Most of the methods in Table 4.3 involve collecting data through interviews. The unstructured–structured continuum is a useful way to organize the diversity of interview methods. Bernard (2018) identifies three types of interviews: unstructured, semistructured, and structured. Each type of interview, in turn, includes a diverse array of techniques. Structured interviews, for example, are used in survey research, in the collection of social network data, and in tandem with formal elicitation techniques such as free-listing and pile-sorting. Semistructured interviews and focus group discussions are likewise similar in level of structure and purpose.
Many medical anthropologists recognize the complementary value of different methods and often combine them in a single study. For example, Singer et al. (2006) designed a five-year study on the prevention of sexually transmitted infections and unwanted pregnancies among low-income, inner-city African-American and Puerto Rican youth in Philadelphia and Hartford. They used a wide range of methods: focus groups, formal elicitation (e.g., free-listing), in-depth individual sexual and romantic life histories, sexual behavior diaries, and structured interviews. This strategy paid off, because different methods yielded different insights. Focus groups helped to identify the range of relevant sexual behaviors and relationship types that people recognized, while in-depth individual interviews revealed the personal, emotional meaning of particular experiences. These complementary findings illustrate the benefit of creating redundancy, or triangulation, by using different types of methods (LeCompte and Schensul 2010, p. 174).
Data Analysis
Some of the methods in Table 4.3 generate text or images; others generate numerical data. There are a wide range of qualitative and quantitative methods for analyzing both types of data. The suggestions for further reading in Table 4.3 provide guidance on appropriate analytic methods for specific types of data. For a comprehensive approach to the analysis of qualitative data, see Bernard et al. (2017). For a guide to quantification in ethnographic research and the “five things you need to know about statistics,” see Dressler (2016). And for a practical guide to the systematic analysis of qualitative and quantitative data, see Dengah et al. (2021).
One example will illustrate the value of having a flexible toolkit of analytic methods. Yoder (1995) explored how mothers of small children in Lubumbashi, Zaire, diagnosed and treated childhood diarrheal diseases. The data he collected were typical of what many medical anthropologists collect. He began with unstructured, open-ended interviews with small groups of mothers, initially attempting to get as complete a list as possible of all the childhood illnesses mothers knew. Later, he probed for detail about symptoms and treatments and eventually identified six illnesses related to the biomedical category of diarrhea: kuhara, kilonda ntumbo, lukunga, kasumbi, buse, and kantembele. Last, he asked new groups of mothers specifically about symptoms, causes, and treatments associated with these six illnesses. Based on qualitative analysis of mothers’ descriptions, Yoder (1995) concluded that mothers’ diagnoses were based on the perception of symptoms and that ethnomedical classification shaped treatment decisions.
Because Yoder published tables of raw data from the mothers’ descriptions, he and collaborators were later able to extend the analysis using other methods. Ryan et al. (2000) conceptualized Yoder’s data as a series of linked lists. Yoder had collected 106 descriptions of the six illnesses. In each description, women discussed the symptoms, causes, and treatments associated with the illness. Ryan et al. converted women’s descriptions into a set of matrices that indicated whether a particular symptom, cause, or treatment was mentioned in each illness description.
Figure 4.5 illustrates the steps in analyzing the relationship between illness and symptoms. The list on the left indicates which symptoms were associated with each description. For example, the women in the first group said that kuhara was associated with frequent stools, vomiting, fever, and no appetite. The matrix on the upper-right converts the textual data into a series of ones (if a symptom was mentioned) and zeros (if it was not). This matrix contains all the information in women’s original descriptions about which symptoms go with which illnesses – although the table of ones and zeros hardly clarifies anything on its on. The trick is that, once the data are in this format, Ryan et al. could use correspondence analysis (Weller and Romney 1990) to visualize the relationships between illness descriptions and symptoms.
Figure 4.5 Steps in analyzing explanatory models of illness, based on Yoder (1995) and Ryan et al. (2000).
In the graph in Figure 4.5, each solid circle represents a symptom; the other symbols represent individual illness descriptions by groups of women. The closer any two points appear together, the more strongly they are associated with each other. Thus, sunken palate and tongue clacking were often mentioned together, and both were associated with descriptions of lukunga. “Frequent stools” are associated with kilondo ntumbo and kasumbi, but women were more likely to mention “very frequent stools” in descriptions of kuhara and buse. The large ovals are 95% confidence intervals that reflect how much the groups of women agreed with one another about which symptoms go with which illnesses.
This analysis is no substitute for Yoder’s original ethnography, of course. But it does add new insight about the level of intracultural variation, the coherence and boundaries of illness categories, and the amount of overlap in symptoms associated with different illnesses. Sibley et al. (2007) recently used the same approach to analyze women’s descriptions of postpartum health problems in Bangladesh. In both cases, the transformation of words into numbers, numbers into pictures, and pictures back into words reminds us of how counterproductive it is to divvy research methods into either qualitative or quantitative.
CONCLUSION: LOOKING AHEAD
Medical anthropologists draw on methods from across the social and health sciences, but they are not only consumers; many are also at the leading edge of developing new methods relevant to interdisciplinary research on health and healing. Some of the most important advances in the last decade include:
participatory, collaborative, and action-oriented models of research (Schensul et al. 2015)
methods for rapid ethnographic research (Sangaramoorthy and Kroeger 2020), as well as for slowing it down (Pigg 2013)
a new measurement model that explicitly incorporates cultural meaning into survey measurement (Dressler et al. 2005; Dressler 2020)
new tools for analyzing the structure and composition of social networks (Borgatti et al. 2018)
new approaches to the study of inter- and intracultural variation in knowledge, beliefs, and practices (Dressler 2018; Dressler et al. 2015; Hruschka et al. 2008)
biocultural strategies for measuring stress-related outcomes (Brewis et al. 2022)
advances in how the choice of data collection method (e.g., face-to-face or online, individual interview versus focus group) influences data quality (Gravlee et al. 2018; Wutich et al. 2010)
research strategies and techniques for cross-cultural, ethnographic research (Wutich and Brewis 2019)
innovative methods for studying water insecurity across cultural, geographic, and demographic contexts (Wutich et al. 2017; Wutich et al. 2019)