Ethnography: where the idea comes from and when we started to recognize the work of production

‘An ethnographic account is conventionally the description of a particular society and culture known to be based at some point on the experiences and observations of a fieldworker who was there’.14 This is how Strathern starts her reflections about anthropological production15 and the many dissonant voices that debate the representation of ‘society’ and ‘culture’ through ethnography. The discussion turns on the understanding that there is no ethnography without the ethnographer. As logical as this may seem, the implications of this insight became the object of debate some decades ago.16 Of course, the presence of the ethnographer had been acknowledged earlier, but the effect of this presence on the very idea of ‘culture’ and ‘society’ became the main discussion. The ethnographer is a ‘medium that already has a form of its own’,17 and the idea of ethnographic account as mere description or transparent representation of a given society or culture came under attack.18 Rather than a neutral vessel through which culture or society can be transported, the work of the ethnographer is the very condition of existence of society and culture.

Let us explain. While in the field, the ethnographer interacts with study participants, observes the relationships between them, and learns local meanings and rules. (S)he accumulates a diversity of experiences in situ.19 However, the interactions between researcher and study participants are not replicated one-to-one in the ethnographic report. Rather, as Strathern points out, it is the generalised information about these experiences—what is common to many of them—that can be aggregated as the ‘culture’ of a group of people.20 In order to produce a coherent text, the ethnographer connects the events in which (s)he participated into an organizing entity. This entity goes by the name ‘society’ or ‘culture’21. At the conclusion of the study, the ethnographer writes a report, which is a holistic, unifying assumption about the integration of meaning regarding his/her experiences in the field.22 This reflection has led authors to conclude that not only is there no ethnography without the ethnographer but also that there is no culture or society without the ethnographer’s unifying and interpretative presence. The work of the ethnographer is already contained within the origin of a given culture or society, as a condition of its existence.

We have borrowed past anthropological discussions about the relationship between text— the ethnographic account—and the notion of society or culture in order to help us reflect on the relationship between our texts—the articles we have studied—and the notion of ‘knowledge’ or even ‘evidence’ in health research. To fuse experiences from the field into a description of an overarching entity resembles the underlying aim in traditional ways of reviewing scientific papers. Commonly, reviews provide descriptions of scientific knowledge published somewhere else—what is already known about a certain theme or phenomenon—in order to provide a foundation for further research or practice.23 In ethnographic research, the idea of culture or society is no longer seen as a mere description of a reality out there. Rather, ethnographic work is the condition for the existence of culture or society. Once this ethnographic rationale is applied to explore scientific publications and these publications acquire the status of study participants, this exploration can no longer claim to be a mere representation of the already established knowledge about a topic, such as health and female homosexuality in Brazil, or to present a unified whole about it. The debate about representation has urged alternative ways of carrying out research, which understand ethnography as means to interfere with the world, ‘not a way of opening a window on the world’.24

It is important to emphasise that anthropology is not, of course, claiming that culture and society do not exist out there. Neither do we argue that our object of study—the knowledge about same-sex attracted women and health in Brazil ‘out there’—does not exist. The debate we engage with here revolves around how research mediates between the world and various accounts of it; how research interferes with participants; how the ‘culture’ in the ethnographic account is dependent on the work carried out by the ethnographer. Rather than understanding ethnography as a tool to access reality, we use it as a medium to interfere with reality.

We draw on ethnography here to state the obvious: the information about health and female homosexuality we provide in this review is the outcome of manufacturing work. Each article was selected, read, re-read, translated from Portuguese to English when necessary, reflected on and written about. There is no literature review without the work of reviewing.

Enlisting participants and draft one

In Table 1, we have listed 14 papers that took active part in the production of this review article. Along the way, we encountered many other texts (such as reports, doctoral and master theses, unpublished material, news reports, and governmental publications) that also contributed to the production of our text (and to the writings listed in Table 1). Just as the ethnographer does not replicate all field conversations in an ethnography, nor accounts for the relationship with each individual informant,25 we also avoided trying to do so. Thus, the list in Table 1 names the participants we came to know the best, not all the informants that contributed to our perspectives.

We came across some of the review participants through database searches for articles related to our main topic of interest. As we expected the number of articles to be small, we aimed to gather all publications related to this topic. We used terms such as ‘lesbian women’, ‘lesbians’, ‘women who have sex with women’ (or the acronym WSW), ‘same-sex attracted women’ and ‘female homosexuality’. We searched for terms both in English and Portuguese. All participating articles focused on health-related issues in Brazil. In a second moment, we used a purposive snowball sampling strategy to include additional review participants among papers quoted or mentioned in the initial sample. Our aim was to select articles that would contribute to a wide variety of perspectives and articles where significant aspects of the phenomena under study were most pronounced.26

In Table 1, we have listed the review participants by publication date (their birth date, as it were) and included some background information about them that we consider contextually relevant. ‘Name of author(s)’ and ‘publication year’ are provided so that we can identify the papers in our text. ‘Title’ reflects the authors’ take on the themes discussed in the articles. When an English title was not available, we decided to keep the original Portuguese wording in the table and provided the reader with our English translation of it. ‘Journal name’ and ‘language’ give an idea of the accessibility and potential scope of influence of each paper, and the number of pages offers an impression of the length of each paper. ‘Approach’ allude to the ways or settings in which knowledge was produced.

All articles were printed on paper and read several times. As one of the authors is Brazilian and reads Portuguese, it was not necessary to translate all papers in order to analyse them. Some parts of the texts were highlighted with text marking pens, and we hand-wrote notes on some of the printouts. Information contained in the texts was written down in order to generate our first set of ‘field notes’.

We thereafter developed the first draft of this article (Draft One), which contained a description of the papers and identified themes that seemed to be common among them. Special attention was paid to cross-references between the articles. An early observation was that all the articles framed health issues mainly in relation to sexual health. We also noticed that the terminology used to group the women together differed between articles. Among terms in use were ‘lesbians’, ‘lesbian women’, ‘women who have sex with women’, ‘WSW’ and several other terms. We decided to always borrow the study participant’s own terms, when referring to their statements.

Draft two: articles as participants

At this point, the task of relating to scientific and scholarly papers as study participants demanded a reflection about theoretical foundations for this move. A second draft of this article was experimented with as we engaged with this task. It aimed to focus on the part articles were playing in the creation of the review.

Theoretical insight from Actor-Network Theory (ANT)27 motivated an emphasis on the papers’ ways of acting. ANT suggests that both human and non-human actors participate in the production of events and phenomena studied in social world.28 However, rather than being a pure and unproblematic source of action, ANT understands an actor as the target of a variety of actions directed towards it.29 ‘An actor is what is made to act’, not the origin of an action.30 This perspective was helpful because it promoted a shift in our way of observing the articles and the relationships between them. We started to consider anything capable of acting: the articles themselves, the authors and the findings; they were all doing something.

We then turned our attention to what was making them act. We refer to this as the review participants ‘surroundings’ or “environment’— even as their ‘context’. However, with regard to the latter of these terms, it is important to emphasise something. The word ‘context’ is often associated with the words ‘cultural’, ‘economic’ or ‘sociopolitical’, as a general dimension that explains actions. Our notion of ‘context’ was rather the opposite. It was very narrow and uncertain. We took the actor as a starting point to observe and describe ‘context’—the associations that were enabling the review participants’ action. The ‘context’ was the direct relationship that made their actions possible: the conditions necessary for action to occur. We did not know what these ‘contexts’ might be at the onset of the study, but these gradually became a main focus of our attention. We recognised that the act of publishing an article was influenced by surrounding conditions, as for instance claims made in other papers, which in turn deserved investigation.

The paper authored by Reis et al 31 was crucial in this respect. It is a case report. In health research, case reports have the function of exploring singular events, investigating events in depth, or raising the importance of something.32 Reis’ report raises the issue that there is a chance for sexual transmission of disease between women:

Both women denied past history of STDs [sexually transmitted diseases] and heterosexual intercourse; but they did declare previous relationships with digital-vaginal and digital-anal practices, as well as vulvar and oral-to-oral contact. Biopsy results for lesions of the lips and vulvas revealed condyloma acuminata.33

Condylomata acuminata is one of the conditions associated with the human papillomavirus (HPV), a virus connected to several sexually transmitted infections, including genital and oral warts.34 In addition, HPV is described as having oncogenic potential.35 The paper by Reis et al contains four photographs: one showing multiple and confluent warts on a vulva, one depicting gum warts and two presenting microscopic images of these lesions.36 The patient in this case referred to herself as lesbian. In the text, she is said to represent the WSW category (women who have sex with women). It is clearly important that she declared that she had been in a stable relationship with a woman for the 16 months preceding the study and that both she and her partner claimed to never have had sex with men. The article states that the woman was a ‘WSW with intact hymen’,37 thus emphasising that professional observation was consistent with the woman’s assertion that she had never had heterosexual intercourse. This emphasis is made because the article draws attention to a significant finding: the woman had contracted a sexually transmitted disease although she had only had sex with other women. By emphasising this association, this case report becomes evidence that STDs may be sexually transmitted between women.

The perspective that an actor is made to act led us to explore the surrounding conditions that allowed for the publication of Reis et al. We noticed that their case report would only be worth writing and publishing in surroundings where sexual transmission of diseases between women was considered an extraordinary event. This perspective turned our observation to these surroundings: the other papers we were reading. They did indeed state that sexual transmission between women was something unknown. We inferred that the ‘context’ necessary for the publication of Reis et al was made up of this very claim.

Bertolin et al 38 is an example of this ‘context’. This study evaluates the knowledge women who have sex with other women have about HPV and states that women do not know that they can acquire HPV by having sex with other women. Here we started to realise that the principle of observing the condition of existence of statements and claims in our papers, and the conditions allowing their publication, as we had done with Reis et al, was useful. It provided us with a way of connecting the articles to each other without having to summarise their information.

We also observed for the first time the relationship between claims in the papers and the general knowledge about their ‘findings’. In addition to the connections between the articles we could now see, this rather unusual way of working had provided us with a way to move forward.

Sex with women, sex with men

In recognising the participating articles as actors made to act, we could observe the ways in which they influenced each other. Instead of asking what a given text was saying, we started to ask why it had been published in the first place. We asked what were the surrounding conditions that allowed the paper’s claims to have meaning.

A first and obvious way of seeking understanding of what motivates the publication of an article, or the claims it makes, is to pay attention to the reasons stated by its authors. Articles usually base their arguments in other articles. This is habitually done by presenting a quote, inserting a name (or a number) after it, and making a list of these names under the heading ‘references’.39 Like we did just now. However, we also noticed that many arguments justifying the publication of articles had no references. In scientific articles, the practice of referencing seems to be associated with the confidence in the claim made.40 When an argument is contested, or when there is doubt around it, the practice of referencing seems to enhance the factuality of the claim. Conversely, in the case a paper makes a claim that requires no discussion, there is less of a need to reference it. In the following (translated) quote for instance, Barbosa and Koyama did not provide any citation.

The emergence of the epidemics of AIDS exposes a growing preoccupation with sexuality, particularly issues related to male homosexuality. Yet, female homosexuality and its relationship with health was kept marginal to this process.41

In attending to the associations between claim making and the practice of referencing, we also observed that absences—that is, a lack of scientific knowledge or publications about a topic—were often not documented. Often, when an absence was claimed, articles had no way of referencing it. Pinto et al provides an example:

The frequencies of different sexually transmitted diseases (STD) and risk factors related to gynecological cancer among women who have sex with women (WSW) in Brazil is little known.42

It is interesting to note that the motivation for the study carried out by Pinto et al is a claim in its own. The references presented throughout the paper relate to studies about the frequency of STDs abroad. Pinto et al claimed that the topic they were investigating was ‘little known’ in Brazil. This claim enabled the study to justify itself. The absence of previous studies created a medium—a context—that permitted tests to be carried out among Brazilian same-sex attracted women in order to detect and quantify the presence of sexually transmitted infections among them. In the Pinto paper, these tests had been translated into frequencies of trichomoniasis, bacterial vaginosis, fungal infections, chlamydia, hepatitis, abnormal Pap smears, HPV infections and HIV. The authors had also carried out a questionnaire survey in order ‘to identify data relating to possible risk factors for STD/HIV’.43 Risk factors were translated into: having sex with men in addition to having sex with women, not using condoms, sharing sex toys and having sexual intercourse while menstruating.44 All these activities and practices were said to be related to the possibility of being infected. Pinto et al claimed that something that had been happening in other countries, happened in Brazil as well.45 Women who had sex with women may have sex with men too and could thus get infected with STDs.

When comparing Pinto et al with Reis et al, we observed that there was a time lapse between the two publications. Also, while Pinto examined the occurrence of sex with men among the same-sex practicing women they studied, and linked this to a risk for STDs, Reis suggested that even sex between women could bring about transmission of such diseases. It became clear for us that a connecting element existed between the two papers: for both review participants, the possibility of infection was framing the work of research and publication. We noted that this argument was still controversial, because whenever the possibility of infection was mentioned, there was a need to provide references.

Yet, we arrived at an even more captivating inference in draft two. There was a strong association between the new ‘evidence of infection’, and the assumption that a lack of knowledge about this possibility exists. We observed that only Bertolin et al engages in evaluating knowledge about sexually transmitted infections. In the other articles, the possibility of infection as new established fact becomes the foundation to claim that very little is known about it. In the following, we have listed some examples of how this occurred in the papers:

• In Reis et al: ‘This case report clearly illustrates the need for specific orientation about HPV infection targeted at women who have sex with women (WSW)’.46

• In Pinto et al: ‘It is evident that there is a need for healthcare professionals to be correctly informed and sensitive toward the healthcare of WSW’.47

• In Barbosa and Facchini (translated): ‘The lack of knowledge in Brazil about the health needs of this population is evident’.48

• In Mora and Monteiro: ‘The notion of STI and HIV risk is not well developed among women who have sex with women’.49

From this understanding, we identified a thin grid holding our review participants together: the work of publishing was, in one way or another, connected to the possibility of infection among same-sex attracted women plus a shortage of information about that very possibility. Figure 1 illustrates these connections. We have placed inside circle A the different activities the papers claim make infection possible. For our review participants, these activities worked as means through which the ‘possibility of infection’ comes to exist. Circle B contains the different persons and settings where the ‘need for information’ about these possibilities was allocated in the papers. We noted that we could associate one or more items in A with one or more items in B. Through this exercise, we were able to identify a basic structure in the process of making claims in the articles under review: the assertion that ‘need for information’ existed (for instance among women themselves or among healthcare providers) was dependent on the existence of the ‘possibility of infection’. In other words: the claims in circle B were dependent on the evident existence of items in circle A.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

A basic structure for claim making.

We further noticed that once a relationship between A and B had been established, the papers were free to focus their attention on the circumstances surrounding this relationship. In the publication by Barbosa and Facchini, for instance, claims are generated on the basis of an association between A and B. The paper argues that women who have sex with women have ‘difficulties in accessing healthcare’,50 specified as gynaecological care. By looking at the condition of existence of this claim, we could observe the presence of the underlying association between A and B as an established fact. Consider the following example (translated) from the paper:

Among those [interviewed women] with none, or few and irregular consultations, it was very common to refer to a lack of ‘necessity’ [of going to the doctor]. The argument included: a “lack of need” due to “not being at risk” and “not having symptoms”. 51

The logic in this quote is: because the interviewed women asserted that they did not need healthcare (they stated that they were not at risk and/or that they did not have symptoms), it affected their search for healthcare. The ‘lack of necessity’ is an argument producing ‘none or few and irregular consultations’. When we analysed the conditions of existence for the claims in the paper, we observed the existence of the association between A and B as a foundation. The article’s claim is produced based on the established understanding that women are at risk but do not know about this risk. In this scenario, where the association between A and B already exists, not searching for healthcare becomes a problem. Since the association between A and B is already in place, it allows the article to question the women’s argument (lack of necessity) and recast it (1) as women’s misconceptions and (2) as causing women not to access healthcare.

Until this point, we were pleased with our ethnographic approach and with the freedom we had gained from the principle of treating articles as study participants. However, we slowly came to recognise that in our attempt to approach articles as participants and looking at their surroundings, we had gradually introduced a bias, a sort of favouritism, in our reflections: the health of same-sex attracted women in Brazil, our main study object, was slowly evaporating from our text. Taking Barbosa and Facchini52 again as an example. As we highlighted the conditions of existence of this paper’s claims, it seemed that we had left out the claims themselves. In other words, was there really a possibility of infection? Was it a problem if lesbian women did not search for healthcare? Were we ignoring the papers’ content by only looking into their ‘context’? Our overarching question now became: What is the relationship between the claims in the papers and the reality outside of them? This question led us to start anew.

Draft three

While we had initially meant to write a single review article, we now came to consider splitting the manuscript so that it would become two different but related articles. The first would provide a theoretical foundation for the way we were investigating written texts, and the second would contain a practical application of the theory explored in the first. The possibility of writing two articles provided opportunity for an investigation of theories about translation and knowledge production and about philosophy of language. We used these theories to expand our understanding of articles as actors by exploring not only the conditions of existence of the health of Brazilian same-sex attracted women in the literature, but by investigating our study object itself and tracing the relationship between the articles and the realities outside of them.

While we first explored philosophy as a tool to understand properties of texts, we later came to recognise its practical application. One of the articles we were reading contained a feature that was particularly helpful in this respect. Almeida is an article based on an ethnographic study among lesbian activists.53 Throughout the article, the author applies the term ‘lesbian body’ (in Portuguese: corpo lésbico), which is said to encompass how the participants in the study presented their bodies and their health claims in a setting of political activism. However, the participants in the study had never used this term themselves. It was an invented term, manufactured by the author in the writing process:

The lesbian body was a fiction (…) sanctioned by the need to understand how a certain group (strategic in publicly establishing a lesbian identity) thinks itself. (…) I have tried to show how - in a certain political and cultural scenario—a group of individuals who are leading a social movement in expansion represented their bodies and their health needs. (translated)54

From philosophy of language, we have applied to Almeida’s ‘lesbian body’ two main concepts: (1) Jacques Derrida’s notion of inscription 55 and (2) the notion of intertextuality, a term first coined by Kristeva.56 While the reflections about inscription follow below because this concept played a substantial role in draft three, intertextuality only emerged in our last draft, and is thus described later in this paper.

Draft four: when we arrived at a closure

The notion of inscription has left us with an open end to close. What are the effects of understanding that the ‘lesbian body’ contains traces of events, experiments, viruses, interviews and so on? Or, in other words, what are the consequences of recognising that the findings in our participating articles contain traces of the events giving origin to them? A critical point in Derrida’s theory is that it opposes the idea that the written entity (which in our case is the ‘lesbian body’) is a storage medium, or a direct link with an outside reality.63 In other words, that the written entity is a neutral representation of an original object. If we come to terms with this philosophical argument, there is no original object to begin with. If we again take the lesbian body as example, there is no original ‘lesbian body’, only interviews between women and researcher left in the past, translated into written material.

To accept this philosophical claim is by no means to claim that interviews did not take place, or that lesbian women are not really vulnerable. (Or, if we bring up again the reflections about representation in ethnography, that the culture out there does not really exist.) It is a way to remind us that what we have at hand when reading a paper is the outcome of a series of transformations. In effect, the use of such philosophical intakes provided us with a system to take up scientific articles and recast them objectively: as mediators between that which has happened and that which continues to happen. We have used philosophical concepts to reassert the obvious: That papers are papers. The most appealing outcome in this approach, however, is the ability to recognise the participation of these objects in the manufacturing of realities. Just as in the example from anthropology (that culture is dependent on the ethnographic account), so also in our case: the vulnerability of same-sex attracted women in Brazil is dependent on, is mediated through, the work carried out by the articles we studied.

We find that the ethnographic approach we have used has multiplied the ways in which the health of same-sex attracted women in the Brazilian published literature can be grasped. Because once vulnerability came into being through the work of publishing, we could again ask what are the surroundings in which such a claim can be made. In this last and final draft, we applied the concept of intertextuality to the lesbian body, in order to take into account the associations participating once the lesbian body could be claimed as vulnerable. From intertextuality, we employ the general notion of ‘several texts within the text’,64 in which the lesbian body can be explained as an ‘intersection of textual surfaces’65 and an intersection of other texts. The following quotation is again taken from Almeida’s abstract:

The epidemic contributed to this continuing invisibility, due to strong beliefs in which the “lesbian body” would be the only one immune to the infection through the sexual via. The hypothesis guiding this paper is based on the idea that lesbians vulnerability is the “passport” for affirmation/inclusion of a certain identity mark in public policies agenda.66

The notion of intertextuality makes it possible to conclude that within the studied articles we could identify other texts, which we had no direct access to. Alongside the associations we had already described—the A and B claim making structure, and the invisible and female features—we could now recognise a political text as co-participant in the manufacturing of a vulnerable ‘lesbian body’. The notion of intertextuality conceives of the creation of a text as a reading and absorption of the anterior literary corpus, as a dialogue with other texts.67 Texts are more than written entities. They embody everything else that was surrounding the creation of the text, as for instance thought, history and political circumstances.68 Because interviews take place at a certain place and time in history, they inevitably respond to and dialogue with the environment in which they happened. Blood tests are required in a certain setting, and produced in relation to this setting. In the case of this review, the setting was a political one.

While identifying that political texts participate in scientific projects may not represent a fresh insight, intertextuality has provided us with a way to trace the political action of scientific publications. By applying both the notion of inscription and the concept of intertextuality to the ‘lesbian body’, we could make a final claim: in the papers we studied, vulnerabilities and the findings endorsing them are the outcome of a process that involves the transformation of events into scientific publications with political implications.

This is not to say that the scientific is coming about only to fulfil sheer political aims. In our review, it means to recognise that a lot more than interviews, tests and scientific publication take part in establishing that same-sex attracted women are vulnerable. The notion of intertextuality blurred the separation between science and politics. The political does not precede the scientific. Political subjectivities are necessary parts in the objective scientific endeavour. Intertextuality enabled a transformation of the line dividing these two arenas into a juncture. A scientific vulnerability became political. The practice of publishing evidence is a political practice.