Introduction and Overview

Every empirical study you have ever read began with a research question. Most introductory methods courses devote a single lecture to question formulation and then move on, as if the rest of the design followed mechanically once the question was set. Bernard, Wutich, and Ryan take the opposite view: the question is the single most consequential choice a researcher makes, and a defensible qualitative project is one in which every later decision, including sampling, instrument, analytic method, and write-up, can be traced back to a well-specified research question. This lesson is about doing that upstream work well.

In the prior epidemiology sequence you were given research questions to answer with established designs (case-control, cohort, regression). In this course the question is yours to formulate, and the texture of your data, namely 20 loneliness transcripts collected with a semi-structured guide, both enables and constrains what you can defensibly ask.

1.1 Exploratory and Confirmatory Qualitative Research

The two terms describe the epistemic position of the researcher at the start of the study, not the analytic technique they use. An exploratory question is one for which the researcher does not yet have a candidate answer (or has only weak, ill-formed candidates). A confirmatory question is one for which the researcher has a specific candidate answer (a hypothesis, a model, a theory-derived prediction) and is testing it against data. In the language of reasoning you have met before, exploratory work is broadly inductive, reasoning upward from particular observations toward general categories, while confirmatory work is broadly deductive, reasoning downward from a general expectation to what should show up in this particular dataset.

Qualitative work has historically been most associated with exploration. The reason is that qualitative methods are particularly good at discovering categories, that is, what counts as a thing, what its dimensions are, how it differs across people, and discovery is the natural mode when you do not yet know what to measure. Bernard, Wutich, and Ryan are careful, however, not to collapse qualitative work into exploration. Grounded theory's later phases (Charmaz, 2006), analytic induction, and qualitative comparative analysis are confirmatory qualitative techniques (a later module will return to them). The point is that the same dataset can be analyzed exploratorily or confirmatorily, and you owe the reader an explicit statement of which posture your study is taking.

Notice that both questions can be answered from the same 20 transcripts. The first asks the data to teach you the categories; the second asks the data to evaluate a category already proposed in the literature.

1.2 The Four Screening Questions

Before you commit to a research question, Bernard, Wutich, and Ryan recommend running it through four screening questions. Each one is a different kind of feasibility check, and a question that fails any one of them is not yet a good question.

Many students arrive at this course with a topic, not a research question. The four screens are most useful as a movement from topic to question. “Loneliness” is a topic. “How do recently-bereaved older women in British Columbia describe the shape of their loneliness, and what social, material, and embodied features of their daily lives are they identifying as constitutive of it?” is a research question. Note how the second one specifies who, what, where, and what kind of answer it expects.

1.3 What Makes a Question Defensible for the Course Dataset

The course dataset constrains the questions an analyst can defensibly ask. Some constraints are obvious and some are less so.

The dataset has 20 transcripts. Twenty in-depth transcripts is a defensible qualitative sample size for descriptive and interpretive work (Sandelowski, 1995). That means questions about prevalence (“how common is…?”) are off-limits. So are questions that require dense subgroup contrasts (a question that splits the dataset 15 ways will have one or two transcripts per cell, which is not enough to characterize anything). Questions about kinds (what kinds of loneliness exist? how is each kind structured?), processes (how do people come to recognize their loneliness? what do they do when it shows up?), and meanings (what does loneliness signify to them?) are well-supported.

The dataset was collected using a specific semi-structured guide. Topics the guide asked about (definitions of loneliness, triggers, coping, social world, stigma, policy preferences) will be richly represented across transcripts. Topics the guide did not directly ask about (clinical symptomatology, family-of-origin attachment styles, neurobiological correlates) will appear only sporadically. A defensible question concentrates on what the guide elicited.

The dataset is synthetic. This is unusual, and any methods section written from it must acknowledge it. The transcripts are pedagogically valuable and analytically coherent, but you cannot draw causal claims from them, and you should be cautious about claims that depend on real-world demographic representativeness (since the synthetic composites were designed for analytic variation, not for representative sampling).

Introduction and Overview

In the prior courses theory often appeared as background: the biological model of disease, the social-ecological model, the proportionate-universalism framing of inequity. You used theory to motivate a study and to interpret findings, but you rarely had to declare which theoretical tradition you were working within. Qualitative research is different. The methodological tradition expects you to name your paradigm and to align your method to it. This section walks through the four paradigms most commonly named in contemporary public-health qualitative research and shows how the same loneliness question would be approached differently in each.

2.1 What a Paradigm Is

The Background box defines a paradigm as a bundle of commitments about ontology, epistemology, and axiology. The bundles are not internally consistent in every detail, but they cohere enough that researchers within a paradigm tend to agree on what counts as a good study. You may find paradigm-talk uncomfortable if your training is quantitative: quantitative epidemiology rarely makes you say out loud that you believe in an external reality knowable through measurement, because almost everyone in the room is already a postpositivist. Qualitative research draws practitioners from very different positions, so the convention is that you declare yours.

2.2 Positivism / Postpositivism

The view that an external reality exists and is, in principle, knowable. Strict positivism (rare today) treats social facts as fully observable; postpositivism, the dominant position in quantitative epidemiology, accepts that our access to reality is fallible and shaped by our instruments, while still treating reality as the target. A postpositivist qualitative study of loneliness would treat participants' accounts as imperfect windows onto an underlying phenomenon and would prioritize procedures (intercoder reliability, triangulation, transparency) that increase the credibility of the inference from accounts to phenomenon.

2.3 Interpretivism / Constructivism

An interpretivist qualitative study of loneliness would treat the accounts as the data, where the loneliness of interest is the loneliness as constructed in talk, not a hypothetical thing behind the talk (Thorne, Kirkham, & MacDonald-Emes, 1997). The metaphors participants choose, their silences, and even their refusals to answer become primary data. Charmaz's constructivist grounded theory (Charmaz, 2006), taught in Lesson 7, Section 2, is the most widely cited contemporary interpretivist methodology in health research.

2.4 Critical and Critical-Realist Paradigms

This lesson calls this family the critical (transformative) paradigms, the label used across the series for what HSCI 207 calls transformative and participatory stances and HSCI 230 calls critical theory. Critical realism, a related position on ontology which holds that real social structures and mechanisms exist and generate the events researchers observe, is grouped with the family here because it shares its attention to structural causes. These are frameworks that explicitly attend to power, structural inequality, and the historical embeddedness of phenomena. Feminist, anti-racist, decolonial, Marxist, and critical-realist approaches share the commitment that research is not neutral and that part of the researcher's job is to make visible the structural conditions that produce the phenomenon. A critical loneliness study might foreground how policy choices (housing affordability, immigration restrictions, healthcare under-funding, long-term-care austerity) produce specific patterns of isolation, and would treat participants' accounts as evidence about structural conditions as well as individual experience.

2.5 Pragmatism

The Background box defines pragmatism. In practice, pragmatist researchers mix methods, draw on different paradigms in different parts of a project, and ground decisions in “what works for this question.” Pragmatism is common in applied health-services and implementation-science research. Its critics worry that it can become an excuse for non-disclosure; its defenders argue that strict paradigm-fidelity can produce studies that are philosophically clean but practically irrelevant.

2.6 How Theory Enters a Study: Sensitizing Concepts and Conceptual Frameworks

Paradigm is theory at its most abstract. Theory also enters a study in a more hands-on way, and two levels are worth naming. A conceptual framework is a chosen set of concepts and assumed relationships that frames the question and gives the analysis a place to start; the framework synthesis method described later leans on exactly this. A sensitizing concept, a term from the sociologist Herbert Blumer, is looser: a broad guiding idea such as “stigma” or “social capital” that suggests where to look without dictating what you will find. This distinction dissolves a false choice that trips up many beginners. Even exploratory, inductive work is rarely theory-free, because you almost always bring sensitizing concepts to the data. What separates exploratory from confirmatory work is not the presence of theory but how tightly it constrains the answer: a sensitizing concept opens up the looking, while a tested hypothesis narrows it to a yes or a no.

2.7 From Paradigm to Methodology: Six Qualitative Designs

A paradigm states assumptions; a methodology is a recognized research design that puts a set of assumptions to work, with its own aims, sampling logic, and analytic procedures. Six methodologies account for most qualitative health research. Each tends to draw on one paradigm, although researchers adapt them, and each connects to one of the five analytic paths (text to counts, themes, schemas, narratives, or talk) set out in Lesson 6, Section 1.

Phenomenology and interpretative phenomenological analysis (IPA). Phenomenology studies lived experience: what a phenomenon is like for the people who live it, described as closely as possible to how they experience it. Descriptive phenomenology, in the tradition of Husserl, asks the researcher to set aside prior assumptions and leans toward a realist stance. Interpretative phenomenological analysis (Smith, Flowers, & Larkin, 2022) is interpretivist: the researcher makes sense of participants making sense of their experience, usually with a small, homogeneous sample and long, minimally structured interviews. A phenomenological study of loneliness would ask what loneliness is like for people in one situation, such as recent widowhood. It follows the text-to-themes path, and Lesson 4, Section 3.1 describes interviewing for IPA.

Ethnography. Ethnography studies the practices, meanings, and shared understandings of a group in its own setting, through prolonged participant observation supported by interviews and documents. Most health ethnography is interpretivist, and critical ethnography brings the critical (transformative) paradigm to the same methods. An ethnography of loneliness might spend months in a seniors' housing complex. Lesson 4, Sections 2 and 4.3 cover participant observation and field notes, and the text-to-schemas path (Lesson 9, Sections 1 and 2) suits ethnographic questions about shared cultural models.

Grounded theory. Grounded theory builds an explanatory theory of a process from the data, through constant comparison and theoretical sampling. The classic Glaserian and Straussian versions lean toward postpositivism, and Charmaz's constructivist version is interpretivist. It suits questions about how a process unfolds, such as how people come to recognize and respond to their own loneliness. Lesson 7 teaches it on the text-to-themes path.

Narrative inquiry. Narrative inquiry treats each account as a story and asks how it is structured, what it accomplishes for the teller, and what identities it builds. It draws mainly on interpretivist and constructivist assumptions. A narrative study might ask how people who came out late in life narrate the loneliness that followed. Lesson 9, Section 3 develops it on the text-to-narratives path.

Case study design. A case study examines a bounded case, such as a program, a clinic, or a community, in depth and in its real-world context, usually drawing on several sources of data (Yin, 2018; Stake, 1995). Yin's version treats the case as a test of stated propositions and leans toward postpositivism; Stake's version emphasizes understanding the particular case and is constructivist. Case studies suit health questions about how and why a program works, or stalls, in a specific setting, for example a community connector program for isolated older adults. Analysis usually follows the text-to-themes path with within-case and cross-case matrices (Lesson 7, Section 1).

Qualitative description. Qualitative description (Sandelowski, 2000) aims for a comprehensive, low-inference summary of an experience or event in the everyday language of the participants, staying close to the data. It is usually described as drawing on naturalistic inquiry and is often chosen for pragmatic reasons. It suits applied health questions that call for a clear account of what people experience or want, such as patients' views of a new service. Analysis typically uses qualitative content analysis (Lesson 8) or thematic analysis (Lesson 5, Section 1.4).

A study of the loneliness interview dataset most naturally takes the form of qualitative description or thematic analysis, because the interviews were collected with a single semi-structured guide in one recruitment phase. The later lessons also read the transcripts through grounded-theory, narrative, and phenomenological lenses, and the design limits what each lens can claim: the dataset was not built through theoretical sampling, and its sample is deliberately heterogeneous, while IPA prefers a homogeneous one.

Introduction and Overview

The literature search for qualitative evidence works on the same logic but with important differences. The searchable concepts are different, the databases are partly different, the inclusion criteria are different, and the synthesis methods are very different. This section walks through what is the same, what is different, and how to conduct a defensible qualitative-evidence search for a qualitative study.

3.1 What Is the Same as a Quantitative Review

The basic discipline of a systematic search transfers directly. You still:

• Specify your search question in advance, using a structured frame. PICO can be adapted, but SPIDER (Sample, Phenomenon of Interest, Design, Evaluation, Research type) is purpose-built for qualitative searches (Cooke, Smith, & Booth, 2012).
• Search multiple databases.
• Document your search strings, dates, and result counts.
• Apply inclusion and exclusion criteria with audit trail.
• Screen titles/abstracts, then full text, with a second reviewer if possible.
• Extract data into a structured table.
• Appraise the quality of included studies.

The PRISMA flow diagram (or its qualitative variants, eMERGe for meta-ethnography and ENTREQ for transparency in reporting synthesis) is still the convention. For reporting your own primary qualitative study, the two dominant standards are COREQ (Tong, Sainsbury, & Craig, 2007) for interview and focus-group studies and SRQR (O'Brien, Harris, Beckman, Reed, & Cook, 2014) for qualitative research more broadly.

COREQ is a 32-item checklist organized in three domains: research team and reflexivity (who conducted the interviews, their training, and their relationship with participants), study design (the theoretical framework, participant selection, setting, and data collection), and analysis and findings (how data were coded and themes derived, and how findings are reported with quotations). SRQR is a 21-item standard that applies to qualitative research of any design, including ethnography, grounded theory, and case studies as well as interview studies, and runs from the title and abstract through the methods, results, and discussion to funding and conflicts of interest. Researchers use the checklists while drafting and journals often ask for the completed checklist with a manuscript. HSCI 207 Lesson 12, Section 2.3 (Writing Up Research) applies both checklists to a worked methods section and is optional reading.

3.2 What Is Different

The differences are substantive enough that doing a qualitative search using a quantitative-review playbook will produce a worse search.

The search terms are different. Qualitative studies are not always indexed with reliable terms like “qualitative” or “interview” in their titles or abstracts. The Cochrane Qualitative and Implementation Methods Group has published validated search filters for several databases that you should use rather than improvising. Common useful terms include qualitative, interview, focus group, ethnography, grounded theory, phenomenology, lived experience, narrative, plus methodologically specific terms.

The databases are partly different. MEDLINE and Embase still matter, but CINAHL (nursing), PsycINFO (psychology), Sociological Abstracts, and ASSIA (Applied Social Sciences Index and Abstracts) are particularly important for qualitative health research. Several databases have dedicated qualitative-research filters built in. Grey literature (theses, reports, working papers) is more important in qualitative reviews because qualitative studies are over-represented in unpublished form.

The quality appraisal is different. The Cochrane risk-of-bias tool is not the right instrument. The Critical Appraisal Skills Programme (CASP) qualitative checklist, the Joanna Briggs Institute qualitative checklist, and the Walsh and Downe (2006) criteria are the most widely used.

The synthesis is profoundly different. You do not pool effect sizes. The next subsection turns to what you do instead.

3.3 The Major Qualitative-Evidence Synthesis Methods

Five approaches dominate the qualitative-evidence synthesis literature. They differ in epistemology and in how aggressively they reduce findings.

Critical interpretive synthesis, meta-summary, and meta-aggregation

The two methods that the Background box leaves out sit at opposite ends of the interpretive spectrum. Critical interpretive synthesis (Dixon-Woods et al., 2006) treats the review question as something that can be refined during the review, samples studies purposively, and appraises them mainly for their relevance and contribution to the developing argument. Its product is a synthesizing argument, a new theoretical account that also critiques how the literature has framed the problem. It suits large, mixed bodies of evidence, such as research on access to health care for vulnerable groups, where a tidy summary would hide disagreements about what the problem is.

Qualitative meta-summary (Sandelowski & Barroso, 2007) sits at the aggregative end. Findings are extracted as statements, grouped into topical categories, and counted. A frequency effect size is the proportion of included studies that report a given finding, and an intensity effect size is, for each study, the proportion of all the distinct findings in the synthesis that the study contains. JBI meta-aggregation (Lockwood, Munn, & Porritt, 2015) is also aggregative but does not count. Reviewers extract each finding with an illustrating quotation, rate it as unequivocal, credible, or not supported, group similar findings into categories, and combine the categories into synthesized findings written as statements that can guide practice. Both methods leave the primary authors' interpretations largely intact, which separates them from meta-ethnography and critical interpretive synthesis; meta-summary reports how often a finding recurs, and meta-aggregation reports what practitioners can do with it.

Choosing a method: the RETREAT criteria

Booth and colleagues (2018) identified seven considerations for choosing among qualitative evidence synthesis methods, summarized by the acronym RETREAT: the review question, the epistemology of the review team, the time frame, the resources available, the team's expertise, the audience and purpose of the review, and the type of data in the primary studies. A team with a focused question, a short deadline, and policy readers who want clear statements leans toward thematic synthesis, framework synthesis, or meta-aggregation. A team with an interpretive epistemology, more time, and conceptually rich primary studies can attempt meta-ethnography or critical interpretive synthesis. Thin, descriptive primary studies support aggregation better than reinterpretation.

How much confidence? A worked GRADE-CERQual assessment

GRADE-CERQual (Lewin et al., 2018) rates confidence in each finding of a qualitative evidence synthesis, meaning how likely it is that the finding is a reasonable representation of the phenomenon. Reviewers judge four components, recording no or very minor, minor, moderate, or serious concerns for each, and then make an overall judgement of high, moderate, low, or very low confidence. Every finding starts at high confidence and is rated down when the concerns are important.

The worked example below uses a hypothetical review. Suppose a thematic synthesis of six qualitative studies of loneliness after the death of a spouse produced the finding "Widowed older adults describe loneliness as attached to the objects and daily routines they once shared with their spouse."

Overall, the reviewers might judge moderate confidence: the finding is likely to be a reasonable representation of the phenomenon, rated down once because the concerns about adequacy and relevance together leave open whether it holds for widowed men. A summary-of-qualitative-findings table reports this judgement beside each finding with a brief explanation, so that readers can see why confidence was rated down.

A literature framing for a primary qualitative study does not require a synthesis. It summarizes what is known about the topic from qualitative research, identifies the gap the study would address, and locates the research question in that body of work. A one-page framing with eight to twelve cited sources is a common length.

Introduction and Overview

This section walks through the process of moving from a public-health topic to a defensible research question, using a worked example built on the loneliness dataset. It then sets out the ethical considerations a qualitative methods section addresses, including the TCPS 2 provisions specific to qualitative research.

4.1 Worked Example: From “Loneliness in Older People” to a Research Question

A typical student arrives at the course with a topic like “loneliness in older people” or “immigrant loneliness.” Each of these is a starting point, not a research question. The work of question formulation is to narrow, specify, and ground the topic.

Step 1: From topic to question stub. What about loneliness in older people interests you? “How older people describe the experience of loneliness.” That is closer to a question, but still vague: “describe” could mean anything from clinical phenomenology to social-network mapping.

Step 2: Specify the kind of answer. What do you want to know that you do not now know? “What dimensions and triggers of loneliness do older participants name in their own accounts, and how do these compare to younger participants' accounts?” Now the question has a comparison built in and is asking for content that the dataset can provide.

Step 3: Ground the question in the dataset. Can the 20 transcripts answer this? Five of the 20 participants are over 65 (P05 Linda, P11 Helen, P14 Kenji, P17 Jacob, P20 Frank). Five is enough for descriptive characterization but not for fine-grained subgroup work. The question is dataset-feasible if you accept that “older” will be defined by those five and that comparison to younger participants will be at the level of patterns, not statistical contrasts.

Step 4: Locate in the literature. The qualitative literature on older-adult loneliness is well-developed (Cacioppo & Patrick, 2008; Hawkley & Cacioppo, 2010; the Cardona et al. 2025 review). What can a study of this dataset add? A defensible answer: a comparison across life-stage from a single, methodologically consistent dataset, where most prior qualitative work has studied life-stages separately.

Step 5: Run the four screens. Personal interest (you do care). Empirical answerability (the dataset can support the question, with stated caveats about sample size). Resources (a few months is sufficient for a five-vs-fifteen-transcript comparative analysis). Ethics (the dataset is synthetic, so the question is REB-clean for instructional purposes).

Final question: “How do older British Columbian adults (age 65+) describe the dimensions and triggers of their loneliness, and in what ways do their accounts pattern differently from those of younger adults in the same dataset?”

4.2 Screening a Draft Question

The worked example is scaffolding for drafting a question of your own, and it works best when adapted to the topic at hand. A finished question should be specific enough that another reader could predict, in broad terms, what the methods section will say.

4.3 Ethics of Qualitative Research: What a Methods Section Addresses

Even with a synthetic dataset, a methods section is expected to address what the ethical considerations would be for a real-world study. The TCPS 2 framework asks researchers to address:

• Free and informed consent: what would participants be told? How would consent be obtained for sensitive topics? How would withdrawal work?
• Confidentiality: how would identifying details be removed from transcripts? What is your data-storage plan? How will quoted excerpts be presented?
• Risk of harm: what is the risk of re-traumatization from an interview about loneliness? What support resources would be provided?
• Justice in subject selection: who is included, who is excluded, and on what grounds?
• Researcher obligations of care: what does the researcher owe a participant who discloses something distressing during an interview?

Bernard, Wutich, and Ryan (Ch. 2, pp. 23–26) are good on the ethics framing. The Canadian-specific operationalization is in TCPS 2 (2022).

TCPS 2 Chapter 10: Provisions for Qualitative Research

TCPS 2 devotes Chapter 10 to qualitative research, because several features of qualitative designs fit poorly with procedures written for experiments and surveys. Four of its provisions bear directly on the checklist.

• Emergent design and evolving interview guides (risk of harm). Qualitative questions often develop during a study. Chapter 10 asks researchers to give the REB all the information available about the general approach, such as draft questions or a thematic outline, and advises REBs not to require a complete interview schedule in advance. Changes that do not significantly alter the approved design need no further review, but changes that could alter the level of risk, such as adding questions about trauma, must be approved before they are used. A risk-of-harm discussion therefore states how the guide may evolve and how changes that alter risk would be brought back to the REB.
• Consent as an ongoing, relational process (free and informed consent). Chapter 10 treats consent as negotiated over the course of a research relationship and recognizes that a signed form can be inappropriate in qualitative work. Consent may be documented in other ways, such as oral consent recorded on audio or in field notes, or a consent log, provided the design explains how consent will be sought and recorded. In some communities, permission from leaders or Elders comes before individual consent. A consent discussion therefore names the modality, how consent is revisited as topics deepen, and how participants can withdraw.
• Observation in natural settings (free and informed consent). When researchers observe people in physical or online settings where those people have some expectation of privacy, Chapter 10 requires a justification for any exception to consent and a description of the privacy protections, and covert observation needs specific justification. Public settings are not automatically free of privacy expectations: religious ceremonies and online groups, for example, can carry them.
• Identifiability and disclosure (confidentiality). Chapter 10 notes that the depth and detail of qualitative data can make individuals, groups, or communities identifiable in publications. The risk is greatest in small communities, where a job title, an illness, and a neighbourhood together can point to one person. Confidentiality is the default, but some participants, as in oral history, may wish to be named; researchers discuss this with participants and record their consent to be identified, provided that disclosure does not harm others. A confidentiality discussion therefore explains how quotations will be edited or attributed and how identifying detail will be handled.

The remaining two checklist items draw on other parts of the policy: justice in subject selection on the core principle of justice, and the obligations of care after disclosure on concern for welfare, which a distress protocol puts into practice.

Bringing It All Together

This lesson has done the upstream design work: you have moved from a topic to a defensible research question, named the paradigm you will work in, and located the question in the qualitative loneliness literature. This work feels disproportionate to its concrete output (a paragraph and a page), but it determines what every later step of the course can do for you. A vague question produces a vague methods section, a vague analysis, and a vague paper. A precise question makes everything that follows possible.

Later lessons turn to the design infrastructure a qualitative methods section is written in: sampling logic (fixed in advance for the course dataset, but still documented defensibly in any write-up) and data collection (also fixed, but the methodological vocabulary is what makes the dataset legible to a reader). Later lessons begin analysis proper.

The final reflection below asks you to bring the lesson together around one question about the course dataset.

Final Knowledge Assessment

This glossary collects the key concepts, methods, and people introduced in this lesson. Type in the search box to filter entries.