remote hamlets in project planning meetings. District-level facilitators supervise the subdistricts, and feed data on all aspects of program planning and outcomes to national program managers in Jakarta.
From 2001 to 2003, the World Bank's CDD program accounted for more than half of its lending to Indonesia (ibid., 116). The Indonesian government assumed ownership of the program and continued its implementation through 2014. Under this original CDD model, communities could theoretically choose to use their grant money for any kind of social 5 Scott Guggenheim, trained as a social anthropologist at Johns Hopkins University, is one the lead architects of the CDD paradigm in Indonesia, and describes the early days of this policy under the World Bank in his chapter in the edited volume The Search for Empowerment: Social Capital as Idea and Practice at the World Bank (Guggenheim 2006). development that benefitted the poor, but in practice they typically chose infrastructure projects such as roads, bridges, and irrigation. To support this typical use of CDD funds, the vast majority of facilitators employed by the program had engineering backgrounds. In order to bypass this bias and to retrain focus on human resource development, the Indonesian government introduced the Generasi program in 2007, which uses the CDD model to improve health and education outcomes while also (it is hoped) stimulating demand and accountability for improved front-line service delivery from these two government sectors that reach most rural villages across Indonesia. In 2015, CDD principles continued in diluted form in the new Village Law (No. 06/2014), which in Jakarta transfers the task of disbursing village development funds from the Ministry of Home Affairs to the newly established Ministry of Village Development, and in the villages transfers the administration of these funds from local facilitators to village governments. Although this massive reorganization did not directly affect Generasi, the politics in Jakarta surrounding these changes indirectly contributed to the delayed disbursement of funds in 2015, as described in this article's introduction. In 2017, Generasi too will be integrated into the Village Law's implementation.
To incentivize (or, in the language of governmentality, to discipline) villages to improve maternal and child health outcomes, Generasi facilitators measure performance based on eight annual indicators. These are: 1. number of deliveries completed in registered health facilities; 2. number of routine pregnancy and postnatal exams with a midwife; 3. number of iron supplements distributed to women during pregnancy; 4. administration of vitamin A supplements for infants; 5. number of babies with complete immunizations; 6. routine attendance at the hamlet's monthly posyandu; 7. routine infant weight measurement at monthly posyandu; and 8. number of babies maintaining normal weight for their age. (PNPM 2008, 111-14) Village performances across these indicators are compared at the subdistrict level, and the best performing villages receive a bonus in their Generasi budget for the following year. Village annual budgets are determined by a formula that takes into account the population of potential beneficiaries and the barriers to access government services. Budget priorities are debated during participatory planning meetings, but the agreed-upon expenses must be in service of improving the annual indicators. This typically results in a short list of expense categories, such as transportation to clinics for exams and deliveries, including emergency deliveries at distant hospitals, and small honorariums to village volunteers or midwives for their service at the monthly village clinics. All facilitators, however, report that the largest expense in their annual budgets is food supplements, which have generally shown poor efficacy in reducing problems associated with chronic undernutrition such as stunting (Bhutta et al. 2013), either for treating individual cases or as a strategy to attract attendance at the monthly clinics. 6 Anthropologist Tania Li (2007Li ( , 2011 describes CDD's emphasis on the outsourcing and incentivizing of routine performance monitoring in the villages as 'rendering society technical'. A proliferation of forms -thirty are included in the appendix of the 2008 Generasi operational manual -guides facilitators through the planning, budgeting, implementation, and monitoring phases of the annual project cycle (PNPM 2008, 92-128). These forms do not just assist with program management and auditing at the village and subdistrict levels: they are the source documents that are collated and filtered up into a national management information system (MIS) in Jakarta. They do the work of rendering society technical, precisely delineating, as Nikolas Rose writes, 'the domain to be governed as an intelligible field with specifiable limits and particular characteristics' (cited in Li 2011, 57).
The data reported in these forms attempt to make intangible social relations visible by documenting, for example, the percentage of poor women who attend participatory planning meetings, or the percentage of children under age five weighed at each monthly village clinic. This is what Rose (1999, 176) calls 'government through community', for these calculations can be 'mobilized, enrolled, deployed in novel programs and techniques which encourage and harness active practices of self-management and identity construction'. But even as villagers take charge of their development within such a narrow and specific set of parameters, these data expose potential deficits in a community's social capital, thus simultaneously justifying external intervention -by nearby program facilitators and distant technocrats and academics -to manage social relations in such a way that maximizes a community's capacity to develop (Li 2011, 63-64). 6 Generasi measures a total of twelve indicators. As noted above, Generasi also supports improved educational outcomes, measuring indicators such as total enrollment and attendance in primary and middle school, and services for children with special needs. Generasi funds may be used for educational expenses such as school uniforms, contract teachers, and transportation to distant schools. This article focuses on Generasi's support for maternal and child health, which accounts for eight of its twelve outcome indicators and most of its expenditures.
Topography and the contours of a community-driven therapeutic landscape Katz (2001Katz ( , 1214 notes that the Oxford English Dictionary's 1971 definition of topography signifies both the thing itself and the description of it, and is thus at once material and metaphorical. The OED definition includes two parts: 'the accurate and detailed delineation and description of any locality' and 'the features of a region or locality collectively'. The detailed work of describing locality, of 'doing topography', Katz (2001Katz ( , 1215 adds, is: purposive, partial, and, of course, interested, and it was usually conducted for political leaders or military commanders. . . . Topographical data are routinely fed to any number of global databases using much touted geographical information systems (GIS) that facilitate resource extraction, surveillance, and rule and, if necessary, attack across geographic scale by various social actors.
In this mode, 'doing topography' is a classic Cartesian 'strategy', as defined by Michel de Certeau (1984), a cartographic snare that secures people and places on a map. 7 Generasi's thirty forms are the starting point for drawing the contour lines of a partial, interested topography of maternal and child health care across Indonesia. Katz (2001Katz ( , 1229 argues that contour lines are the most important organizing metaphor on a topographical map, as they are 'lines of constant elevation that connect places at precisely the same altitude to reveal a terrain's three-dimensional shape'. As a critical 'geographic metaphor for aiding in the understanding of how the healing process works itself out in [and across] places' (Gesler 1992, 743), the contour line maintains the distinctness of place while connecting it analytically to other places (Katz 2001(Katz , 1229. The circulation, collation, and aggregation of data in the MIS generate and draw out Generasi's contour lines to national scale. These are the biopolitical tools with which a national program asserts its knowledge and mastery over distant localities across the archipelago. But before these numbers can travel and scale up, Generasi stakeholders on the ground -thousands of facilitators, village volunteers, and affiliated actors such as midwives -must produce a local topography in the villages. They first apprehend the unique features of a local therapeutic landscape, then translate those features into portable numeric descriptions whose contours enable comparisons. 7 De Certeau (1984,(35)(36) defines strategy as 'the calculation (or manipulation) of power relationships that becomes possible as soon as a subject with will and power (a business, an army, a city, a scientific institution) can be isolated. It postulates a place that can be delimited as its own and serve as the base from which relations with an exteriority composed of targets or threats (customers or competitors, enemies, the country surrounding the city, objectives and objects of research, etc.) can be managed'.

Manggarai village topographies
After a few days in the rural mountains of Manggarai Timur, a careful observer will notice distinguishing features of the landscape that explain differential access to health care. These differences begin with unmistakably material features of the Manggarai topography, the hills mostly, but it is important to understand that these are not strictly natural barriers to care. During my visits to Bonde subdistrict, I spent time in three of its twenty-four politically designated village units. Many of these villages are actually aggregations of hamlets of considerable distance from one another, so while the term 'village' suggests a single community to outsiders, what we observed were disparate communities with unequal access to services.   One morning in May 2013, my two research assistants and I observed a posyandu in Hamlet B, held inside the large front room of a volunteer's house (figure 6). A silent elderly man was the first to arrive; he sat in the middle of the room, but everyone ignored him. The hamlet headman explained to me that the old man was a victim of the Soeharto military dictatorship . He was beaten on the head in the 1970s for opposing Soeharto's ruling party during the campaign for one of his sham presidential elections. No one could remember the details of this man's protest, only that he had been silent ever since. Meanwhile, the village midwife who we interviewed the day before was not present, though she testified that she attends every posyandu. Had she been there, her job would have been to administer immunizations, consult with parents, and supervise the five volunteers who run the posyandu. The owner of the house has hosted posyandu for the past eleven years. He was filling in the books, writing down the weight of each child, an important origin point for drawing our metaphorical contour lines across Generasi communities. My data-oriented task was to examine his bookkeeping skills, so I went over and sat next to him, and I quickly discovered that Hamlet B often goes months at a time without holding these clinics. I tried to ask him questions about the frequency of posyandu and how they manage children at risk for undernutrition, but he wanted to talk about other things. I could barely hear his voice over the rambunctious noise of so many infants, toddlers, and parents in the room as he quietly told me, so that no one else would hear, how tired he was and how he no longer wanted his home to be the host for these clinics. His back ached, and expensive administrative barriers prevented him from signing up for the new national health insurance scheme. Neither the Generasi facilitators nor the public clinic staff in Bonde had given the volunteers their honoraria in months. Li (2007, 12) writes about the distinction between practices of government, 'in which a concept of improvement becomes technical as it is attached to calculated programs', and practices of politics, which might be any 'expression, in word or deed, of a critical challenge'. Such a challenge might begin with 'refusal of the way things are' and 'open up a front of struggle. . . . Government, from this perspective, is a response to the practice of politics that shapes, challenges, and provokes it' (ibid.). This gentleman registered a plaintive but inchoate challenge to the neglected and taken-for-granted health service he has hosted in Hamlet B for more than a decade. His critique was barely audible, much less legible, to the governing agents who might address (or even ignore) his incipient political critique. I was caught off guard by his challenge, and hardly even recognized it as a political act until after the event, because it did not match what I ostensibly came to Dantena to study: the technical aspects of community-driven maternal and child health care (inspecting the books, observing the posyandu activities, monitoring maternal and child nutrition, etc.).
After the parents and children left, I tried to convene a kind of focus group discussion with all five volunteers together, but the formal village leaders and Generasi facilitators (all from Hamlet A) took their responsibilities as hosts to us researchers from Jakarta seriously. They hovered around us, effectively shutting down opportunities for honest discussion. Separately, my research assistants found out that Hamlet B volunteers must retrieve food supplements for at-risk households themselves from Hamlet A. The Generasi facilitators buy the supplements, but they do not deliver them to Hamlet B.
At some point in this region's history, a decision was made to build the road through Dantena's Hamlet A and locate all front-line services there. The prioritization of Hamlet A is seen even in markers of faith, as the state-sanctioned chapel is in Hamlet A, while the ritual drum house representing Manggarai's ancestral traditions is in Hamlet B. In short, Hamlet A is marked for progressive development interventions; Hamlet B looks left behind by design. 8 These meaningful distinctions required several days of fieldwork to discern, if not fully explain, but the Generasi program largely sees Hamlet A and Hamlet B as one Dantena village, and the data reflecting some of these internal inequalities does not register on the radar of Generasi program officials at the subdistrict level or higher.
After my attempted interview with the volunteers, our hosts from Hamlet A arranged a special lunch for us. We learned from several informants, and observed directly in many households, that the dietary mainstay in these Manggarai villages is called makan segitiga, a 'triangle diet' whose three components are water (air), vegetables (sayur), and salt (garam). Either squash or cassava leaves are boiled in salted water, and then poured over an enormous plate of rice. This is what our hosts served us, but what made the lunch special was the addition of instant Indomie-brand noodles as an additional side dish. To be sure, these makan segitiga meals are not served in households living in extreme poverty. Most farming families own their land; they are not sharecroppers. But their gardens on these fertile volcanic slopes are cultivated for cash crops, mostly clove and coffee. Most households keep chickens and many also raise pigs and goats, but these are not for personal consumption or even for financial investment. Rather, they are reserved for enormous social obligations known as sida, a contribution to large family events such as weddings and funerals. We heard a lot about sida in Manggarai. Allerton (2013, 69) defines sida as 'requests for money'. Many of our respondents described sida as an onerous financial burden that perpetuates cycles of debt and prevents most households from overcoming poverty. Despite owning their own livestock, most households in Bonde have no routine source of protein in their daily diet. Young mothers in Dantena told us: 'Our livestock is for income, not to eat', and 'Even when we have chickens or pigs, it wouldn't feel appropriate [kurang pantas] to eat it ourselves given that we still need money'. Most Indonesian household surveys, much less Generasi's forms, cannot render these social relations technical to partially explain the perpetuation of local inequalities.
In 2015 we visited Dantena's neighboring village of Semang. On our first day there, we visited the village secretary's house, accompanied by our Generasi hosts from the subdistrict and district levels. We conducted an interview with one of the Generasi facilitators, who happened to be the village secretary's wife. She did not know the answers to some of our basic questions about how the Generasi program works in Semang. Her husband and the village's Generasi activity manager jumped in repeatedly to answer for her. It became apparent that in Semang there is no division of the technical tasks that are typically separately assigned to village facilitators, activity managers, and formal village leadership; they simply work on and decide everything together. From one perspective, this is a practical approach, especially for the purpose of completing all the forms and meeting other administrative requirements, but from a technical Generasi perspective, these practices reek of collusion and exclusion, leading to elite capture of community development resources. This example also illustrates a profound gender disparity in Generasi's social topography in the Manggarai highlands, and, based on similar interviews with female village facilitators I conducted in 2013, I suspect these gender disparities are the norm. The Bonde subdistrict facilitator confirmed this assessment when he lamented to us how difficult it was to secure women's participation. Without prompting, he mentioned that in the discussion meetings that Generasi had facilitated in the hamlets, once a man or village elite spoke, ordinary women were reluctant to speak up, especially if they had a different opinion.
Manggarai social arrangements in some ways reflect the landscape, with steep hierarchies of gender, class, education, and other status markers such as physical or mental disabilities precluding the kinds of participation scenarios that Generasi expects would transform village governance. This includes some particularly stigmatizing discourses deployed against certain neighbors, such as households disparagingly described as 'lazy' and others as 'newcomers', and, in one case of maternal mortality, a woman's death was brushed aside due to her mental illness. These discourses reveal local hierarchies of 'deservingness' that affect access to health and other resources (Willen 2012).
Here is where my findings superficially appear to differ with Allerton's, who describes a relatively 'undifferentiated' Manggarai community in a one-clan village with descendants from a common ancestor, or at least undifferentiated relative to other eastern Indonesian societies with clear groups of 'nobles, commoners, and the descendants of slaves' (Allerton 2012, 182). I say superficial because Allerton conducts her study of the Manggarai landscape mostly in another register, reporting on the symbols, rituals, histories, and memories that tie Manggarai communities to the land and one another. In her account, the state intervenes as a stiff governing agent in resettlement and heritage preservation schemes, but resolutely does not 'get' the complex Manggarai relationships to the land. Indeed, from my perspective, the Generasi agents I have met are foreclosed from these phenomenological aspects of Manggarai social life, at least when they are doing their Generasi jobs, as they are caught up in another kind of phenomenology of state governance preoccupied with technical data, process, and forms.
A powerful example of this disjuncture is in childbirth practices. Health officials, Generasi actors, and donors all agree on the policy prescription: all women must deliver in accredited health facilities, with biomedically skilled, state-certified midwives delivering babies. This seemingly universal consensus reflects the latest global health trend 'toward hospital deliveries for safe motherhood that supplanted decades of efforts to train and use skilled traditional birth attendants' (Adams, Burke, and Whitmarsh 2014, 181). Vincanne Adams and colleagues argue that 'shifts like this are fueled by a research industry that demands globally comparable metrics', something easy to count like number of hospital deliveries (ibid.). Critics argue that global comparisons based on simple metrics like this are drained of all meaning and value, especially when some accredited facilities include the aforementioned hive-invested delivery room at the clinic in Semang. It should come as no surprise that around half of mothers in Semang still deliver at home, with the support of a traditional birth attendant, which frustrates the clinic staff to no end, as they work under pressure to increase deliveries in their facilities under donor-funded schemes with militaristic names like the 'Maternal and Child Health Revolution'. 9 The reasons for using the services of a traditional birth attendant over a biomedically skilled midwife are familiar, persistent, and long documented in both the global and regional public health literature, and they include topographical barriers to access (Anderson 2014;Hildebrand 2012;Kruske and Barclay 2004;Metherall 2015;Niehof 2014;Østergaard 2015;Stein 2007;Thaddeus and Maine 1994). Semang's village midwife does not actually live in the village, not least because the aforementioned clinic, which includes living quarters for the staff, is unfit for residence. She lives more than an hour away (by motorbike) in the neighboring district capital of Ruteng. By the time families call the midwife, it is usually too late. A more decent clinic with staff on site is too far, across difficult terrain, and often too expensive to reach. A traditional birth attendant is more familiar and nearby; provides comprehensive pre-, peri-, and post-natal support for little or no charge; and brings comforting traditional medicine. Allerton's (2013) helpful account of the ritual and mundane practices associated with childbirth at home goes much further than these material concerns. The birthing room is 9 The Australian-Indonesian Partnership for Maternal and Neonatal Health (AIPMNH) program describes NTT province's 'Maternal and Child Health Revolution' policy in English here: http://aipmnh.org/web_en/index.php?option=com_content&view=article&id=162&Itemid=224. A video presentation of the program in Indonesia is available here: https://www.youtube.com/watch?v=68qlL8Et37g. likened to a transitional womb, a 'room-womb'. Across four pages of a thickly described single childbirth event, we learn how deliveries include many assisting members of the community (Allerton 2013, 24-27). Various rituals insist upon mother and infant staying in the room-womb, maintaining close proximity to the placenta and one's ancestors to ensure the formation of beneficial social relationships. Allerton's description reveals a profound 'field of care', which Gesler (1992, 738) earlier defined as 'networks of interpersonal concern'. Biomedical interventions seek to territorially transfer, wholesale, and by way of military metaphor (the aforementioned 'Maternal and Child Health Revolution'), birthing and other health-related practices from the intimate spaces within Manggarai houses to clinics. What I find interesting, but not surprising given our team's lack of deeper familiarization with Manggarai communities, is that none of these compelling details were revealed to us when Generasi actors, health workers, and hired consultants poked around and asked questions. At most, and if we were lucky, we were introduced to one or two traditional birth attendants and had an opportunity to interview them, but this barely begins to scratch the surface.
Conjuring scale in Bonde subdistrict Anna Tsing (2005, 58) writes about 'scale making projects', defining scale as the 'spatial dimensionality necessary for a particular kind of view, whether up close or from a distance, microscopic or planetary', and reminds us that scales are hardly neutral frames for viewing the world: they are 'proposed, practiced, and evaded, as well as taken for granted'. In short, scales must be conjured into being, and are manifested in what I call 'scalar imaginaries'. As described above, CDD programs like Generasi invest the political scale of the village with national-scale standardized tools for harnessing social capital and other resources to achieve Generasi's vision of community development.
In Bonde subdistrict, Generasi has six staff and additional support from various officials in the subdistrict government. About a five-minute walk behind and above the Generasi office is subdistrict clinic, where a coordinator manages a large group of midwives in various stages of professional development. Many are assigned to particular villages as midwives, though few of them actually live in the villages full time, and they set the schedules for each hamlet's posyandu. How do these low-level officials within the national hierarchy effect their role as governing agents over the constituent villages in their subdistricts? By my reckoning, they generate and validate their authority through practices of material documentation. At both the clinic and the Generasi office, data reporting, entry, and collation from the villages begins in earnest. These data are the raw materials Generasi and the government use for recasting villages as drivers of their own development and then drawing the contour lines that secure and connect villages into discretely but only partially known objects. These village-level data are aggregated into their respective subdistricts, then to district level, and finally up to the national level, becoming useful for the justification of larger-scale political projects.
These agents of local health governance also deploy spatial metaphors, in speech and occasionally in practice, to project their authority. When we interviewed the village midwife for Dantena, she described a monthly three-day ritual of 'sweeping', using the English word as it has been adopted in the Indonesian language. In Indonesian, sweeping is a security procedure, often equated with raids, typically conducted by police or vice squads on the hunt for illicit vendors of sex, drugs, and alcohol. It involves door-to-door inspections until a specified area is secured. Sweeping has a negative connotation in the Indonesian political imaginary. 10 In the context of posyandu, 'sweeping' means going to the home of every mother with a child under five years of age; on the first day, mothers are reminded that the posyandu will be held the next day. On the second day, the posyandu is held. On the third day, a followup sweep is conducted with every mother who did not attend. Sweeping has spatial connotations; it projects total coverage, of an entire hamlet, for example, leaving not one household untouched. The door-to-door aspect suggests a systematic approach, and its borrowing from the security sector has an intimidating quality. The moral valence of sweeping adds stigma to absenteeism from posyandu.
Dantena's midwife confidently told us that she not only attends each hamlet's monthly posyandu but also conducts each sweep herself, along with the village volunteers. As reported already, when we attended the posyandu in Hamlet B the next day, not only was she absent but we learned that the posyandu is only semiactive. The volunteers informed us that the midwife only attends posyandu when it is time for immunizations or Vitamin A supplements, roughly twice a year. If sweeping occurs, the village volunteers do it themselves. In neighboring Semang, a posyandu volunteer told us that their village midwife sends a text message every month to remind volunteers to sweep the hamlet the day before posyandu. Hardly a utilitarian and systematic operation, the volunteer complained that sweeping is rather time consuming because every visit is a social occasion, an opportunity for each house she visits to serve coffee and gossip. Her description of sweeping unwittingly and momentarily unsettled the fantasy of spatial and martial mastery that the term ordinarily conjures. 10 In her book The Wisdom of Whores, Elizabeth Pisani (2008, 90-91) provides an evocative description of how 'sweeping' operations during election season impede HIV surveillance among vulnerable communities in Jakarta.
Since 2013, I have been puzzled by the falsehoods Dantena's midwife so confidently told us during our interview with her. She also claimed that all the women in Dantena deliver their babies at clinics with midwives, and that all deliveries have been safe during her tenure, even though we knew that just a few weeks prior to our arrival, a woman with mental illness died while delivering her eighth child. When we asked her about this recent case, she expressed surprise that we knew about it, and asked us if the clinic coordinator had told us (she had, as had Generasi staff and members of the Dantena community). Her embarrassed response was telling: 'I purposely did not tell you about it at first because her death could undermine program targets'. She added that the mother's death would not be counted in their clinic reports because of her mental illness; in fact, every version of the story we heard about this woman's death featured this 'yes, but she had a mental illness' caveat, revealing shared assumptions about 'deservingness'. Under pressure to meet targets in NTT's Maternal and Child Health Revolution program, this mother's exceptional status made it easier for the clinic staff to discount her death in their formal accountability procedures. While we might make sense of this story in terms of the pressures and moral hazards under which clinic staff operate, I was still puzzled by the midwife's surprise, her confidence utterly shaken. How could she have possibly thought we would not find out about the woman's death, or that she does not usually attend the posyandu she supervises, or that many women still deliver at home with traditional birth attendants?

Scaling up and down
My puzzlement led me to consider how people in this context understand changes in scale. The hierarchy of down-scale travel from Jakarta -from province, to district, to subdistrict, to village, to hamlet -is captured in the Indonesian phrase commonly used for going to the field, 'turun ke lapangan', which actually means to 'descend to the field'. This downward action is also a bit stigmatizing, as it is used in phrases like 'descend to the world of politics', 'descend to the streets', and 'descend to the nightlife world'. 11 The sense is that even if there is pleasure in it, one feels a bit dirty for doing it, or perhaps one should. Going down to the field is therefore regulated; it typically requires permission, coordination, and reporting at every level. One brings a letter, secures an introduction by phone, or is accompanied to the field. This allows for stage management and secures appearances. The Dantena midwife probably felt secure in her representation of the maternal and child health situation in her village because she did not know that we were going to the next day's posyandu in Hamlet B, much less that we would actually walk down a steep hill to attend it. 11 These are translated as 'turun ke dunia politik', 'turun ke jalan', and 'turun ke dunia gemerlap', respectively. As I recounted above, our visit to Hamlet B was arranged and chaperoned by Generasi, not the clinic staff, which explains not just our access but also the surveillance by a different set of actors. The clinic expects all donor activity to be channeled vertically through the health ministry, but CDD programs in theory operate directly with communities, from the bottom up. In actuality, they operate through a government channel parallel to the health ministry, and the alternative access that Generasi provides poses a potential threat to government service providers.
The practice of unannounced visits to the field is a new trend closely associated with the current administration. President Joko Widodo earned his populist credentials while serving as mayor of a Javanese city, and then governor of Jakarta, by making unannounced spotchecks called 'blusukan' (from Javanese) designed to embarrass nonperforming government officials and hold them accountable (Tapsell 2015, 41-43;Hatherell 2014, 445-47). We listened to mid-level bureaucrats complain about this trend because it violates longstanding protocols of a system that benefits them. Our first visit to Semang was a textbook example of a well-regulated and authorized turun ke lapangan, a descent to the field, accompanied by a group of Generasi chaperones from subdistrict and district levels. On our second day in Semang, though it was not a blusukan spot-check, we made a much less formal follow-up visit to the village secretary so that we could speak with him privately, without Generasi's higher level stage managers all around. Even though we asked him the day before if we could visit a second time, and he gave us permission, he was nonetheless a bit rattled by our next call because he never expected we would actually come back. His revealing first question was whether we brought a letter that authorized our research.
Martial metaphors and circulating forms project mastery over scales across a therapeutic landscape. Documents such as posyandu registers and Generasi forms distill but also distort local topographies into the contour lines of national development policies and programs. In other words, these documents enable scaling up, connecting Dantena, Semang, and other Manggarai communities analytically with Generasi villages across Indonesia, and allowing comparisons among them. For example, from Jakarta I selected Manggarai Timur for inclusion in our 2013 study based on its promising topography of indicators relative to other NTT districts in the Generasi MIS.
Another set of documents like travel authorization letters and text messages facilitate regulated descents into the field; these are the procedures for traversing down scale that preserve a territorial fantasy, the invested contour lines drawn by middle-and lower-level stakeholders. This includes the fantasy of 'government through community', as facilitators, volunteers, and other 'participants' in Indonesia's hamlets and villages enlist in narrowly circumscribed forms of governance and do the work of making therapeutic (and other) landscapes legible to the governing agents who recruit them (Rose 1999;Rushton 2014).
All of these forms make some aspects of the local topography visible while concealing others. Concealment may be based on false pretenses, but more often it is because the forms capture only thin slices, a few contours, of data that when emphasized for describing program outcomes cannot yield anything but distortions. These distortions create 'truths' that require strategic maintenance and accountability through the management of appearances on carefully stagecrafted site visits. 12 Topography and ethnography Katz (2001Katz ( , 1216 proposes using the strategic tools that generate contour lines in service of a critical counter-topography that would 'encourage and enable the formation of new political-economic alliances that transcend both place and identity and foster a more effective cultural politics to counter the imperial, patriarchal, and racist integument of globalization'. Contour lines may enable comparisons while also keeping an anchor tethered to the distinctiveness of place, but I hesitate to rely exclusively on this method because it risks introducing the aforementioned distortions into the analysis, a fetishization of the comparative advantages offered by contour lines that travel and scale easily, in service of external stakeholder agendas (progressive or otherwise), but at the expense of a more sustained engagement with local contexts. In the case of Generasi and other wide-reaching programs, the forms train technocrats and facilitators to see maternal and child health in particular ways that utterly preclude looking for, much less acknowledging, local fields of care as culturally specific as room-wombs.
Anthropologist Kirsten Hastrup (2005, 145) has proposed a 'topographic turn' that does not 'return to cartography and mapmaking' nor posit 'the map as a direct representation of the world'. Rather, the topographic turn is distinguished by taking seriously both the movements of social agents, and the paths they carve out, physically and socially, through their wayfinding. The concreteness and materiality of topography thus defies the abstract map (the territory as represented), and is closely linked to experience and practical mastery of the environment (ibid.).
Just as the physical terrain is the first order of description when Generasi actors describe the challenging topografi of program implementation in Flores before moving on to a description 12 Generasi's endless circulation of forms recalls Anelise Riles's (2000, 3) definition of a network. Forms are productive artifacts of Generasi's own therapeutic landscape, generating its own reality by repeatedly reflecting on itself, but this internal and reflexive focus does not see, and therefore excludes, larger fields of care. of the social terrain, I argue here that comparative contour lines are but a first-order and partial description when apprehending the interested representations of a particular therapeutic landscape. At most, the contour lines that governing agents draw provide one point of entry; they require ethnographic inquiry into how certain contours acquire their legibility over others. 13 Hastrup's topographic turn should feel familiar to ethnographers, 14 simultaneously acknowledging abstract maps and defying them, beginning with the concreteness and materiality of the landscape, and then linking it with less strategic but more tactical way-finding, lived experiences, and practical mastery across it. 15 For the short-term ethnographer studying a health intervention in Indonesia, entry into a stage-managed therapeutic landscape requires a more tactical style of fieldwork -something akin to the spirit of blusukan (if not the actual media-savvy political campaign strategy it has become) over more formal and accompanied descents into the field -in order to access other contour lines, some more legible than others (and others still inaccessible, as Allerton's longer-term ethnographic insights suggest), that describe a larger, more contested field of maternal and child health care. At the very least, this requires more than just an ethnographic sensibility attuned to structural dynamics of power and hierarchy, and to the constraints those dynamics impose upon each informant and the organization of the research itself. It also demands more time than what typical program evaluation budgets can afford, as it takes at least a few days to move through accompanied descents into the field at the district, subdistrict, village, and hamlet levels, before one can engage more comfortably and honestly with program stakeholders. Through more tactical acts of fieldwork, it is possible to learn, for example, what sweeping operations really look like, or that 50 percent of childbirths in Manggarai villages do not take place in health facilities, or that the midwives in the clinic keep a different set of posyandu books than the volunteers in the village. These revelations generally escape the strategically layered topography of the Generasi program as it builds scale through the district, provincial, national, and global levels. Through its international donors, Generasi commissions world-class randomized control studies to measure the 13 For example, while it was not part of this research, I believe the high prevalence of martial metaphors, such as 'sweeping', that came up during this fieldwork, speaks to provincial and national histories of political violence. 14 Citing Hastrup, Adia Benton (2015, 28) describes the first chapter of her ethnography HIV Exceptionalism: Development through Disease in Sierra Leone as a topography of the physical and social terrain of HIV/AIDS programs in Freetown. Accounting for donor-driven agendas and the distortions they produce on the ground are also an important part of this topography. 15 In contrast to his definition of strategy, de Certeau (1984, 37) defines a tactic as 'a calculated action determined by the absence of a proper locus. No delimitation of an exteriority, then, provides it with the condition necessary for autonomy. The space of a tactic is the space of the other'. impact of the community-driven development paradigm. 16 These studies depend on Generasi's documented contours in the MIS and other centralized databases in Jakarta for their sampling frame and definition of variables. This is a topography with accumulating layers, adding to the purposive, partial, and interested archeology of past therapeutic interventions on the Manggarai landscape and across the nation.