Making Methodological Decisions

Deciding which qualitative research tools to choose requires consideration of the research paradigms or approaches that guide or frame a study. It is also important to determine whether a study requires experiential understanding of the setting or community context within which the people who are the focus of the study live, or whether it requires researchers to learn from respondents their point of view on a topic. Even if direct experience in the study site is not part of the study design, understanding something about the cultural setting is critical in considering which tools of inquiry are likely to be most appropriate and when. For example, it is important to know enough about a setting to determine what procedures must be set into place in order to conduct a confidential in-depth interview on a sensitive subject, such as HIV risk behaviors. This can be a challenge in densely inhabited residential areas where there is little space and where neighbors may be curious, as in many low-income urban areas of India. Or, for example, deciding where, when, and how to interview patients in an emergency room setting calls for the acquisition of considerable information about the emergency room beforehand.

In addition to learning about the study context, researchers should consider the following: (1) which guiding research paradigm or paradigms they follow; (2) what their research questions and subquestions are in the initial stages of research; (3) what type of formative conceptual model they can develop with the information they have; (4) where, when, and with whom they will conduct the study; and (5) what their sampling plans are. They also should consider what methods and tools they will use to collect their data. These topics are considered in more depth in the following sections.

Guiding Paradigms

Qualitative research can be conducted within the framework of a number of different guiding paradigms. The term paradigm derives from the work of Thomas Kuhn (1970), who suggested that scientists are influenced by dominant ways of or frameworks for conducting science. In the social sciences we speak of several different influential views. The most commonly referenced are the positivist, interpretivist, critical, and participatory paradigms. Positivists believe that reality is external to the self, that it can be observed, and that the tools used in the conduct of research can produce information that is reproducible and potentially replicable if collected under similar circumstances. Positivists generally believe that researchers are observers and should minimize their interactions with and effects on the subject matter of the research while they are gathering data. Researchers who take a positivist position often prefer to test preexisting theories rather than to derive them inductively from the study situation.

Interpretivists take the position that social or cultural phenomena emerge from the ways in which actors in a setting construct meaning. The researcher comes to understand behaviors and the meanings attributed to them through immersion in the setting and interaction with the study participants. The earlier case example illustrating design decisions for the study of depression among older low-income adults of diverse ethnic and racial backgrounds included an interpretivist qualitative component examining the language and meanings associated with sadness, loss, stress, and life dissatisfaction in this population.

Mixed-paradigm research combines interpretivist and positivist approaches, highlighting the voices and views of the participants, in interaction with the results and interpretations of the researchers. Research on the use of the drug MDMA (Ecstasy) carried out in Hartford, Connecticut, with 120 young adult Ecstasy users provides a good example of this complementarity (Singer & Schensul, in press). The researchers explored the meanings and rituals these young adults attributed to the use of Ecstasy through in-depth interviews, and collected survey data on drug use, locations where drugs were used and bought, and other self-report data. More than half of the respondents maintained that, by balancing the risks of use with risk mediation, they were able to benefit from the use of Ecstasy and were in good control of their Ecstasy use. But their perspective did not include their other drug use. A separate analysis from the researchers' perspective showed variation in control based on the amount and frequency of Ecstasy use in conjunction with the use of other drugs. This analysis suggested that when Ecstasy users combined this drug with other drugs, their overall control of their use of substances, including Ecstasy, declined.

Critical researchers believe that social and political structures shape and hold power over the lives of individuals, creating various types of disparities. Critical researchers always locate the behaviors and meanings held by individuals and groups within larger systems of dominance and control. Although they may focus primarily on the structural determinants of disparities in their research, they always are concerned with how these factors differentially affect people living in communities by reinforcing and replicating benefits for those with more resources and reducing benefits for those with fewer. Critical researchers often base their work on secondary sources and historical reconstruction, using maps and graphs to illustrate structures of dominance; they also rely on in-depth interviews to reveal ways in which dominance and persistent inequities are transferred to the behaviors, opportunity structures, and meaning systems of vulnerable populations.

To illustrate, in a study of the role of temporary housing in the lives of injection drug users, researcher Julia Dickson-Gomez and her colleagues (Dickson-Gomez, Convey, Hilario, Corbett, & Weeks, 2007) were able to show that despite somewhat more progressive policies guiding permanent housing options for injectors, agency staffs that had the responsibility for helping drug users find permanent housing did not do so, in part because they did not believe that drug users could maintain a more permanent housing situation and benefit from social services. As a consequence, many drug-using men, for whom the chances of quitting drug use and getting a job were enhanced by permanent housing plus services, were deprived of this option. These men reported a variety of problems and challenges in finding housing other than temporary shelters. Their continued association with other drug users made it difficult or impossible for them to leave drug use behind and find a job.

The positivist position coincides with the critical paradigm insofar as each conducts observations through the lens of externally developed theoretical frameworks and each calls for the collection of data to demonstrate the veracity of hypotheses derived from the theory.

Researchers who take a participatory or collaborative approach join forces with stakeholders to conduct research toward some form of social action (see Chapter Eighteen as an example). Stakeholders may include community residents concerned with making positive changes in their community, state and local agencies, community organizations involved in the study problem or topic, and researchers. This approach acknowledges researcher expertise in scientific methodology, and at the same time engages stakeholders in contributing their knowledge and experience and decision making to research design, data collection, analysis, interpretation, and use of findings (Berg & Schensul, 2004; Minkler & Wallerstein, 2003). Participatory or collaborative research requires bringing positivist, interpretivist, and critical research approaches and tools to bear on issues and challenges presented by partners. The interaction of these frameworks or paradigms in actual field research results in improvements in theory, research methodology, and research tools, as well as interpretations and results that have local or partnership meaning. Partners can use these jointly forged results effectively to move toward desired social change. Social scientific research and social settings are complex, and a single research paradigm or approach may not be able to answer research questions or fulfill collaborator needs. Researchers should be aware that several research paradigms can be combined to guide a study, and many if not most research tools can be adapted for use within each of these paradigms.

Defining the Research Questions

Qualitative researchers are most likely to raise research questions based on three factors: (1) what has personal meaning to them; (2) what they read and discover to be gaps in the literature; and (3) what they perceive during their first exposure to the field, or the study setting. Researchers usually frame their questions based on the research paradigm with which they feel most comfortable. For example, a qualitative researcher who favors a positivist perspective will build theory based on scientific literature, frame research questions from the literature, and structure data gathering to prove the theory. The researcher then gathers qualitative data to support the theory.

An interpretivist may identify research questions based on discussions with participants in the field about the meaning of an activity, ritual, artifact, or series of events. For example, researchers Stephen Schensul and colleagues (Kostick et al., 2010) found that the primary reproductive health complaint of women in Mumbai was white discharge (referred to in Hindi as safed pani). This complaint had no basis in infection or a medical problem. So the researchers undertook to learn from women what meanings they associated with the concept of safed pani, why they thought it was a health problem, and to what they attributed it in their lives. In-depth interviews with forty women revealed that they associated it with various sources of tension or conflict in their lives, including insufficient income, negative marital relationships, and abuse.

A critical researcher is more likely to raise questions about the factors that contribute to health disparities in minority communities, or how people experiencing disparities mobilize to address them. Young researchers in Hartford, for example, explored structural factors contributing to racism. They identified media, education, and economic disparities, and collected in-depth video-recorded interviews on those factors from other youth and adults (Mosher, in press).

Regardless of the researcher's choice of dominant perspective or paradigm, however, qualitative research questions tend to focus on explorations of behavior or social organization, of the many factors that might contribute to these, and of their meaning to the study population. Historical influence is also important in qualitative research, contributing to questions about how patterns of culture, social structure, beliefs, and behaviors came into existence; what historical factors might have contributed to current injustices in policy and practice; and whether things change over time.

Formulating a Conceptual Model

If research methodology or design constitutes a roadmap for the study as a whole, a conceptual model constitutes a theoretical roadmap. In many fields (psychology, sociology, economics), research begins with a theoretical model to be tested. Here the instruments are chosen in advance in relation to the components in the theoretical model. In qualitative research a conceptual model is a theoretical starting point, a pictorial map of the conceptual direction of the study. The conceptual model is a diagram that identifies the primary research domains that are likely to be addressed in the study based on initial assessment and the literature. A domain is a broad area of culture that a researcher considers to be important in the study.

For example, in a study of adolescent drug use, the starting point might be the hunches that youth involved in drug use learn through watching others and that the contexts in which they are exposed to drug use make a difference in what they use and why. Here the domains are involvement in drug use, learning to use drugs, and family history of drug use. Qualitative researchers would investigate and unpack each of these domains. For example, in-depth interviews might reveal information about current drug use, including the different drugs people use; how they use those drugs; how much they use; where they obtain the drugs; and the reasons why they use them, and with whom. The domain learning to use drugs would include any information about how a person started learning to use each of the drugs in his or her repertoire. This could include such subdomains as watching others use that drug, seeing close friends or parents use the drug, being taught by someone how to use the drug, anticipating effects of the drug, and first experiencing using the drug. Family history of drug use might include a history of drug use among family members and household members, involving what drugs were used, how they were used, and whether household patterns of drug use involved children—including the participant reporting the information. An initial working model would link these domains and their subcomponents. Thus different components of family history of drug use and learning to use drugs might be connected to a person's current drug use.

Models often are portrayed as diagrams. Domains are arranged as predictor and outcome domains (for example, learning by watching friends results in more drug use than trying drugs alone). Here drug use is the outcome domain; drug use may be “operationalized” or described in many ways. Learning is a predictor domain, which can include all the ways that people learn or become socialized into learning how to use drugs. These models are not quantitative tests of association. They are ways of illustrating patterns of interaction among domains and subdomains or qualitative variables. Domain associations can also be portrayed as causal chains or flow charts stated as hypotheses concerning how one domain may lead to another. For example, domain A (regulations combining regular and charter schools—an antecedent event) leads to domain B (overcrowding—a current condition) leads to domain C (more student arguments in hallways—observed event). Matrices can also be used to show patterns of association among domains or subdomains across units of analysis. A unit of analysis is the social unit that is being compared with others to identify patterns of interaction among domains, subdomains, or qualitative variables. Units of analysis are usually persons, places, events, or things.

Domains can be deconstructed during or prior to fieldwork as ideas emerge and are clustered under domain headings. For example, in a current study of Ecstasy use in relation to sexual risk, researchers at the Institute for Community Research identified the use of Ecstasy for other purposes, including coping with negative life situations. This general theme initially emerged during in-depth interviews. Later we classified 118 in-depth interviews on Ecstasy use into two groups: use of Ecstasy for coping and use of Ecstasy for recreational or sex-related reasons. Approximately half of the interviews could be classified in the first group. Within this subgroup of cases, we identified five major subdomains, such as managing abusive situations, dealing with life stresses, and coping with the loss of a loved one. These subdomains are the independent variables predicting ecstasy use as well as the pattern of use (frequency and amount) (Moonzwe, Schensul, & Kostick, in press). In another example, researchers from the Institute for Community Research and the University of Connecticut Health Center, in collaboration with researchers from the Mauritius Family Planning Association and the University of Mauritius, produced an initial model in which peers, family, and media all were perceived to play a role in how young women and men related to one another. The model predicted that these relational domains were linked with sexual behaviors and consequences, such as HIV (Schensul, Oodit, Schensul, Ragobur, & Bhowon, 1994). Researchers modified this model in the field by asking respondents about their activities with their peers and their relationships with their family, as well as how they obtained information through the media and elsewhere about sexuality and HIV; what intimate behaviors they were involved in; and what the emotional, physical, and social consequences of these relationships were. The initial model was then expanded. Such models as these are developed in a rudimentary format, expanded in the field, and finalized during the analytic phase of a study.

Where, When, and With Whom the Study Will Be Conducted

All researchers must make decisions about study sites, the time period during which a study is to be conducted, and the boundaries of the study population. In this section I discuss some of the elements that researchers should consider in deciding where, when, and with whom their study is to be carried out.

Where: Study Location

Qualitative studies generally take place in one or more physical locations called study sites. A study site may be a sociopolitical community (a neighborhood, a municipality, a village, or even a city) or multiple communities (for example, a cluster of villages) included for comparative purposes. Or it may be an institution within a community (for example, one or more early childhood learning centers, health outposts, university campuses, clinics, or parks). Lately some researchers have been turning to the Internet to conduct qualitative research, using blogs, wikis, and social networking sites to collect data and share them with respondents. Social networks of individuals, organizations, or both may also constitute the focus of a qualitative study. Some qualitative research focused on topical interests, however, may not be place based, such as a study of couples who have chosen in-vitro fertilization or of individuals who have experienced sexual abuse as children. Here the location may be less significant than the identification of a scarce or hidden population.

When: Timing and Duration of the Study

All researchers have to make decisions about the timing and duration of their study. Timing refers to the time of year and time during the day, evening, or weekend when the research will take place. Duration refers to the length of time during which the study will take place. Many factors may determine both the timing and the duration of a study. Such external factors as available funds or the time period available to the researcher (for example, a sabbatical or half-sabbatical, or a six- or twelve-month period in the field required for dissertation work) are important considerations. Timing is important in research that focuses on the plans and activities leading up to a ritual occurring only once a year, combined with interviews about the actual conduct of the ritual. Interviews about conducting and participating in the ritual and observations of the ritual require that the researcher be in the field during the period of time that covers planning activities, the ritual, and its aftermath. Researchers doing qualitative work on asthma emergency room visits might want to time their study so that it coincides with peak periods (fall and spring in the Northeastern United States, for example) and regular times in order to see how experiences of asthma differ.

Taking into consideration recall time, or how long a person can remember the details of a specific event, is also important. Unless the event is highly unusual and dramatic, most people do not recall the details of an experience very accurately more than two or three days after it occurs. Research that calls for accurate recall of such an event as childbirth, a wedding, or a funeral; an eviction or the shift from one type of residential situation to another; or a recent visit to the doctor for a health problem should take place within a week of the event. Life narrative interviews, however, do not require such timing because they call for recollection of a person's entire life history rather than a specific recent event (see Chapter Nine).

Finally, researchers need to take into consideration at what time during the day or week observations or interviews can best be conducted. Interviews with adolescents have to be timed for the after-school period between about 2:30 P.M. and 5:30 P.M. to avoid conflict with homework; but if adolescents are involved in after-school athletics, interviews may have to be conducted in the evening. Homeless people must be interviewed during the day because they are required to be registered in a shelter by later in the afternoon. Entrance and exit interviews with dance club clients can only be conducted in the evening and late at night as they are arriving at or leaving their club of choice. Families from northern India living in Mumbai usually go home for an extended period prior to the arrival of the rainy season in June; thus May through June is not a good time to plan to interview family members.

With Whom: Study Population

The term study population refers to the people who are the focus of the study. In a qualitative study there may be several study populations; for example, for a study on treatment programs for substance users, the two study populations may be service providers offering programs for drug users and the people who are using drugs. The study populations are always chosen in relation to the study topic, and the reasons why they are chosen—that is, their expected contribution to the study—must be given as part of the study design. As units of investigation and analysis, study populations may be chosen at different levels—census districts, town administrative bodies, local experts, family or household units, or individuals. In qualitative research the study population may or may not be equivalent to the study community, a place. Researchers conducting a study of barriers to female condom use might consider including those involved in reproductive health policymaking; potential or actual distributors, such as pharmacies and clinics; health care providers (for example, primary health care providers, nurses, health outreach workers, peer educators); and individual end users. Similarly, in a comparative study of the effects of social development curriculum on secondary school children's behavior, all middle schools would constitute the first-level study population. The research question would address differences in ways the curriculum is managed and delivered across schools. The second-level study population would be teachers who deliver the curriculum. Here the question might focus on how the curriculum is delivered and what facilitates or constrains delivery in each classroom. The third-level study population would be the students who engage with the curriculum; questions for these students might address their engagement with the curriculum, what they understand, what they might like to change, and how they feel about curriculum delivery. And the fourth-level study population might be parents, who could be asked about their knowledge of the curriculum content, how it is delivered, or their children's responses to it. A good study design might obtain some form of representative sample from each level or constituency to gain a holistic perspective on the study topic.

Sampling in Qualitative Research

Qualitative researchers sample for reasons that differ from those of quantitative or survey researchers, who usually prefer random or systematic sampling strategies that allow them to generalize or extend their results to the broader population from which a sample of respondents or events is selected. This is because qualitative research questions tend to focus on processes; on detailed contextual or historical descriptions; or on the meanings, interpretations, and explanations people assign to events, activities, and behaviors. Many qualitative questions are more concerned with validity (the degree to which the data and interpretation fit the situation) than with generalizability (the degree to which the study results can be generalized to the broader population from which the study sample is drawn).

Qualitative researchers use cases to illustrate the interactions among variables, recognizing that the same (or even different) variables may configure in different ways in other places or with other populations over time. They seek less for randomization and control than for understanding of the range of variation of behavioral phenomena or meanings in a population. Thus qualitative researchers turn to forms of purposive (also called purposeful), targeted, or systematic sampling, such as criterion sampling (identifying cases based on a set of criteria—for example, expertise in club drug use) or theoretical sampling (filling cells by characteristics of the study population defined in advance; cells may be filled according to such theoretical criteria as age, ethnicity, size, or level of empowerment). Another sampling option is extreme or midpoint sampling when the range of variation in a study population is known. Extreme sampling refers to choosing examples that represent the extreme ends or opposites of a continuum, whereas midpoint sampling involves choosing examples that are known to be typical of the study population. Or researchers may screen by study topic. In a study of drinking behavior in Mumbai, for example, researchers sought out married and unmarried men who had drunk a “little” (several times) or “a lot” (once a week or more) in the past thirty days, approximately twenty men in each of four cells. Finally, qualitative researchers use a variety of sampling strategies other than randomization to obtain representative samples from which generalizations can be made. For targeted sampling, researchers map all sites, or as many sites as possible, where people in the desired group gather; choose a sample of sites; and identify respondents from that site sample using preestablished protocols for choosing the number of respondents from each site. In respondent-driven sampling, respondents from randomly identified locations present recruitment cards to three others who volunteer for participation; these participants are asked to present cards to three others. This form of sampling is believed to result in an unbiased sample of the general population after approximately five such cycles (Heckathorn, 1997; Ramirez-Valles, Heckathorn, Salganik, & Heckathorn, 2004; Vázquez, Diaz, & Campbell, 2005) and is thus preferred to other forms of network sampling, provided that the respondents engage in behavior that is known to be networked. As with site selection, the researcher must justify each sampling approach. And it should be remembered that to choose a proper sample for validity or generalizability to the rest of the population or others like it requires knowledge of the study site, which is best obtained by spending as much time as possible learning about where people gather and in what numbers (sometimes referred to as site mapping or ethnographic mapping) (Tripathi, Sharma, Pelto, & Tripathi, 2010).