Re-thinking children's agency in extreme hardship: Zimbabwean children's draw-and-write about their HIV-affected peers

We compare two analyses of the same ‘ draw-and-write ’ exercises in which 128 Zimbabwean children represented their HIV-affected peers. The ﬁ rst, informed by the ‘ New Social Studies of Childhood ’ , easily identi ﬁ ed examples of independent re ﬂ ection and action by children. The second, informed by Sen's understandings of agency, drew attention to the negative consequences of many of the choices available to children, and the contextual limits on outcomes children themselves would value: the support of caring adults, adequate food, and opportunities to advance their health and safety. Conceptualisations of agency need to take greater account of children's own accounts of outcomes they value, rather than identifying agency in any form of independent re ﬂ ection and action per se.


Introduction
We draw on a study of Zimbabwean school children's accounts of the lives of their HIV-affected peers to call for an extension of current understandings of children's agency. We argue that the 'New Social Studies of Childhood' (NSSC) view of children's agency that dominates research on the health and well-being of children in extreme settingsfails to take enough account of children's own experiences of their lives, unwittingly masking the degree of children's suffering at a time when welfare and development aid for children are being reduced in many settings. We develop this argument at two interlinked levels. At the theoretical level we argue that the NSSC view of agency is often applied in a decontextualising and top-down manner. It is de-contextualising through paying inadequate attention to structural constraints on children's agency and on the negative outcomes of many of the allegedly agentic choices available to children. It is top-down insofar as it fails to pay enough attention to children's own accounts of their lives and the futures they would like to see, foregrounding researchers' concern to showcase evidence for independent action per se, irrespective of its outcome. Methodologically we support our argument through comparing two different analyses of the same set of children's draw-and-write exercises which generated different findings. The first analysis was informed by a NSSC framework and used thematic content analysis to analyse the draw-and-writes. The second analysis was guided by Sen's emphasis on the need to understand peoples' experiences in the light of their own accounts of the lives they hope for, and used a more holistic narrative approach to analysis.
The NSSC was spearheaded by James and Prout (1990) in the early 1990s. They argued against politically problematic and factually inaccurate stereotypes of children as passive victims of impossible social circumstances, in social worlds dominated by adults. This work inspired two decades of research into children's agency (e.g. Evers et al., 2011;Hutchby and Moran-Ellis, 1998;Qvortrup, 2005;Christensen and James, 2000), advancing for example understandings of the life situations of street children (Panter-Brick, 2002;Clauser, 1997), working children (Porter, 1996;Nieuwenhuys, 2005;Abebe and Kjørholt, 2009), young carers (Wihstutz, 2011;Skovdal et al., 2009) and children living with disabilities (Davis and Watson, 2002;MacArthur et al., 2007). It is now generally accepted that even small children are able to think and act independently of adults, often in ways that ensure their survival in extremely challenging settings.
Recognising children's ability to exercise agency is regarded an antidote to various problems associated with the concepts and practices of particular child-focused academics and development agencies. James and Prout (1990) sought to pull together the work of a scattered but growing group of critical intellectuals increasingly uncomfortable with views of children as passive and dominated by adults that prevailed in the 1970s and 80s. More recently, Bordonaro and Payne (2012) have identified an emphasis on children's agency as an important counter to a worrying tendency to construct children as 'social problems'. For Skovdal and Daniel (2012), robust understandings of children's ability for independent reflection and action in challenging social environments is key to developing strengths-based development policies and interventions that respect and build on coping strategies developed by children themselves. They contrast these with top-down approaches that impose strategies framed by external health and welfare professionals, that may have a poor fit with children's specific circumstances and existing coping skills (Skovdal, 2012). For Kesby et al. (2006), an emphasis on the agency and resourcefulness of children is part and parcel of a respectful recognition of the value of 'other childhoods'. They argue that such recognition is necessary in the face of a tendency to regard privileged western notions of childhood as the norm against which all other childhoods are evaluated, and found wanting.
However there have also been critiques of the emphasis on children's agency. Discussing her research with migrant children in Lesotho and Malawi, Ansell (2014) argues that children's ability to exercise agency is heavily limited by the economic factors that threaten their survival, and intertwined cultural and relational practices that silence their voices in their families and in the national and international contexts that contextualise efforts to improve their health. Durham (2008: 151) makes a similar point in her study of anthropologists' research into young peoples' involvement in political protest in Botswana. She accuses researchers of 'romanticisation' in rushing to identify youth participation in social protest and religious groupings as evidence for agency, neglecting the extreme limitations on the nature and outcome of such participation.
In studies of children of the HIV/AIDS epidemic in Africa, there are also rumblings of unease that a dogged insistence on children's agency potentially masks constraints on the outcomes of choices available to children in conditions of poverty, violence and abuse, and may play a role in 'normalising' the extent of their suffering. Andersen (2012) points out that the types of agency available to Kenyan HIV-affected children in the short term may have severely negative consequences in the longer term. She argues that agency may be a blunt analytical tool to describe a 12-year-old girl's choice to have unprotected sex with an HIV-positive older man, to generate income to feed her younger siblings, or a child's decision to drop out of school to care for a dying parent.
Who decides what constitutes agency? Seckinelgin (2012) argues against the trend for assessing poor people's life situations and possibilities in terms of externally derived criteria formulated by academicians, health professionals or international development agencies. Influenced by Sen's arguments in Development as Freedom (1999), he argues that situations should be assessed in terms of the extent to which people's actions open up opportunities for them to lead the lives that they themselves would want to lead. We argue against a tendency to view agency as a child's ability to engage in any form of independent action per se. More attention needs to be paid to the extent to which children have access to the resources they would need to act in ways that take them closer to the lives they themselves would want to lead.
Using the NSCC notion of agency, one might indeed say that children are exercising agency in choosing to drop out of school to care for dying parents who have no other source of care. However, we will argue that such a conceptualisation of agencythat makes no reference to children's own aspirations (to complete their education in this case)is a hollow and academic exercise, drawing on a problematic neoliberal version of the subject which needs to be challenged and re-thought (Ruddick, 2007a(Ruddick, , 2007b. In discussing the lives of rural Zimbabwean children with HIV/ AIDS, Parsons (2013) argues that the current theoretical toolkit of psychology and anthropology lacks the tools to adequately capture the depth and extent of AIDS-affected children's suffering. We argue that a critical rethinking of the notion of children's agency is part and parcel of tackling this challenge.

What is agency?
Agency is a core concept across a range of social sciences. This paper is framed by symbolic interactionist accounts of agency, rooted in social psychology and sociology. Social psychologists tend to focus on actions that result from choices between different options. Gillespie (2010) defines agency as the exercise of choice, where a socially located person acts independently of an immediate situation, weighing up and choosing between alternative responses to social demands, in the light of goals that may be motivated by concerns outside of the immediate situation. Sociologists (e.g. Giddens, 1979;Mead, 1934) speak of agency in the context of the structure-agency relationship. They recognise the constraints of social relations (structure) on individual action, and their moulding influence on the individual, whilst also acknowledging that individuals are often able to resist or reshape the social contexts in which they find themselves (agency). They emphasise that agency and structure should not be seen in binary terms, but as different aspects of the single process through which societies and individuals are co-constructed. It is this symbolic interactionist perspective that informs the research below. Our starting point is that opportunities to exercise agency are heavily constrained by social contexts (Williams, 2004), including the resources available to children in coping with life challenges in extreme situations (Kesby et al., 2006), the nature of the social relationships in which an actor is embedded (Cleaver, 2007), and the limits or opportunities presented by a person's physical health and strength (Nguyen, 2005), particularly relevant in our context of HIV/AIDS, poverty and food insecurity. Sen (1999) and Nussbaum (2000) emphasise an additional element in their definition of agency. They define an agent not only as someone who acts and brings about change, but also as someone whose achievement can be evaluated in terms of his or her own values and objectives, and their own understanding of what would constitute a good life. We will argue that attention to children's own values and objectives is currently lacking from the way in which much research in the NSSC tradition conceptualises agency. This is arguably ironic given that the NSSC repeatedly emphasise the importance of including children's voices in research. Whilst this intention tends to be honoured through research methods that elicit children's own views, it is not always reflected in interpretative frameworks that foreground these views in making sense of the resulting data.

Our study
Against the background of our view of agency as an inherently social and relational phenomenon, and our interest in foregrounding children's own perspectives on their lives and options, we explore rural Zimbabwean children's own understandings of the agency of their AIDS-affected peers, through an analysis of draw-and-write exercises. We view draw-and-write as an appropriate method for exploring the socially negotiated and shared symbolic frameworks or social representations that this group of children used to interpret their experience and guide their actions. We use the resulting findings to support our argument that whilst children repeatedly depicted their AIDS-affected peers as engaged in independent and resourceful action, these actions were often so limited by situational constraints that it makes little sense to regard these as agentic choicesin Sen's sense of the term. Whilst these actions often constitute evidence of heroic courage and immense resourcefulness, sometimes under conditions of almost unimaginable hardship, they cannot be seen to constitute 'agency' in Sen's sense of the word. In the difficult life contexts of the AIDS epidemic in rural Zimbabwe, children's actions often achieve little more than their very basic survival, with outcomes often falling very short of the lives that they themselves would like to lead.
Our interest in draw-and-write is framed by social representations theory (SRT) (cf. Moscovici, 1984) a social psychological perspective that explores the symbolic frameworks that communities or groups of people jointly construct to make sense of their social worlds. Moscovici (1973: xviii) defines social representations as "systems of values, ideas and practices with a twofold function; first, to establish an order which will enable individuals to orientate themselves in their material and social world and to master it; and secondly to enable communication to take place among the members of a community by providing them with a code for social exchange and for naming and classifying the various aspects of their world and their individual and group history." The resulting common-sense knowledge is deeply shaped by local context. Social representations are constantly negotiated and renegotiated in the course of daily interactions between people, groups and institutions in particular historical settingsthrough dialogue and communication as people go about their daily lives. Understanding children's social representations is particularly important for those seeking to develop child-centred interventions and policies because representations serve as a key mediator of how children respond to life challenges, and how they engage with external efforts to support them.

Context of research
The draw-and-write study we report on here is one strand of a larger multi-method study of HIV/AIDS-affected children in rural Zimbabwe, which received ethical approval from the Medical Research Council of Zimbabwe (MRCZ) and the London School of Economics Research Ethics Committee. In rural Zimbabwe, one in six adults (15%) and one in 40 children under 14 (2.5%) are HIV positive, and 20% of children have lost at least one parent to HIV/AIDS. At the time of our study (August 2012 to May 2013) Zimbabwean people's access to anti-retroviral drug therapy (ART) was 31.5% (28,149) for children and 59% (298,092) for adults (National AIDS Council Zimbabwe, 2011) and on-going treatment availability was not assured. HIV constituted a death sentence for many, with the epidemic still shrouded in high levels of stigma and denial.
In his study of children with AIDS in rural Zimbabwe, Parsons (2013) argues against regarding ART availability as inevitably supporting an improvement in the children's lives. In his words, for many HIV-positive children, "the horror of life (with AIDS) threatens to render all experience unspeakable" (2013: 2560). Parsons documents children who eventually chose to discontinue potentially life-saving treatment, worn down by a range of factors. These included pressures by spiritualist churches that regarded acceptance of ART treatment as a failure of faith, the cumulative suffering arising from factors such as grief at the loss of parents, the daily experience of social annihilation in the face of stigma and the immensity of the daily grind of coping with malnutrition in contexts of extreme poverty and precarious livelihoods (particularly given that effective ART depends on eating properly) and patchy or inexistent social support.
In rural Zimbabwe, most people live in extreme poverty with minimal access to health or welfare support from public sector or local or global NGOs, in the wider climate of national political and economic instability, and fluctuating global commitments to development aid Campbell et al. (2012). Schools have been particularly disrupted in recent years. Thus for example around 2008 due to hyperinflation, teacher's salaries were extremely low, with many schools closing altogether. Since then, with new currency developments and a concerted programme of assistance from international donors, the situation has improved, but remains unpredictable. Thus for example, in recent years, UNICEF provided extensive funding for the BEAM programme, providing school fees for many impoverished AIDS affected children, who would not otherwise have been able to attend school, though the future of this programme is uncertain.

Research methods
Our draw-and-write exercises were drawn from a larger study that included interviews and focus groups with teachers, community leaders, lay community members and children, ethnographic observation of children in school settings, and Photovoice with high school children (Campbell et al, 2014). Draw-and-writes were collected by two Shona-speaking fieldworkers, both qualified social workers, a man and a woman, in classroom settings in a rural primary school. Participants included 128 children (58 boys, 70 girls), aged 10-14. The exercise took 90 min. Teachers were not present. Half were asked to write about an HIV-affected boy, half about an HIV-affected girl, with the task worded as follows: How does HIV affect girls/boys at your school? Write a story of a girl/boy who is affected by HIV. This child can themselves be HIV-infected, or have relatives suffering from AIDS, or be an AIDS orphan. What challenges does s/he face? How does the school help him/her to overcome difficulties in his/her everyday life? Draw a picture to accompany your story.
Children were not asked to disclose whether they were themselves HIV affected. However given that one in five children in our study site had lost at least one parent to AIDS, that one in six adults were HIV positive, and that 40% of adults with full-blown AIDS were not accessing drug treatment it is likely that all children in the class would have had first-hand contact with the impact of the epidemic either on their own families or the families of peers or neighbours. As will become clear from the data below, some stories suggest that particular children may be referring to their own personal experiences; others do not. However these are informal observations. Our theory (SRT) and research design do not equip us to make judgements of this nature. Rather than seeking to make claims about the everyday realities of particular children's lives, our data throw light on the collectively negotiated symbolic fields available to children for making sense of their daily challenges and their social relationships in particular contexts.
Drawings and essays were analysed as a single unit. We provide some illustrative examples of these immediately below. In the 128 drawings and essays, 63 of the children's accounts referred children who were AIDS orphans: Surely when problems start they keep coming one after the other. There is a boy who is facing serious challenges. This boy is an orphan his parents died of AIDS. When he goes to school he goes without eating any food and he wears rags without shoes. At school he doesn't have any books or pens and he doesn't pay school fees, the donors pay for him. Some children want to play with him but some children stigmatise him thinking he has HIV. At home he is abused by the grandmother. He goes to herd cattle without having eaten any food and when he comes back it will be already night and he sleeps without eating. He sleeps on the floor with only a mat and nothing to throw over him. (See Fig. 1, entitled: 'Draw-and-write of an AIDS Orphan', for associated drawing).
Sixty-two of the draw-and-writes referred to children who were themselves suffering from AIDS: I am writing my composition talking about a girl who is sick with HIV/AIDS. Surely it is sad because this is a killer disease. There is a girl who is suffering from this disease. This girl faces a lot of challenges wherever she goes. She can't do household chores because she will be sick. At school she can't do other activities because she is sick and at times she doesn't go to school because she will be sick. The biggest problem is that she can't eat properly because she is sick. Surely a person needs to see this for themselves. (See Fig. 2, entitled: 'Draw-and-write of a child suffering from AIDS', for associated drawing).
Finally 35 of the draw-and-writes referred to children who were caring for sick parents: Truly there are sad cases in this world. I feel sorry for this girl called C. Her parents are suffering from HIV. Every weekend she goes to the river to wash blankets that are spoiled by her parents' faeces. At times she comes late to school as she first bathes and clothes her parents. The school helps her by giving her money for the grinding mill and food. (See Fig. 3, entitled: 'Draw-and-write about a child carer', for associated drawing.) These numbers do not add up to 128 because 32 draw-andwrites referred to children in more than one category.

The story of our data analysis
The argument of this paper evolved out of our comparison of two separate analyses we conducted of the same draw-and-write material. Each was informed by different methods and conceptual frameworks, and generated different findings. We briefly summarise the findings of our first analysis, which are reported elsewhere, before turning to report in more detail on the findings of our second analysis.
When we started to analyse these draw-and-writes we were heavily influenced by the NSSC's emphasis on children's agency and the associated academic literature on AIDS-affected children in sub-Saharan Africa, which emphasised the role played by children's agency in ensuring their coping and resilience (Skovdal and Daniel,  . We were also informed by the growing body of research literature documenting teachers' support of HIV-affected children, for example through patoral care (Bhana et al., 2006;Ogina, 2010), mentoring and counselling (Theron and Engelbrecht, 2012) the adaptation of school codes of conduct to foster a caring and safe learning environment for learners (Mampane and Bouwer, 2011), including for example, showing flexibility in allowing pregnant teenage girls to stay in school (Bhana et al., 2010) and so on. Relatedly, we drew on literature describing school-based interventions in support of HIV-affected children, such as teacher-led HIV prevention activities (Mathews et al, 2012), school-based sexual and reproductive health interventions (Renju et al., 2010) and the reading of cultural stories to boost positive adjustment (Wood et al., 2012). We came to the drawings well-primed to look for, and to find, evidence of children exercising 'agency' to overcome adversity, supported by teachers and backed up by local church and women's groups in schools interfacing neatly with health services.

First analysis and findings
Our first analysis of our material (Campbell et al., 2014) used thematic content analysis (Attride-Stirling, 2001), isolating and extracting data segments from individual stories to highlight particular themes (e.g. 'caregiving responsibilities' or 'teacher support for children') taking the whole data corpus as our unit of analysis rather than individual stories. Here we focused on the role of schools in facilitating successful coping by AIDS-affected children, and we did indeed find evidence for what many would call children's agency, albeit in a minority of the draw-and-write exercises. We grouped this evidence into three categories, according to the categorisation of agency developed in our earlier study of children's resilience in Kenya (Skovdal et al., 2009). On this basis we emphasised children's agency in constructing positive social identities, in mobilising social support and in contributing to household reproduction. In relation to positive identities, our draw-and-writes made frequent references to the positive moral qualities of AIDS-affected children, suggesting that such children were succeeding in achieving social recognition as responsible and moral citizens. AIDS-affected children were also portrayed as drawing on school attendance and the symbolic value of schooling as a source of positive selfhood. School was also depicted as constituting a welcome distraction from difficult home life. In relation to mobilising social support, the draw-and-write exercises made references to children mobilising positive practical support from teachers, including pens, uniforms and referrals to NGOs; from peers, who offered emotional support and companionship as well as practical help; and from NGOs, CBOs and health services. Children were also depicted as exercised agency in sustaining their households, with multiple references to their roles in household chores (esp water, firewood and cleaning) and caring for sick relatives.

Second analysis and findings
As we continued to explore these drawings, however, it became increasingly clear that references to positive forms of agency and coping were relatively few and far between. Most draw-andwrites portrayed AIDS-affected learners as isolated, exhausted, stigmatised and bullied, neglected, hopeless and overwhelmed by the burden of household and caring chores, either in households where their own parents were too ill to care for themselves and their homesteads, or in foster families often depicted as callously viewing orphaned children as a source of free labour.
In European literature on children's stories, much is written of children's use of stories as symbolic vehicles for channelling unconscious fears of the chaotic, potentially dangerous and unpredictable nature of human existence (Von Franz, 1996;Propp, 1968). Prototypical western children's stories are full of exaggerated and highly imaginative accounts of wicked stepmothers, child-eating witches, dangerous monsters and so on, more a product of children's imaginations than a reflection of their daily realities. Similarly, it might be argued that our research participants were using the story medium to provide an exaggerated account of the unconscious fears that assail all human beings in all cultures and contexts, rather than a description of the particularly challenging realities of their daily lives.
However, the accounts in children's draw-and-writes are supported by an extensive research literature that gives equally negative accounts of the devastating suffering of AIDS-affected children. They also resonate accurately with information gathered in other parts of our own multi-method study from teachers, older high school children, and other community members in the same study site (Campbell et al., 2013(Campbell et al., , 2014(Campbell et al., , 2015. Furthermore, a key feature of the western literature is how the underlying structure of children's stories and fairy tales nearly always involves some resolution, where problems are overcome and dangers warded off at the last minute, by helpful adults, by lucky chance or by a child's own ingenuity (Bettelheim, 1976). The methodological research literature on the structure of narratives makes a similar point (Abell, 1987). As will be discussed below, such 'narrative resolution' was seldom achieved in the stories in the draw-and-writes.
As stated above, our first analysis used thematic content analysis. Data were analysed through clustering isolated segments of textoccurring across the pooled corpus of children's storiesinto themes in a manner that took no account of a text segment's location within the wider context of a particular child's overall story. Our second analysis approached the same draw-and-write data using a more holistic approachone that took a child's whole story as our basic unit of analysis, and did not isolate themes out of the context of the particular story. This approach resonates more with the narrative analysis approach (Riessman, 2008) than with thematic content analysis.
This second analysis of draw-and-writes highlighted how nearly all of the stories spoke of children whose independentreflecting-and-acting was implemented in situations that presented few opportunities for what children regarded as positive outcomes. One account after another spoke of children's desperate life struggles in situations where they often had no power to take control of any aspects of their lives at all, or to make what might be considered meaningful choices in the context of their own worldviews. Some were depicted as completely at the mercy of relatives or other adults who had reluctantly taken them in when their parents died, expecting them to do endless physical work, or not feeding them properly so that they fainted at school. Other children were exhausted by caring for dying parents, not able to concentrate at school for fear their parents might be dying at home, or falling asleep in class from lack of sleep or food or their own illness. Some were depicted as bullied by peers at school and beaten by children in the neighbourhood, crying uncontrollably.
In short, whilst our first analysis of the draw-and-write exercises indeed highlighted instances of the exercise of agency understood as independent reflection and action by AIDS-affected children, our second analysis emphasised how these instances of agency took place in social environments that limited the outcomes of these, often depicting children in extreme distress, despite all their best efforts to act. We turn to pay more detailed attention to the results of the second analysis with attention to four key features that emerged from this more holistic analysis: children's implicit accounts of what they might regard as hallmarks of a good life; representations of support; the ambiguity inherent in their accounts of action; and degree of narrative resolution in the stories.

Children's implicit accounts of a 'good childhood'
As discussed above, Sen (1999) and Nussbaum (2000) highlight the importance of taking account of relevant actors' assumptions about the constituents of a 'good life' in conceptualisations of agency. What did our respondents themselves see as the prerequisites for a 'good childhood'? There was a clear implicit account of these in participants' stories of the challenges and problems faced by their AIDS-affected peers, as well as a more explicit account of them in participants' accounts of the supports available to them. Interestingly, in our bigger multi-method study adults (teachers, community members) tended to emphasise the material challenges facing AIDS-affected children (food, clothing, school fees). In draw-and-write children tended to lay far greater emphasis on the emotional challenges facing peers (Campbell et al., 2013).
Appendix A provides detailed information about the nine criteria underlying the stories, their frequency across the data set, and representative quotes from the stories. Following Sen (1999:18 ) we might consider these as the factors that children considered as prerequisites for a good life, and a 'good childhood'. These included freedom from emotional and physical abuse and neglect (alluded to in 44% of the stories); household duties that left the child time for school and play (31%); opportunities for personal health and hygiene (29%); supportive and sustaining peer networks (28%); recognition of a child's integrity and worth (20%); enough food to get through the school day without fainting or falling asleep (18%); opportunities for regular school attendance (17%); a clean and orderly living space (15%); and the material necessities for school (fees, uniforms, pens) (15%). We discuss these further below.

Agency through mobilising social support
In Skovdal et al.'s (2009) account of the agency of young carers in rural Kenya, great emphasis is placed on children's ability to mobilise social support, through a range of strategies such as helping neighbours with housework in exchange for food, approaching teachers to ask for referrals to NGOs, asking adults to assist with taking sick parents to hospital and so on. Whilst our analyses were qualitative in nature, we include crude frequency counts of the proportion of stories that depicted children with or without social support. In the draw-and-writes, one third depicted AIDS affected children as lacking any source of support or help whatsoever: The boy I want to talk about is a boy who used to be sent to do many errands. This boy did not refuse to do the errands. When he was sent to do anything he would run very fast. When he returned home he would get beaten and was not given a chance to say anything, and he would be accused of playing instead of doing what he was told to do. That is why I am saying this child is being abused. His feet would be numb with the cold because he was sent on errands early in the morning as if he was going on a journey. (See Fig. 4, 'Child lacking support' for associated drawing) A minority of 1.5% of the drawings depicted an AIDS affected child as well, loved and adequately cared for. This included a story of a girl with AIDS, who lived a happy life surrounded by loving relatives. We construed a further 3% as positive stories insofar as the story consisted solely of accounts of the support the protagonist had been able to mobilise.
She faces problems of getting ill. Many people help her by providing good food. The school head assists her by carrying her in his car. He goes every morning to fetch her water in his car. Her parents do not work, they stay at home. The headmaster pays school fees for her. When its time to eat she eats at the teachers house, if she does not have a pen she is given by the teacher. People help her when she faces other problems (M125).
The other 62% of the draw-and-writes referred to situations where the protagonist had been able to solicit some source of support. This was often fairly minor or piecemeal in naturea teacher giving a child a bowl of porridge when they fainted from hunger and were unable to walk home, or some kindness shown to a crying child by a concerned peer. Furthermore references to isolated cases of support where often framed by a grim wider context, with the support only tackling one piece of a much wider set of problems.
I am writing this composition about a girl whose parents died from HIV. She is always sad because she doesn't have a brother or a sister. No one in the community would give her money for school fees. The teachers are the ones that gave her money so that she would go to school as the other children are doing. Other children who are mean beat her as if she is not a human being. (See Fig. 5, 'Presence of piecemeal support' for associated drawing.)

Ambiguous agency
Stories made repeated reference to children acting in ways that contributed significantly to their own, or other household members', well-being. Their main activities centred around their performance of household chores (particularly collecting and carrying water and firewood), caring for sick parents, relatives and siblings and making superhuman efforts to get to school, having to rise very early to complete their chores before school, and then often arriving at school exhausted and not having eaten. However in many of the stories, the unsupportive wider context meant that their actions yielded few benefits for the children beyond their minimal survival. In the story below, a hard-working orphaned child exercises agency through helping an aunt who has taken him in after his parents died. So this is arguably a story with elements of both social support and agency.
I know of a boy who is an orphan. His aunt takes care of him. He wakes up and fetches firewood and water. By the time he goes to school everyone is there already. Other children at school are cruel to him because he no longer has parents. When he comes back from school there is no food for him and he starts working without eating anything. This boy is in grade five. His parents were suffering from HIV and AIDS. The boy's aunt never gave him rest. When he came back from school he would be told to go to the grinding mill. (F32) Yet whilst he exercises agency through helping the aunt (fetching firewood and water, and working in the grinding mill) in exchange for accommodation and 'care', there is also a suggestion that she is exploiting him ('never gave him rest'), and that there is no food for him when he comes home from school.
In the story below, there is another example of a boy who could be perceived as exercising agency by choosing to defy the school rules, 'running away' during school hours to care for his sick relatives, and so on. However here again, this agency is exercised in a limited and limiting context that calls into question the analytical sense of viewing this behaviour as evidence for his agency: There is a boy who stays at a household where there are people who are sick with AIDS. Before he comes to school he makes sure that he has fed them, given them their medication, and seen that they have all that they need for the day. The boy faces lots of difficulties at school, like having to run away so that he goes to feed the sick and give them their medication, and most of the time he faints because of hunger. (F74)

Narrative resolution in stories
The narrative research literature highlights the significance of the ending that a respondent gives to a story (Riessman, 2008). To what extent did the draw-and-write exercises end in ways that implied any hope of the types of outcomes that children would hope for as outlined above? Frequency counts pointed to the number of stories with positive, ambiguous and negative endings.
About 10% of the stories did indeed have a positive ending, often through peers offering some sort of support and hope to the AIDS affected child: I am writing a composition describing a girl whose parents are sick because of HIV. The girl is called Paulan. Her parents are all infected with HIV. When people knew the problem they started laughing at her. At school she was so lonely and no one got near her, saying if your parents have HIV you have it too. Sometimes she spends most of her time in tears. At school we invented a club called Health Club. We bought some blankets for her parents and we comforted her. Sometimes we went to their house and helped her wash her parents and their clothes. After that everyone played with her and she became very happy again. (See Fig. 6, 'Story with a positive ending' for associated drawing) We categorised the resolution of 19% of the stories as ambiguous. On the one hand they referred to a series of challenges facing the child. However, there was also some indication that some of the problems (generally a minority of them) might be resolvable, although others were not, Other stories in the 'ambiguous' category consisted of a negative story, but one that ended on a potentially positive note: Tom is my next door neighbour. We live in X village. Tom's parents are dead in 2008. His father had many girlfriends. His mother had only one partner. Tom's father caught HIV and AIDS. It lasted for only two months and he was dead. Tom always buys fruits to eat. He was tested HIV and was found HIV positive. He always vomits. He is talented in school work. He does not have anyone to pay his school fees. We see him everyday begging money from people. Tom is very kind to others. I hear most people making fun of him when he tries to play with them. He does not have clothes. He is 12 years old. He has two brothers. He looks after them everyday. He pretends to be a good mother. He is always happy with his brother. (See Fig. 7, 'Story with ambiguous ending', for associated drawing).  However most stories (70%) failed to indicate any sense of hope for the child's future.
There is a girl called C, her parents are HIV positive, therefore she was infected from birth. Her parents tried by all means to get their child cured, but it was in vain. By the time she was going to school she was already on HIV medication. When she was at school some children would refuse to play with her because she had HIV. One day she made friends with a girl called R. R asked her what her problem was and she told her that her parents gave birth to her when she was HIV positive. R tried by all means to help her friend but she failed. That is my composition. (F78)

Conclusion
In line with the New Social Studies of Childhood, our first analysis dutifully sought out and documented instances of independent reflection and action by children. To what extent might we regard these as evidence for agency? Children were depicted as exercising agency in the sense of reflecting and acting, but because they often had so little access to significant power, resources or support, the actions they performed did not bring them closer to the outcomes they would value. We argue for the need for renewed debate about how best to conceptualise children's actions in contexts that (i) provide them with a highly constrained set of options for exercising initiative, (ii) where the exercise of choice in one arena of their lives might be associated with negative long term outcomes in another; and/or (iii) where the outcomes of their actions may not take them any closer to their own perceptions of a good life. In conceptualising agency, instead of positing independent action as an end in itself, we prefer to regard independent action as the means to end that actors themselves would value.
In a tangentially related debate in the field of gender studies, critical researchers of women in the global south are increasingly concerned by a tendency to exaggerate the agency of women in situations of extreme subordination and coercion. Attempts by well-meaning academics, activists and policy makers to avoid depicting marginalised women as victims have led to a situation where "the search for agency in the least favourable situations has reached almost epidemic proportions" (Madhok et al., 2013: loc 554). In the process, they argue that feminist scholarsmotivated by the desire to avoid potentially offensive depictions of 'third world women' as passive victimshave sometimes unwittingly aligned themselves with individualistic neoliberal understandings of agency and personhood. They have done this through advancing understandings of agency as any form of decontextualised individual choice and through celebrating actions by women that lead to nothing more than their basic survival. In the process they "neglect the oppressive structures of material and discursive power", generating understandings of agency that undermine attention to, and analysis of, gendered oppression (Wilson 2012): loc 2253).
Similar points have been made in another tangentially related debate in research about children's legal status in the USA. Ruddick (2007aRuddick ( , 2007b argues that unrealistic emphases on children's agency, in the absence of attention to the constraints on children's reflection and action, represent an inappropriate commitment to the 'neoliberal notion of the human subject'. They over-emphasise the potential for people to improve their lives through individual action in negative wider contexts which may make such improvement unlikely. Such a notion regards individual children's actions out of the context of wider contextual constraints on the outcomes of these. Scholars in the fields of international development (Harcourt, 2009) and health (Campbell, 2003) have highlighted a problematic tendency for development and welfare agencies to overemphasise the agency of poor and excluded groups. This masks the potentially devastating impacts of widespread reductions in welfare and international development support on the poorest of the poor in many contexts, particularly in wider contexts of global austerity.
As researchers concerned with foregrounding children's voices, the accounts we provide of such voices are not objective or neutral. They are constructions that we ourselves generate, and are heavily marked by our own personal politics, theories and research methods. We argue that there is an urgent need for greater attention to the interpretative frameworks used by researchers, as well as by social development and welfare agents who would seek to interpret children's voices, and represent their opinions. Our experiences outlined above suggest that even two analyses of a single set of drawings by the very same researchers might lead to very different conclusions, depending on their starting assumptions about what constitutes agency.
There is no doubt that, as the NSCC has now firmly established, children are able to act, show resourcefulness and survive, often with little help or input from adults. To that extent they are 'competent social actors'. This was a vitally important point to make in the 1990s. However in the light of the strong body of research generated by the NSCC tradition, we believe this can now be taken as a given. We argue that the next step for researchers is to pay greater attention to the factors that mediate between so-called agency and its outcomes, and, most important, pay particular attention to children's own accounts of their hopes for the future, and children's own visions of what would constitute a 'good life' from one social setting to another.