Understanding the social context in which individual risk behaviour takes place is essential for the design of effective interventions that aim to improve protective behaviour.1^–3 Structural aspects of society beyond the individual’s control—such as discrimination, economic marginalisation and resulting social inequity— have an important effect on epidemiological risk factors.1 Risk behaviours can be seen as proximate causes of sexually transmitted infections (STIs) through which the more fundamental processes generate ill health.4 Broadening the focus of research to include the root causes of sexual ill health means going beyond sexual risk factors and studying the social relations that underlie the sexual ones. In Pakistan, HIV remains concentrated among high-risk groups5 and, besides injecting drug users (IDUs) and female sex workers (FSWs), the National AIDS Control Programme (NACP) recognises male sex workers (MSWs) and transgender sex workers (hijra or khusra) as groups at higher risk. Individuals in the transgender category, known across South Asia as the “third gender”,6 are born as biological males or inter-sexed and have taken on a largely female persona and often become sex workers.7^–9 Social norms surrounding sexuality in Pakistani society, stigmatising attitudes towards sex work as well as the illegality of both sex work and male-to-male sex considerably constrain the individual’s ability to act and reduce risk. The understanding of how groups themselves experience, interpret and act on these structural determinants becomes crucial, not only for designing effective interventions, but also for minimising the likelihood of any unintended adverse outcomes.

In order to gain this “insider perspective” of these social aspects of STI transmission, specific methods of qualitative research are needed.10^–12 These methods depend on building relationships of trust with members of the target groups, yet all the aspects which marginalise these hard-to-reach groups also hinder their participation in research. Ethnographic studies, in which the researcher is placed long-term within a community for “participant observation”, is most suited for gaining an in-depth understanding but is generally far too time-intensive to be practical in applied research. An alternative is involving members from the vulnerable groups in the research process, building rapport and drawing on pre-existing relationships of trust within their network. Peer ethnography13 is such a method, using average members of vulnerable groups as “peer researchers”. In this paper we illustrate how different the contexts are for three distinct groups of sex workers while, at the same time, showing fluidity within the constructed categories. We also demonstrate the power of engaging peers in the research process to improve the understanding of the context in which vulnerable populations live and take health risks.

METHODS

The main premise of peer ethnography is to train members of the target population as “peer researchers” (PRs) to conduct interviews with peers in their social network. The approach aims to deliver a more insightful and realistic reflection of everyday realities as they are experienced by marginalised populations.13

Recruitment of PRs

PRs from MSWs and transgender sex workers (khusras) were recruited through a local non-governmental organisation (NGO) doing outreach work with them. The method specifically stresses training “average”—often non-literate—“peers” from the target group rather than “programme insiders” who occupy a distinct role or status such as peer educators. Involving gurus (head of sex worker networks and important gate-keepers) in the recruitment of transgender sex workers was essential for facilitating access.

Training of PRs and field work

Two workshops were run in December 2006 to train 15 PRs in each of the two groups separately. Each workshop consisted of a 4-day training programme in which research themes were discussed and adapted to the PR’s own conceptualisations. Using an interactive training approach, interview probes were grounded in the PR’s vernacular and rehearsed in role plays. The PRs were encouraged to elicit “stories” from their peers to get a thick description and for easy recall. Ethical considerations of anonymity and consent inherent in the peer ethnography method were emphasised: PRs were trained to ask questions in the third person (what do other friends like you say about …?), asked specifically to avoid accounts of personal experience and ensure names were not used during the conversational interviews. Pilot interviews were conducted in a field test towards the end of the training and probes were refined.

Over a period of 5 weeks, each PR interviewed three of their own peers. Each of these three peers was interviewed on three separate occasions using a pictorial topic guide developed during the training workshop to explore three main themes: (1) issues of social lives; (2) problems and risks; and (3) health and health-seeking behaviour. After every (or every second) interview with each of the three peers, the PR reported back to supervisors at the NGO, who took notes. These field notes were translated and continuously reviewed by local anthropologists to enrich their understanding of the context and to identify themes/develop probes for individual interviews with PRs. Each PR met with a supervisor at least six times during the field work. Thirteen khusra and 12 MSW PRs completed the field work.

One-to-one interviews and analysis workshop

In February 2007, after reviewing supervision field notes, in-depth interviews with the PRs were conducted by the anthropologists and social scientists of the main research team. PRs were given the chance to tell their stories and express their understanding of issues, while researchers probed further on emerging issues. Twelve khusra and 11 MSW came for individual interviews. These interviews were taped (except for three khusra who did not consent), transcribed and translated from Urdu/Punjabi into English. For each group there was a final 1-day workshop using participatory analysis to consolidate the shared understanding of issues by PRs and the main research team, as well as refining categories used in the instruments for the bio-behavioural survey (see paper by Hawkes et al elsewhere in this supplement).14

Analysis of data

The analysis of the qualitative data is based on the transcripts of the one-to-one interviews which are considered the “primary data” in this method. However, the anthropologists and social scientists were present during the training workshop, in-depth interviews and final workshop, with “observation” adding valuable insight into the meaning of what the PRs reported. Informal interactions between local anthropologists and PRs also informed the interpretation of the text-based data. The data from each group were analysed by at least three researchers. The software Atlas-ti was used for data management, with transcripts coded under each main theme with subsequent subthemes being identified.

RESULTS

During the training workshops, early on in the research process, we had ample opportunity to interact with and observe the interaction among people from the target groups. It became clear that there are two different subgroups within the MSWs: khotkis or feminised males and banthas with a mainstream male identity. The latter were partially excluded and partially mocked by the khotkis during the MSW training. While banthas do not have a stigmatised sexual identity, for khotkis and khusras the sexual identity is both a cause for discrimination as well as central to social organisation and access to peer support structures. Hiding both sex work and a stigmatised identity from family, neighbours and authorities is a daily preoccupation. Incorporating these concerns and the PRs’ own conceptualisations is part of how research objectives are negotiated and how rapport evolves through engagement with both the research team and the research itself. An incident occurred during the training of khusra PRs where a member of the research team had unwittingly probed too much into relationships with FSWs. One PR had become very angry, with palpable levels of discomfort among the entire group. Because of vehement denial, all reference to sex with women was left out of the topic guide. As explained later, this was not at the cost of data loss.

Distinguishing between the three groups of sex workers

The khusras are tightly organised around common transgender identity within a hierarchical community led by gurus. They are very visible in the society and accepted for their tradition of begging, as well as giving “blessings” and performing in their neighbourhoods, especially at weddings and ceremonies for the birth of a son. They dress and walk like women, use make-up and refer to each other in the feminine (“she” and “her”). The most important social relationship within the khusra network is guru-chaila (master-disciple). A chaila pays part of the monthly earnings (usually 50%) to her guru in exchange for learning the skills of the trade, customers and a safe place to have sex, as well as a designated area to beg and a chance to perform at functions. This financial contribution as well as respect and loyalty to the guru ensure a recognised position (a name and identity) in the khusra community with the support, solidarity and protection this brings. Khusras live away from their native families, having migrated to another town and living with other khusras in dairas (rented places). Their occupation remains mostly secret from their families. Despite the physical separation, the connection to the family remains strong with frequent visits and remittances home. They still perform important cultural male roles like “shouldering the coffin” during funerals of relatives.

Khotkis also have gurus and maintain very strong links with other khotkis around a common feminised identity. In contrast to khusras, most still live at home with less freedom to cross-dress for risk of being publicly recognised as a sex worker in their own community, thus bringing dishonour to their family. They may adopt a “feminine gait” which helps in attracting clients. Younger khotkis who aspire to a khusra identity with its freedom to express feminine features more openly are held back by family obligation and their (seemingly inaccurate) perception that khusras have sacrificed and severed all ties with families. Still living at home, khotkis have less scope to lie about employment and sources of income, and most have other jobs providing “legitimate” income (not considered as impure) to contribute to the family’s expenses. More resentment was expressed about the economic transfers to gurus, with some opting for the less secure option of meeting clients outside the daira to avoid payment. Penalties may range from fines to extreme punishment by refusal of access to the daira and by instigating a social boycott by all members of the community. Some khotkis had the intention to give up sex work once they get married, with worries expressed about their guru’s response in terms of loss of income. For both khotkis and khusras, gurus act as an influential protective figure when dealing with those in authority, especially the police. The guru may pay monthly “protection” money and/or offer his chailas for free sex to police who threaten or “create problems”.

In contrast to khusras and khotkis, banthas do not seem to unite or operate in networks and, indeed, they do not seem to have a sexual identity issue that excludes them from the main society. They do, however, hide the fact they have sex with, and sell sex to, other men. Since only two of the MSW PRs were banthas (and one of them did not finish the field work), our knowledge was gained through the eyes of khotkis and khusras who expressed resentment and judgement about banthas selling sex. Since banthas are viewed as “complete” men, they can have access to any job and lead a normal life, unlike the feminised men and transgenders who experience discrimination. For khusras and khotkis, providing passive sex (being penetrated) is seen as a core expression of femininity which they cannot help, and active sex is considered as male. “Men” (with male soul) who penetrate other men is acceptable, but “men” getting penetrated was seen as offensive. Banthas were known to take both roles. Besides moral objections against men taking passive roles, banthas were also seen as competition intercepting clients on their way to a daira.

Blurring distinctions: squashed between social and group norms

Their sexual identity, life styles and occupation as sex workers make khusras and khotkis vulnerable to violence and abuse from society, clients and police, as reported by Mayhew et al elsewhere in this issue.15 However, the internal social networks among khusras and khotkis with guru-chaila and peer-based relationships serve as strong support structures. While these parallel social structures provide emotional and material support, they also come with distinct social norms which are based on expected “feminine” behaviours in line with the shared identity. This includes norms that khusras do not get married, do not “father” children and do not have active penetrative sex with men or women. This “consensus” view portrayed during the training workshops was later contradicted in private one-to-one interviews following the revelation that one PR had not attended all interviews with the research supervisor because s/he had gone home to get married (for the second time). He described his new wife as beautiful and told us he really wanted children with her. Other PRs also revealed khusras providing financial support to wife and children as well as their parents. This social reality of a double life remained unspoken in front of other khusras, as any reference to procreation or sex with women was censored again in the subsequent analysis workshop at the end of the research process. It is clear that khusras and khotkis find themselves transgressing both social and group norms, with the risk of losing support and isolation from both communities.

DISCUSSION

From an epidemiological perspective based on the common risk factor of engaging in anal sex with multiple partners, the three categories of sex workers could be seen as one group. From a sociological perspective they are very different groups. We demonstrate that, by maintaining the transgender/male dichotomy in current intervention and behavioural surveillance, an important group of sex workers will be left out since they are more hidden, less obviously visually recognised as selling sex and less easily contacted through peer networks. We gleaned this insight through observation of interaction between PRs followed by further probing. Yet with only two banthas PRs, we know too little through this peer ethnography alone to fully describe the group. However, by recruiting a separate category of banthas for the subsequent bio-behavioural survey, we show that they are important in numbers and have a different risk profile.14

A picture of complex social realities for khusras and khotkis has emerged from these data, showing fluidity in both behaviour and identity according to the situational context. Qualitative research in health behaviour has been criticised for failing to discriminate between normative statements and actual practices and for paying insufficient attention to the context within which respondents’ statements are produced.16 17 This results in accounts that may give strategic representations of the self,18 and actual behaviour is often less discriminative than portrayed in verbal accounts.19 These fears are especially valid in research with highly stigmatised groups engaging in illegal behaviour, and PRs were clearly drawing on a range of normative notions. We found the informal observational components during the training and analysis workshop (inherent in the peer ethnography method) crucial for the correct interpretation of the data.

As illustrated by the example of the incident during the training workshop, the rapport developed with the PRs is of a very different quality from that achieved through in-depth interviews or focus group discussions. More than gaining trust by putting respondents at ease, rapport evolved in a process of establishing unspoken rules. By dropping contentious areas of interest (sex with women), we shifted ownership of the probe list towards the PRs with the research team giving up some control. We did not, in fact, lose data; the way in which any discussion on sex with women was silenced showed how strict group norms are, with little scope to admit to any divergent views or behaviour in front of peers. In private, PRs more readily discussed the “traditional” male roles performed by khusras, especially as providers to their family including wife and children. This shows how participant accounts of behaviour and relationships are shaped by the research contexts including who interviews them, at what stage of familiarity in the research process and, most importantly, who may overhear the conversation. Observations from the informal interactions and group dynamics were more powerful than verbal accounts in highlighting the normative character of statements produced. This calls for reconsidering what constitutes “data”, going beyond verbal accounts to include the observation and analysis of incidents of discordance and conflict.20

Since the peer method relies on interviews conducted in the third person, personal accounts are deliberately avoided and the method relies heavily on stories or gossip circulating among the target group. However, when talking about “what friends of friends think or do”, PRs do also talk about themselves and express their own views. Like other methods based on “hearsay evidence” (second-hand and filtered by someone else who may add meaning to the story),21 these data serve for exploring how PRs construct reality when probed by a social researcher. Since we are not seeking social “truths” but a rich “dynamic social commentary”, stories are ideal to illustrate the normative framework of rules with which communities live out their social lives,13 usually presented from the point of view of the ruling interest.18

How do we place faith in the validity of our findings? Validity is about improved understanding rather than improved accuracy or agreement among different sources,22 23 and the inherent “triangulation” in peer ethnography is through use of different methods (one-to-one interviews, observation of group interactions), and through involving different researchers including peer researchers and (Pakistani and non-Pakistani) social scientists in the analysis and interpretation.

In this study we can also triangulate with the survey findings.14 In constructing our categorisations of khusra, khotki and banthas as three distinct groups with separate identities, we were constantly challenged by the fluidity of the supposedly distinguishing features. In designing the survey instrument, we could include some characteristics (whom they lived with, whether taking an active role in sex, sex with women, etc) that seemed to differentiate the three groups. Our understanding of sensitivities guided the sequence of some of the questions (eg, questions on marriage, children and female partners were asked at the end). While the qualitative data forewarn a degree of normative responses in terms of denying both sex with women and anal penetrative sex among khusras, 10% of survey respondents reported being married.14 Some khusras reported living at home with native family (10%) and some khotkis lived in dairas (13%),14 corroborating the observed fluidity. If the qualitative research had been limited to a few focus group discussions to improve the language of the survey instrument, we would have incorporated misleading generalisations, imposing a false clarity without the ability to qualify these by documenting the everyday reality.

There were certainly limitations to our study; for example, the selection of PRs. Recruitment went ultimately via the gatekeepers with a guru as one of the khusra PRs. Hierarchies and divisions exist within every peer group and may not only be hard to predict, but also hard to cater for. We do not know how different the data may have been if the guru had not been around, although we doubt we could have had any data without his consent. We certainly reaped the benefits of his presence and power in mobilising the khusras to attend the workshops. Through his central role, we learned that the relationships with the gurus are the link through which transgender culture gets transmitted from generation to generation. The interconnectedness of the entire network seemed to support the observance of khusra cultural rules. Going beyond what is possible in an FGD, the peer ethnography process at least allowed us to learn some of the rules, although it can be no substitute for an ethnography that observes what happens to members who get excluded from these communities.6 Another limitation is that it is hard to generalise the social dynamics beyond the city in which we did the research. Indeed, in smaller cities there is bound to be less anonymity and less opportunity for networks of alternative sexualities to form and, indeed, the survey showed a different distribution in the second smaller city with fewer khusras and khotkis sampled.14 Finally, because this was a formative research study, we feel we have not capitalised on the participatory dialogue initiated with the members of the target community to translate this immediately into interventions. When interventions are targeted at these groups in future, peer research for detailed intervention design should be used to increase relevance and impact.