Obligation to family during times of transition: care, support and the response to HIV and AIDS in rural South Africa

ABSTRACT In rural South Africa, high HIV prevalence has the potential to affect the care and support that kin are able to provide to those living with HIV. Despite this, families seem to be largely resilient and a key source of care and support to family affected by HIV. In this article, we explore the motivations for the provision of care and support by kin. We use the results of a small-scale in-depth qualitative study conducted in 10 households over 6 months in rural KwaZulu-Natal, South Africa, to show that family obligation and conditional reciprocity operate in varying degrees and build social capital. We highlight the complexity of kin relations where obligation is not guaranteed or is limited, requiring the consideration of policy measures that provide means of social support that are not reliant on the family.

In South Africa, political upheaval, socio-economic control of individuals and migration, along with regime change, have changed family composition but also put pressure on family functioning and traditional norms (Nkosi & Daniels, 2007). Death or long-term illness from AIDS has the potential to further erode family systems that function to provide financial support, in-kind assistance and physical care, particularly in high-prevalence communities. Given the national antenatal HIV prevalence estimate of 29.5% (National Department of Health, 2013), changes associated with death or longterm illness from AIDS have the potential to place pressure on family systems of organisation that are intended to ensure that vulnerable members of familial networks are taken care of and financially supported.
This led to a body of research exploring the epidemic on families (Ankrah, 1993;Seeley et al., 1993), particularly the ability of the family to respond by providing care and support to sick members or orphaned children (see for example Heymann et al., 2007;Louw, Dunbar-Krige, & Fritz, 2010;Schatz & Ogunmefun, 2007;Seeley et al., 1993). Much of this research is framed by questions about the resilience of family functioning and the ability to resist dissolution in the face of the potential impacts of HIV and AIDS in the era largely before wide-scale access to treatment. South African evidence on the impacts of HIV on family suggest that although families experience difficulties providing support and may even be a burden, families in general are an important source of support and care for people living with HIV (Cross, 2001;Hosegood, Preston-Whyte, et al., 2007;Iwelunmor, Airhihenbuwa, Okoror, Brown, & BeLue, 2006;Smit, 2007). Thus, enabling resilience among those directly affected (Smit, 2007). However, much of this literature precedes widespread access to treatment and there has been little analysis of the factors motivating the provision of financial support, physical care and material assistance by kin to affected individuals and their families. In this article, we attempt to explore and understand the motivating factors that determine this provision of support and care to those affected by AIDS-related illness and death. This paper uses data collected in 2008 during a period of rapid roll-out of antiretroviral therapy (ART) in this context and during a period of more limited access to ART. Despite the relatively dated nature of the data and the fact that the outlook for those with HIV was more grim than currently the results are still significant in the sense that families are still dealing with repercussions of HIV and their members still require support and care.
Theoretical models of kinship, family obligation, resilience and social capital In his recent analysis of kinship, Sahlins (2011aSahlins ( , 2011b argue that, as exemplified in the myriad examples of ethnographic work he cites, kinship is a "mutuality of being" (Sahlins, 2011a, p. 2). Kinship is therefore based on principles of intersubjective belonging, inherent dynamics operating both as a function of this "mutuality of being" and in to maintain kinship ties. Prior research with South African black families suggests that normative forces of family obligation may have shaped the provision of support (Sagner & Mtati, 1999;Siqwana-Ndulo, 1998;Viljoen, 1994). Norms of family obligation determine the extent to and ways in which family and kin provide support, and highlight the motivations of doing so, including for practical reasons to ensure that policy (and associated programmes) aligns with social reality (Van Bavel, Dykstra, Wijckmans, & Liefbroer, 2010). The research we report here provided us with an opportunity to explore whether family obligations are still important in the provision of care and support after two decades of the AIDS epidemic.
The moral obligation to help or support kin is often conceptualised in the literature from the global north as a family "obligation", commonly characterised by exchanges within the nuclear family or with others also closely related by blood (Del Corso & Lanz, 2012;Finch, 1987;Finch & Mason, 1991;Van Bavel et al., 2010). Sahlins' (2011aSahlins' ( , 2011b conceptualisation of kin and Mkhize's (2006) argument for the collective existence inherent in African families extends this conception of family and associated obligations to the wider kinship network, whether its members are related by blood or through social relations. Such obligations are social norms, although individual commitments are constantly renegotiated depending on membership, personal and family circumstance, and interpretation (Van Bavel et al., 2010). Obligations are invariably complex, deriving from a socially sanctioned duty to family, a desire to help family based on the quality of a relationship, or because of covert or implicit self-interest operating at the individual level. Kinship has historically shaped many of the social norms by which individuals negotiate their relationships, interactions and responsibilities to other people in Zulu and other South African families and society (Preston-Whyte, 1974;Russell, 2003Russell, , 2004Vilakazi, 1962). The patrifocal lineage system has been important in determining not only norms of residency but also domestic responsibilities and obligations within the family and amongst close kin. Historically, ties within wider kinship networks were cemented through the integration of individual household or homestead production into that of the kinship network and ensuring reciprocation, sharing and co-operation in production (Sansom, 1974).
Norms of household composition have changed and household typologies increasingly vary because of a range of factors, including high levels of circular migration, fertile extra-marital sexual unions and fewer marriages (Amoateng, 2004;Hosegood, McGrath, & Moultrie, 2009;. These structural and compositional changes also affect associated obligations and social norms. However, despite these changes and related shifts in social norms, the research suggests that many of the traditional norms of obligation, social networks and relationships continue to function in South Africa (Nkosi & Daniels, 2007;Siqwana-Ndulo, 1998).
Although complex and not always simply defined, the concept of resilience and its use in the analysis of the impact of AIDS on households and families in the South African context persists (Ankrah, 1993;Samuels & Drinkwater, 2011). Resilience in the context of HIV and the family suggests the capacity to manage and in some cases mitigate the impact of HIV (Loevinsohn & Gillespie, 2003;Seeley, 2015). While the bulk of the literature focuses either on the emotional or psychological resilience of socio-economic resilience (livelihood), this paper adopts a more functional approach to the notion of a resilient family. Firstly, the resilient family manages to largely maintain its integrity in terms of remaining a construct both in definitional terms but in the mind of the family members. In addition, it remains both functioning in terms of the family activities and cohesion. The integrity of the family as a social structure is maintained through social systems of exchange, moral obligation and link to the development of what some call "social capital".
Social capital although also contested has been used for examining issues such as food security, livelihoods and the burden of HIV and AIDS (Burger & Booysen, 2006;Misselhorn, 2009). For this paper, we adopt a definition proposed by Burger and Booysen (2006); they differentiate between the claims people have on resources and support from a social network, and the building blocks of social capital, including the norms of reciprocity, familial obligation and trust developed within social networks.

Study setting
In this article, we present the results of a qualitative study conducted in 10 households during 2008 within the UMkhanyakude district of northern KwaZulu-Natal, South Africa. Study households, situated in both rural and peri-urban areas of the district, were located within the Africa Centre for Health and Population Studies Demographic Surveillance Site (DSS) area. The population in this district highly mobile, with frequent changes in living arrangements and large numbers of non-resident and multiple household memberships Tanser et al., 2008). Households may include members of multiple generations and people considered to be kin either through biological or social ties (Ross, 1996;Spiegel, 1996). Consequently, as elsewhere in South Africa, household membership is complex and changeable. Analysis of the DSS data for this district shows increased rates of orphanhood related to AIDS between 2000 and 2005. Living arrangements of children in this context are complex, and child-headed households remain a small minority with orphaned children living with relatives and patterns of familial fostering prevalent prior to the continuing HIV epidemic (Hosegood, Floyd, et al., 2007;Tanser et al., 2008). The livelihoods of households in this community are dominated by social welfare grants paid by the state, paid employment but not necessarily in the community and agriculture (although practised on a smaller scale) (May, 2000;Muhwava, 2007).
Antenatal HIV prevalence in the district was just over 35% in 2012 (National Department of Health, 2013). Analysis of the HIV incidence in the study area between 2003 and 2007 showed little decline in incidence with an overall incidence of 3.4 per 100 personyears ). The availability of antiretroviral therapy (ART) since 2004 has reduced AIDS-related mortalityapproximately 22% for women and 29% for men between 2002 and 2006 (Herbst et al., 2009). By 2008, in the DSS population of about 85,000 people, about 7500 people had initiated on ART (Hontelez et al., 2011). Despite this, AIDS remains and was at the time of this study the leading cause of death within this community (Herbst et al., 2009). Those who were testing and accessing treatment were still doing so at a low CD4 count of <200 cells/µl and Stage 4 symptomatic illness with an increased risk associated of both morbidity and mortality despite access to ART (Herbst et al., 2009;Houlihan et al., 2011). The burden of HIV-related morbidity in South Africa has meant that the bulk of those with HIV cannot be cared for in facilities and even those who are quite unwell are discharged for care within the community supported to varying degrees by home-based care organisations and community health workers (Singh, Chaudoir, Escobar, & Kalichman, 2011).

Methods
All households were selected from within the DSS community, and were purposively sampled according to whether there had been a death of an adult household member from AIDS or whether it had a member living with HIV, to ensure a range of experiences of HIV illness and death, six months prior to the study. As a result of the sampling criteria, it was necessary to pre-identify households with experience of HIV and where either the cause of death, or in the case of illness, the index person's HIV status was disclosed to another household member. In order to fulfil these criteria, we employed various means to identify households and to ensure the inclusion of households with varied characteristics. Five households were identified by the local Catholic Church home-based care programme, one by the verbal autopsy staff from the Africa Centre and yet another household through an opportunistic contact. Three additional households had been part of an earlier study that investigated the household-level impact of HIV prior to ART (Hosegood, Preston-Whyte, et al., 2007;Montgomery, Hosegood, Busza, Timaeus, & Timaeus, 2006).
A series of six semi-structured interviews (each guided by a topic guide and informed by previous interviews) were conducted with members of each household, in conjunction with non-participant observation, which is conducted at both interview and subsequent household visits. The use of detailed and regular interviews, and observation of household circumstances, enabled us to collect detailed retrospective and contemporary data about the changing household situation, context and their experiences of illness and death, over a 6month period. In total, 60 interviews were conducted. Household genograms enabled the collection and collation of household composition data while household events map chronicled important episodes or events relating to illness and death. Both techniques provided important context and history (Adato, Lund, & Mhlongo, 2004). Frequent data collection and the longterm involvement of the research team with householders encouraged trust, rapport and the sharing of sensitive information (Christensen, 1992;Murphy, Dingwall, Greatbatch, Parker, & Watson, 1998). Multiple individual interviews were conducted with both male and female household members (Montgomery et al., 2006), providing differing perspectives. Data were analysed and updated throughout the period of fieldwork, allowing for consistency checks and theory testing with respondents (Ezzy, 2002;Green & Thorogood, 2004).
In order to prevent inadvertent disclosure and to protect households from HIV-related stigma, issues specific to HIV status were only discussed with respondents or, with their permission, other household members. Prior disclosure by the affected individual of their HIV status to at least one other household member was a prerequisite for household inclusion. While interviews dealt generally with issues of illness and death, some respondents voluntarily disclosed their HIV status to the study team.
All adults in the households received detailed study information and participants provided written informed consent during the first encounter. One household withdrew after three months, citing a sick member's discomfort with our presence. Ethical approval was obtained from the Humanities and Social Sciences Research Ethics committee at the University of KwaZulu-Natal, and the Research Ethics Committee at the London School of Hygiene & Tropical Medicine.
Interviews were conducted by a trained, locally resident research assistant along with the principal researcher. The data were collected in isiZulu, transcribed and translated concurrent to data collection by the research team, with results used to inform further interviews. NVivo software was used to code transcripts and fieldnotes cross-sectionally using framework analysis (Mason, 2002). The process was iterative, and a coding framework based on the primary study objectives and emerging issues was revised throughout (Green & Thorogood, 2004;Ritchie & Spencer, 2002). Data were also collated and analysed to develop household case studies, providing detailed descriptions of household's experiences in context and changes at a household and individual level (Mikkelson, 1995;Russell, 2005). The case studies and cross-sectional analyses were then compared. Pseudonyms were given to participants to protect their identity.
Data to corroborate these findings were collected from a present and willing adult during working hours on weekdays. These were the safest times for travel within the community as a result of crime and poor roads and access and the times when logistical support from the Africa Centre was available. One respondent was interviewed near her workplace so as not to exclude households with working members. These time limitations determined the study sample, so that in the end respondents largely comprised the elderly and the sick, the unemployed, school-going individuals, those on holiday or shift workers. Respondents were therefore mostly women and relatively old, potentially introducing bias in gender and age into the sample.

Results
The rural household was the unit of analysis for this study. The household is a contested concept, especially in South Africa where composition is complex, membership is fluid, and not defined by place of residence or biology but rather by affiliation Wittenberg & Collinson, 2007). While this study addressed the household-level impacts of HIV and AIDS, the results demonstrate that the role of broader family networks, extending the definition of family beyond that of the nuclear and including kin both resident in and external to the household and vital to family functioning. Therefore, while the locus of study and those interviewed were mostly resident household members the relationships to those outside the household were considered and where possible non-resident household members were also interviewed.

Familial care and support
This study considers the care and support that family are able to provide to each other; this is more broadly defined than just physical care but extends to the financial support and in-kind assistance required to meet the needs of those affected by HIV (Moyer & Igonya, 2014). This paper is not about access to health-care but certainly access to adequate care and support as provided by family can facilitate access to formal health-care and treatment. Family members were a key source of direct support, care and assistance for HIV-affected individuals and households. Despite the difficulties that the households faced in providing support and responding to the needs of sick people, affected families were able, in most cases, to mobilise their often-limited resources to do so.
Direct access to financial support from social grants enabled household members to be provided with financial support and care for their physical needs by their family members (Knight, Hosegood, & Timaeus, 2013). To a lesser extent, income from employment was also important although it was often received in the form of remittances and more likely to be specified for spending and less likely to be pooled than social welfare income. For example Thembi Nkosi's access to a disability grant facilitated not only her own and her husband's access to health-care but also enabled her to feed her family. Other families pooled formal and informal sources of household income, including remittances and social grant monies. This joint income was used to support all members of the household through the provision of food and also facilitated care of or access to care for those who were unwell. For example in the Dlamini household, income from a number of child support grants and one old age pension was pooled to ensure that the children being cared for in the household were fed.
Family members also provided financial assistance for general household expenses including illness and funerals. Other family members contributed surplus food from their fields or gardens, or purchased extra food for affected households when they went grocery shopping. Precious Sibaya and Nomsa Bhengu both spoke of the in-kind assistance their households could rely on from family members who lived elsewhere in the community when they were in need. Some other households received other in-kind assistance such as helping to check on those who were sick, preparing food or assisting with weeding or planting.
An obligation to support family, defined broadly, was expressed in varying degrees by all participants. This obligation differed between families in terms of how it was operationalised and influenced by various personal and societal factors that determined the types of support that were provided and to whom (Van Bavel et al., 2010). In the results that follow, we illustrate the general trends we observed throughout the course of this study.

Unconditional obligation
The general sense of obligation to those considered family was noted among all participants in this study sample and is illustrated by this comment: "I just help my family. I think it is right to help other people" (Gugu Dlamini, daughter of female household head, 33 years). These unconditional obligations observed and narrated in this context seem to be influenced by social norms dictating a duty and responsibility to family.
An individual's obligation to provide support or assistance to kin was often influenced by affective ties or a close kin relationship. Tina, for example, physically cared for her HIV-positive and orphaned grandson. She lived with this grandchild and two of her single sons. I don't have any problem with [caring for the child] because I know that he is my child's, so he is mine too. (Tina Ntuli, female household head, 63 years) As in this instance, mothers felt a significant sense of obligation to provide support, mostly in the form of physical care to their children, or for their grandchildren, whom they saw as an extension of their children. The bonds and obligations between parents, children and grandchildren were not strictly biological, as social parents who were assigned these roles within society also took on the accompanying obligations (Mkhize, 2006).
Nobantu (a 55-year-old female household head who was caring for her sick adult daughter and two older single and otherwise unsupported brothers, all requiring substantial care and support as a result of illness) demonstrates the bonds and strong obligation of parents to their children. Nobantu fed, cared for and helped her brothers take their medication for their symptoms every day. She provided more active support for her 39-yearold daughter, Lindiwe, paying for her hospital attendances and ensuring regular clinic visits. Nobantu also cared for Lindiwe's newborn baby. Lindiwe, along with her three siblings, two cousins, her two children and a nephew, lived in her mother's house, while her uncles, single with no children living nearby, slept in an illequipped and unfinished structure separate to the main house. Nobantu's decision to support her brothers seemed to be dominated, in part, by their extreme need but also by an obligation to kin, as she explained: "They are my brothers, they have no-one else". Despite this, the levels of observed support and the strength of her motivation to provide it were different for her daughter and her brothers, whether she consciously made this decision or not.
The quality of the relationship with family, influenced by frequency of contact, trust and feelings of closeness, played a role in Tina's family. Tina explained that relationships which might be considered distant, such as her relationship with her uncles, was much closer in reality and this was reflected in the support that they show to her family: "It is my uncle's home [who will provide support]. They are just brothers … If we have problems like we are hungry, they help us." (Tina Ntuli, female household head, 63 years)

Gendered obligation
While often without conditions for support, family obligations and expectations are often gendered. As dictated by traditional gender norms, responsibility for regular domestic activities, day-to-day decision-making and care, primarily falls to women in the study householdsmothers, grandmothers, sisters and daughters. As Gugu's mother explained, "(i)t is [Gugu's] job now [to take care of things and people within the household] … it is because she is a girl and also because she was born here" (Ntombizodwa Dlamini, female household head, 70 years). It is also possibly important that Gugu is the present and capable female child of the household head increasing her responsibilities to the household as the head's duties are delegated to her by her very elderly mother. Men's role within the family was also largely defined normatively with both men themselves and their families positioning them as the expected breadwinners with responsibilities to work outside of and provide mostly financially for the household. Yet, the situation was different for Zinhle Bhengu, whose work for the household, although very important, was not within the norms of that expected for her gender. Here her mother describes her adult daughter in the masculine as the provider: "[Zinhle] was the man [of the house], there were deliveries to the house before she was sick but now there is nothing" (Nomsa Bhengu, female head, 61 years). Thus, even when gendered norms are not practised, the discourse around gendered norms reinforce the status quo rather than working against them. This masculine support role taken on by Zinhle was in contrast to Gugu's sense of responsibility to take care of the domestic realm.
Although households are no longer necessarily organised according to traditional principles, a normative gendered division of roles is still pervasive within the narrative about household responsibilities. This determines the social expectations of and obligations to family.

Reciprocal support
In addition to a sense of unconditional obligation influence by social norms, many families either overtly expressed or demonstrated through their organisation of interactions and support, a hope that their provision of assistance to kin would help us to maintain future mutual support. This sense of reciprocity has a long history in southern Africa: child fostering practices and care for and by the elderly has long been a part of a mutual system of social support among kin (Foster, 2000;Monasch & Boerma, 2004;Schatz, 2007) although not always guaranteed (de Klerk, 2012). Nomsa's obligation to support her children is informed by her own hope for security and support as she gets older, even though she cannot guarantee this, and by her children's expectations; "There is nothing else for me to do [other than help] or my children will grow up and [neglect] me if they think that I was not helping them" (Nomsa Bhengu, female head, 61 years).
In some cases, the reciprocal nature of the exchanges in times of need was clear and mutually beneficial. Here Precious provides examples of in-kind and financial support she and her family receive from other family members and she also describes how they return the favour when called on.
Oh, they are so helpful because when I don't have something they give it to me and what they don't have I give to them. [They give] any kind [of help]even if it's money, they had a funeral and we helped them … Even if we don't ask they give it to us and if they have a lot of something they share with us. And we also help them … It helps us get out of trouble if we get what we need. (Precious Sibaya, wife of head, 36 years) Family obligations and expectations are maintained by family bonds and trust, as described above, but also by economic or material reciprocity over time acting as social insurance. For example, non-resident members, such as Zinhle who as an adult, over a relatively long period of health and working prior to her illness and moving home, remitted earnings and brought necessities when they visited their rural households. This helped us to maintain relationships, and the investment over time ensured that, according to norms of family obligation, they could make claims on support, care and assistance when needed.
Expectations of support from family existed in explicit statements about obligations of family members to each other, and was strongly felt where there is an observed ability to provide support on the part of other members. These expectations of support were tracked and noted by family members over time. Here, a step-mother speaks about her sick step-son's failure to contribute financially to the needs of the household.
Mandla doesn't give us anything. He keeps his money [from disability grant] in his pocket … ..We don't know how he spends his money, he doesn't help us to buy food … . Their father buys the food … ..and I also buy [food] with the money from the child support social grant. (Precious Sibaya, wife of head, 36 years) Mandla, whose health was deteriorating and who was refusing to take his HIV medication or remain in the hospital to control his symptoms, risked his stepmother refusing to care for him in the future by withholding pooling his financial means with those of the household. At the time of the study, however, he was also receiving in-kind support and physical care from his mother's family. This knowledge that care was being provided elsewhere meant that Precious had no current responsibility to care for Mandla, but it was clear that she was keeping track and that his unwillingness to fulfil his obligations may affect whether Precious provided care in the future.

Socially sanctioned and displayed support
The support of family is complex and driven both by a sense of personal or conditional obligation and by a concurrent expectation and desire to be seen to be acting in a socially acceptable way. Tina expected material support and financial assistance from an adult son, and spoke of the potential for public embarrassment should it not be provided: We returned from our journey with no food, I was starving. I was going to be mocked by the other women for not having supportive sons. He said he didn't have time but that he would try … He said he would try to come and give me provisions that I need. I said I need juice, meat and buns without sugar. (Tina Ntuli, female household head, 63 years) The social expectation that families contribute in some way, and the reciprocal nature of this support, although pervasive in terms of influencing a duty or responsibility to family, was particularly apparent at the time of deaths and funerals. Most family members, even those without close relationships to the bereaved household, provided some financial support or in-kind assistance with an expectation of future return. In-kind support included the preparation of food and general assistance with preparations for the funeral. Funerals therefore provide people with opportunities to demonstrate explicit support, both affective and practical.

Complex and conflicted circumstances
Although in general non-resident or extended families provided the most financial and in-kind support for affected households where they were able, this was not without its challenge. One reason was that poverty and the widespread impact of AIDS limited families' available resources. Another reason was that some individuals had conflicting obligations to new families, limiting their ability to act "like a member of the family" as described by Tina Ntuli: "I miss [my son], because he helped me … but now, he has the worst girlfriend. She doesn't want him to share his money with his family" (Tina Ntuli, female household head, 63 years). This shift in obligation to another family is not a new phenomenon, but impacted on relationships and obligations among certain family members and may have been exacerbated by changes in living arrangements sometimes causing a move away from communal extended family living (Hosegood, Benzler, & Solarsh, 2005). This seemed to be common among men whose obligations were split between their wives and children, with whom they resided often in urban areas, and their birth families in the rural areas.
Certain households felt isolated within their kinship networks, when relatives were unwilling to provide support or to be associated with them: "I have nobody, there is my brother in Durban but he doesn't care about me … He has money. He is supposed to help" (Thobela Nkosi wife of head, 36 years). The lack of close family links, such as those broken by the death of Thobela's parents or exacerbated by migration, led to family members being distanced from one another. The one-sided nature of this data means that it is not possible to understand that families reasoning for not providing this expected support. Isolation was more difficult in households where livelihoods of household members were mostly reliant on or dependent on their social relationships.

Discussion
Despite the burden AIDS placed on the households and the difficulties they experienced in responding to the consequences of illness or death, our results support existing research that argues for the continued ability of South African families to respond to the impacts of illness and death (Goudge, Gilson, Russell, Gumede, & Mills, 2009;Sagner & Mtati, 1999;Siqwana-Ndulo, 1998;Viljoen, 1994). Families in this study provide an important safety net, albeit one "with holes" as argued by Seeley et al. (1993) almost two decades ago, to both affected households and individuals. The kinship network is a resilient source of social capital for the majority of those affected. In a context where in-patient healthcare is limited and those infected and affected by HIV require not only physical care but also financial support and in-kind assistance, these and other results demonstrate the remarkable ability families have to adapt to and respond to the implications of HIV in rural South Africa and beyond (Baylies, 2002;Cross, 2001;Iwelunmor et al., 2006;Seeley et al., 2008;Smit, 2007). This is not a new finding and not the major contribution of the paper but has relevance because it provides us with evidence from a dark period and place where despite access to ART, HIV was still the most common cause of illness and death. Despite the burden of HIV and the way in which HIV changes the family dynamics and household composition in this study, the results show that families were still actively deciding to provide care and support.
In addition, the results presented here unpack the social forces underpinning the decision by family members to support and care for family or kin affected by HIV and AIDS. These building blocks of social capital for rural households are assumed within the literature about family care and support, both preceding and based on the impacts of AIDS in South Africa, but are never fully explored nor understood as the mechanisms underpinning ongoing support and care in the current context of access to treatment. This support and care is influenced by varying degrees of family obligation depending on the quality or intimacy of the relationships and also strongly dependent on social norms. We have also noted that family members' obligations were in many cases gendered but also depended on personal circumstances and standing within the family. Obligation to family may have conditions attached or bring the expectation of reciprocity. Despite examples of support as a result of family obligations and reciprocal norms, family dynamics are complex and the results provide examples of affected individuals or families who were inadequately or not supported, or where obligations were disputed or complicated by personal circumstances.
The obligations operating in the study households ensure family assistance, support and care through the reorganisation of resources to provide for those directly and indirectly affected by HIV. The persistence of norms of family obligations observed within this article support arguments for the pervasive nature of family members' obligation, duty and responsibility to one another despite problems or barriers as observed within research in East Africa (Moyer & Igonya, 2014;Reynolds Whyte, 2005). The obligation to support family and kin is governed by culturally and socially constructed norms. Norms observed in other South African studies of family support and care outside of the context of AIDS, and argued for within the theory of familial obligation (Bozalek, 1999;Finch & Mason, 1991;Sagner, 2000). These moral obligations to provide support in this context were often felt more strongly within close familial relationships, such as filial or sibling relationships. Individuals' obligations to support family are therefore fostered through close ties, trust and affection, and desire to provide help. Intimate and personal care, including of sick householders and dependent children, support the assumption of a "hierarchy of obligations" felt most acutely by nuclear families (Finch & Mason, 1991. Within the household, a gendered element exists to the obligations household members feel to provide assistance and support, which reflects traditional social roles. Unmarried women have a greater responsibility for their household than men who, although they are expected to support the household financially in times of crisis, seem to have less of a responsibility for the day-to-day running and functioning of the household (Preston- Whyte, 1974;Sansom, 1974). This is supported by more recent South African research that shows that the domestic realm and responsibilities for physical care in the age of HIV tend to fall to women (Harber, 1998;Schatz, 2007;Schatz & Ogunmefun, 2007). This is supported by anthropological evidence from Ugandan research also in the context of HIV (Reynolds Whyte, 2005). More recent evidence, however, calls these predominant stereotypes into question and suggest that they are not always played out in reality in other families affected by HIV within this study context (Montgomery et al., 2006). This is supported by the example from our study. Depending on a range of circumstances, women may be the breadwinners while men provide physical care. Such circumstances include the feminisation of the low or unskilled workforce in South Africa (Casale & Posel, 2002). Even so, in the study area, conventional gendered norms, obligations and expectations still dominate discourse about social expectations and obligations.
As the historical evidence on fostering and remittance behaviour in South Africa and the results suggest, the ties and bonds between family members were not only practical and social, but at times involve financial assistance or material support for the rural household or individual (Madhavan & Schatz, 2007;McDaniel & Zulu, 1996;Posel, 2005). While financial support, in-kind assistance and physical care provision for those affected by HIV were generally underpinned by unconditional moral obligations to kin, not all the motivations for support and care were easy to classify. In many cases, the results show that the decision to provide support appeared to be influenced by both a sense of obligation and a conditional desire for reciprocity. Bray (2009) has described exchanges of childcare and financial support between siblings affected by HIV, and Abebe and Skovdal (2010) have illustrated reciprocal relationships between orphans and adult carers. Conditional obligations to family were often tracked historically with past contributions noted by family or kin, and influencing future support, almost acting as future social insurance.
Family obligation and reciprocity also play an important role in rural South African families not affected by HIV and AIDS, and are dictated by social norms (Everatt, Habib, Maharaj, & Nyar, 2005;Haddad & Maluccio, 2003;Russell, 2003). These close relationships and norms of obligation and reciprocity help affected individuals and families respond to the impacts of illness and death. The importance of an obligation to family in South Africa has been acknowledged by Ross (1996), Sagner and Mtati (1999) and Bozalek (1999) as motivating various forms of support and care within the household and kinship networks.
The provision of support and care, whether conditional or unconditional, is often also socially sanctioned, expected and monitored. This may reflect a desire to want to be seen to be what has been termed "doing family" and presenting a public display of cohesion and quality of family life to the outside word. Funerals and other social gatherings offer perfect opportunities for such displays of support for family (Finch, 2007). This display of kinship solidarity has been suggested as significant in the motivations for supporting households at a time of death by Bahre (2007) in his research in the Western Cape. This also serves as a reciprocal relationship and the norms associated with the event mean that contributions to affected households within the kin network secured reciprocal assistance from these households in the experience of a death in the contributing household. Elsewhere evidence suggests that failing to provide care has been portrayed very negatively and is also strongly socially sanctioned (de Klerk, 2013;Moyer, 2012).
High levels of expectation of family obligation and care were observed in the study and are supported by findings from East Africa that suggest an expectation of a "right to care" from family (Moyer & Igonya, 2014, p. 138). While these and other authors argue that this right to care is countered by a moralising blame for those who are sick by kin unable or unwilling to provide the level of expected care it is less clear in this setting and study where the kin in the study were not those failing to provide care (Dilger, 2008;Moyer & Igonya, 2014). Despite high levels of expectation belonging to familial networks does not automatically lead to support and care. Some families had limited resources and their own vulnerability influenced their ability to provide support. Others were constrained by conflicting obligations, family disputes, or emotional and geographical distance. Other research shows that households affected by AIDS and experiencing conflict were more likely to feel stigmatised and unsupported (Hosegood, Preston-Whyte, et al., 2007). This resulted in affected family members who received inadequate support and some that were completely excluded from the kinship network and were without social capital. Similar conclusions about the exclusion of households from social networks as a result of poverty and an inability to reciprocate were drawn in Nombo's (2007) work in Tanzania and suggested by the likes of Seeley et al. (2008)in their conclusions. In the examples in this study the exclusion of the household was made more severe because family members seemed to have actively chosen to distance themselves from the study household, despite the high level of expectation. This makes consideration of possible social isolation important in assessing the ability to make claims on kin, and may result in increased vulnerability and poverty for those isolated (Adato, Carter, & May, 2006).

Conclusion
The results suggest that despite changes to the household that have been observed by the likes of Viljoen (1994) and Amoateng (2004) and the disappearance of certain aspects of tradition, the norms which govern obligation, social ties and familial relationships in the households enrolled in the study and affected by HIV have largely managed to remain intact and ensure a certain amount of resilience. These findings therefore point to the fact that social forces such a social obligation and conditional reciprocity still function and are important determinants of whether families affected by HIV are able to rely on social capital from within the kinship network. In the South African public health context where HIV is community-based, family support and care, and understanding how best to facilitate this is central to HIV care and treatment interventions (Betancourt, Abrams, McBain, & Fawzi, 2010;Lewis Kulzer et al., 2012;Richter et al., 2009;Rotheram-Borus, Flannery, Rice, & Lester, 2005). An example of such a family-oriented programme is the provision of skills and support for those providing home-based care, thereby supporting the existing roles of the family. An important policy intervention would also be the provision of adequate social welfare to facilitate both direct and indirect household coping and enable family support (Knight et al., 2013;Richter et al., 2009). Although not providing families with the capacity to develop social capital access to social welfare enables family members to provide financial support and in-kind assistance. Therefore possibly facilitating access to care, and enabling individuals to fulfil the obligations they have to family. In addition, social welfare also provides a vital emergency financial safety net for affected households who are isolated and unable to rely on social capital from their various networks.
Family support, care and assistance for those affected by HIV and AIDS is important but inconsistent and dependent on a number of factors. Understanding the resilience of the family and its continuing ability to provide for the needs of its members requires understanding and acknowledging the complex relationships, norms and traditions that underpin it. In this article, we have shown that processes of family obligation, both unconditional and reciprocal, together with conditional support and social expectations, all are important in ensuring the provision of financial assistance, material support and physical care. They also contribute to the resilience of the family, enabling them to cope financially or provide better care for those who are unwell, in the face of AIDS. At the same time, families are greatly impacted by negative social and economic repercussions of AIDS. This means that family support is not always a given; as described two decades ago, it remains "a safety net with holes" (Seeley et al., 1993, p. 117). Our findings suggest that norms of familial support still function and are underpinned by traditional values that maintain the significance of family obligation and reciprocity, but there are examples to the contrary reflecting both societal changes and the hardships caused by HIV.