Sociological research methods: surveys, ethnography, experiments, mixed methods

Intuition Beginner

Sociologists study society using systematic methods, not just personal opinion. If you want to know whether Americans are becoming less religious, you cannot rely on what your neighbors happen to believe. You need to ask many people the same questions, record their answers, and look for patterns. This is what separates sociology from casual observation: it follows rules of evidence that let other researchers check the findings.

Surveys ask large numbers of people standardized questions. The General Social Survey, run since 1972, has tracked American attitudes on religion, work, politics, and family for over fifty years. Because it asks the same questions year after year, sociologists can see how the country actually changes — not how it seems to change from the headlines.

Ethnography takes the opposite approach. Instead of asking thousands of people a few questions, a researcher spends months or years inside one community, watching daily life unfold. William Foote Whyte lived for years in a Boston neighborhood to write Street Corner Society. Mitch Duneier worked among street vendors in Greenwich Village for Sidewalk.

Experiments manipulate one variable while holding others constant, to test cause and effect. Comparative-historical methods trace how entire societies change across decades or centuries. Each method has different strengths: surveys describe broad patterns, ethnography reveals meaning and detail, experiments establish causation, and historical methods show long-term change.

No single method can do everything. A survey can tell you that something is happening across a population, but not what it feels like to live through it. Ethnography can capture that experience in depth, but cannot easily generalize to a whole country. This is why most sociologists use mixed methods — combining approaches so that each compensates for what the others miss.

Visual Beginner

The table below compares the four major research methods introduced in this unit. Each method answers a different kind of question.

Method	Core question	What it produces	Main limitation
Survey	How common is X across a large population?	Broad patterns, percentages, trends over time	Cannot explain why people answer as they do
Ethnography	What does daily life mean to the people living it?	Deep description of norms, practices, and meanings	Hard to generalize beyond one setting
Experiment	Does variable A cause outcome B?	Evidence of causation	May not reflect real-world conditions
Comparative-historical	How and why did societies differ or change over time?	Explanations of large-scale transformation	Depends on incomplete historical records

Key term	Plain-language meaning
Variable	Something that can differ (income, age, religion) and that a researcher measures or compares
Sample	The subset of people a researcher actually studies, standing in for a larger population
Cause and effect	Showing that changing one thing (the cause) produces a change in another (the effect)
Generalize	To argue that what you found in your sample also holds for the wider population
Mixed methods	Combining two or more methods so that each fills in what the others miss

Worked example Beginner

Example 1: The General Social Survey

The General Social Survey (GSS) has asked Americans the same core questions since 1972. One recurring question asks whether a person who believes that Black people are genetically inferior should be allowed to give a speech in the respondent's community. In the 1970s, a large majority of white Americans said no. Today, that proportion has dropped sharply. The survey did not change the question — the country changed.

This is the power of a repeated survey. By asking identical questions over decades, researchers can track real shifts in attitudes without guessing. The GSS also asks about confidence in institutions, personal happiness, job satisfaction, and religious belief. Sociologists use this data to describe how American society has actually evolved, grounding claims about social change in evidence rather than impression.

Example 2: Pager's audit study

Devah Pager wanted to know whether having a criminal record hurt job applicants — and whether the penalty was worse for Black applicants than for white ones. In 2003, she sent trained young men to apply for entry-level jobs in Milwaukee. The applicants were matched in pairs: same resume, same interview script, same demeanor. One key difference was randomly assigned.

Some applicants were told to indicate they had a felony drug conviction; others were told to present a clean record. Because the criminal record was randomly assigned and everything else was held constant, any difference in employer callbacks had to come from that single variable. The result was stark: a white applicant with a criminal record received more callbacks than a Black applicant without one. The experiment isolated race and criminal history as causes, something a survey alone could not do.

Check your understanding Beginner

Formal definition Intermediate+

A research method is a systematic procedure for collecting and analyzing evidence about social life. The choice of method is not neutral: it determines what questions can be asked, what evidence counts as an answer, and what kinds of claims the study can support.

Validity and reliability

Validity refers to whether a study measures what it claims to measure. Construct validity asks whether the operational definition of a concept (say, "religious commitment" measured as church attendance) actually captures that concept. Internal validity asks whether the study's design supports a causal inference — whether the observed effect can be attributed to the variable the researcher manipulated rather than to some confound. External validity asks whether the findings generalize beyond the specific people, setting, and time in which the study was conducted.

Reliability refers to consistency: whether the same measurement, repeated under the same conditions, would produce the same result. A reliable survey item produces similar answers when administered again to comparable respondents. A method can be reliable without being valid (consistently measuring the wrong thing), and valid without being reliable if a single measurement is noisy.

The five method families

A survey collects standardized responses from a sample of respondents using a structured questionnaire (closed-ended or open-ended items). Its strength is breadth: with probability sampling, findings generalize to a defined population.

Ethnography (participant observation) is sustained immersion in a social setting over weeks, months, or years, combined with field notes, interviews, and reflexive analysis. Its strength is depth: it captures the meanings, norms, and practices that participants themselves may not articulate.

An experiment manipulates one or more independent variables while controlling others, using random assignment to create equivalent treatment and control groups. Its strength is causal inference: randomization balances confounders, so group differences can be attributed to the treatment.

Comparative-historical research analyzes cases (societies, revolutions, welfare regimes) across time and place, using Mill's methods of agreement and difference, process tracing, and archival evidence. Its strength is explaining large-scale transformation.

Content analysis systematically codes texts, media, or cultural artifacts according to a defined scheme. Its strength is making cultural material analyzable at scale, either quantitatively (frequency counts, sentiment scoring) or qualitatively (thematic interpretation).

Mixed-methods research deliberately combines two or more of the above. The quantitative component typically establishes that a pattern exists and how much; the qualitative component explores how and why. Triangulation — using multiple methods to study the same phenomenon — strengthens confidence when findings converge and generates insight when they diverge.

Key concepts: the major research methods Intermediate+

Survey research

Survey methodology rests on two pillars: questionnaire design and sampling.

Questionnaire design. Each item must be valid (measuring what it claims), reliable (producing consistent results), and neutrally worded. Leading questions ("Do you agree that the government wastes your money?") bias responses. Double-barreled questions ("Do you support freedom and democracy?") conflate two issues. Likert scales measure attitude intensity on ordered response options (strongly disagree, disagree, neutral, agree, strongly agree). Pilot testing catches ambiguity before the full study launches.

Sampling. A population is the entire group the researcher wants to generalize to; a sample is the subset actually studied. Probability sampling uses random selection so every member has a known, nonzero probability of inclusion. The main types are simple random sampling (every member has an equal chance), stratified sampling (the population is divided into strata — e.g., by race or region — and random samples are drawn from each), and cluster sampling (randomly selecting groups, such as schools or neighborhoods, then sampling within them). Non-probability sampling includes convenience sampling (whoever is easiest to reach), snowball sampling (participants refer others), and purposive sampling (the researcher selects cases for theoretical relevance). Only probability sampling supports formal statistical generalization.

Sampling error is the random variation between a sample statistic and the true population value; it shrinks as sample size grows. Representativeness is the degree to which the sample mirrors the population on relevant characteristics. Nonresponse bias arises when the people who refuse or are unreachable differ systematically from those who respond — an increasingly serious problem as response rates to telephone and door-to-door surveys have declined over the past three decades.

Major survey infrastructures include the General Social Survey (United States, since 1972), the World Values Survey (cross-national, waves since 1981), and the survey programs of the Pew Research Center. Online panels have partially replaced in-person interviewing, raising new questions about coverage bias (who is online) and data quality (inattentive respondents).

Ethnography and participant observation

Ethnography's central concept is thick description, from Clifford Geertz (1973), who borrowed the term from the philosopher Gilbert Ryle. A thin description says "he closed one eye." A thick description distinguishes an involuntary blink from a conspiratorial wink from a parody of a wink. The physical movement is identical; the meaning is different. Thick description places behavior in its interpretive context so that the reader understands not just what happened but what it meant to participants.

Gaining access to a field site requires negotiation: finding a gatekeeper, establishing a role, and building rapport — mutual trust that makes candid observation possible. The researcher is the research instrument; their identity (race, gender, class, age, language) shapes what participants will show and tell. Reflexivity is the disciplined practice of analyzing how the researcher's own position affects the data collected. The risk of going native — over-identifying with participants to the point of losing analytical distance — is a recognized hazard, though some argue that deep empathy and critical analysis are not mutually exclusive.

The ethnographic canon includes foundational community studies: Whyte's Street Corner Society (1943, Italian-American corner gangs in Boston), Duneier's Sidewalk (1999, Black street booksellers in Greenwich Village), Anderson's Code of the Street (1999, the negotiation of respect and violence in Philadelphia), Bourgois's In Search of Respect (1995, Puerto Rican crack dealers in East Harlem), and Desmond's Evicted (2016, housing insecurity in Milwaukee). Each combined sustained observation with interviews and, in some cases, quantitative data.

Digital ethnography extends participant observation to online communities: social media platforms, forums, gaming worlds, and messaging apps. It raises questions about what "the field" means when interaction is mediated by screens, whether online behavior is performative in different ways than offline behavior, and how consent is obtained in semi-public digital spaces.

Experiments in sociology

Sociologists distinguish three types of experiment. Laboratory experiments take place in controlled settings and offer maximum internal validity but may sacrifice realism. Field experiments take place in natural settings — Pager's audit study sent real applicants to real employers. Natural experiments exploit events outside the researcher's control (a policy change, a natural disaster, an administrative boundary) that create treatment and control groups as if by random assignment.

Random assignment is the defining feature of a true experiment. By randomly assigning participants to treatment and control groups, the researcher makes the groups equivalent in expectation — confounders, both known and unknown, are balanced across groups. Any post-treatment difference is then attributable to the intervention rather than to preexisting differences. The trade-off is internal versus external validity: tight laboratory control strengthens causal inference but may make the setting so artificial that findings do not generalize.

Vignette studies present respondents with hypothetical scenarios (a short story about a job applicant) and vary one element across versions to see how it affects judgments. Audit studies go further: they send matched testers — or matched applications — into real hiring, housing, or lending markets, varying only the characteristic of interest (race, gender, criminal record) while holding all else constant. Pager's 2003 Milwaukee study is the paradigmatic example, finding that a criminal record reduced callbacks by roughly 50 percent and that the racial gap was larger than the criminal-record gap.

Comparative-historical methods

Comparative-historical sociology analyzes large-scale processes — revolutions, democratization, welfare-state formation, colonialism — across cases and over long time spans. The method draws on two of J. S. Mill's methods of agreement and difference. The method of agreement identifies what cases that share an outcome have in common, inferring that the shared factor may be causally relevant. The method of difference compares cases that are similar in most respects but differ in the outcome, inferring that the factor that varies is causally implicated.

Process tracing opens the black box between cause and outcome by tracing the causal chain step by step within a single case, testing whether each predicted link holds against the evidence. Path dependence describes how early choices (or accidents) set institutions on trajectories that become difficult to reverse, as later steps are constrained by what came before. Critical junctures are moments when options are unusually open and the path chosen has lasting consequences.

Key practitioners include Charles Tilly (state formation, contention, and the historical construction of inequality), Theda Skocpol (States and Social Revolutions, 1979, comparing the French, Russian, and Chinese revolutions), and Barrington Moore (Social Origins of Dictatorship and Democracy, 1966, tracing how agrarian class structures produced different political outcomes — "no bourgeoisie, no democracy," but also "no peasantry, no communism").

Content analysis and cultural sociology

Content analysis systematically codes cultural material — news articles, advertisements, films, television shows, social media posts, legislative texts — to identify patterns that would be invisible in unsystematic reading.

Qualitative coding develops categories from the data through iterative reading. In the grounded theory tradition of Glaser and Strauss (1967), coding proceeds in stages: open coding (generating initial categories by closely reading the material), axial coding (relating categories to one another), and selective coding (identifying a core theme that integrates the analysis). The goal is theory that is grounded in evidence rather than imposed from above.

Quantitative content analysis applies a predefined coding scheme to count occurrences and test hypotheses — for example, measuring how often newspapers associate different racial groups with crime, or tracking how often the word "terrorism" appears in coverage of different regions. Computational approaches now extend this to large corpora through word-frequency analysis, topic modeling, and sentiment analysis.

Mixed-methods designs

Mixed-methods research is not merely using two methods; it is a deliberate integration with a logic connecting the parts. Sequential designs use one method's findings to inform the next — a survey identifies a pattern, then interviews explain it. Concurrent designs run quantitative and qualitative components in parallel and integrate the results at the interpretation stage. Transformative designs use a theoretical framework (feminist, critical race, participatory) to guide the integration, ensuring that the study serves the communities it studies.

Complementarity means each method addresses what the other cannot. Triangulation means using multiple methods to converge on the same finding, increasing confidence. When methods produce divergent findings, the discrepancy itself is analytically productive: it points to a gap between what people say (survey) and what they do (observation), or between the broad pattern and the lived experience within it.

Exercises Intermediate+

Advanced results: epistemology, case selection, and computational frontiers Master

Epistemological debates: positivism, interpretivism, critical theory

The choice of method is not merely technical; it reflects a position on what kind of knowledge sociology can produce. Three traditions define the spectrum.

Positivism, descending from Comte and Durkheim, holds that sociology should emulate the natural sciences: social phenomena are governed by regularities that can be discovered through systematic observation, measurement, and hypothesis testing. The positivist ideal is objective, value-neutral knowledge of causal laws.

Interpretivism, descending from Weber's Verstehen and developed by Geertz and the phenomenological tradition, holds that social life is fundamentally different from natural phenomena because it is constituted by meaning. People act on the basis of how they interpret their situation, and the sociologist's task is to understand those interpretations, not merely measure their behavioral outputs. An interpretivist does not deny causation but insists that causal accounts of human behavior must pass through meaning.

Critical theory, in the tradition of the Frankfurt School and Jürgen Habermas, holds that sociology is never value-neutral. Knowledge is always produced within — and serves or challenges — structures of power and domination. Critical theory does not merely describe society; it aims to emancipate. Its methods include ideology critique, which exposes how dominant frameworks present contingent arrangements as natural or inevitable.

These positions are not mutually exclusive in practice. Most working sociologists draw on all three, but the relative weight given to each shapes the methods they trust, the questions they ask, and the standards by which they judge knowledge claims.

Standpoint epistemology and positionality

Standpoint epistemology, developed by Sandra Harding and Dorothy Smith, argues that knowledge is produced from a social location, not from nowhere. The standpoint of marginalized groups — women, racial minorities, the working class, colonized peoples — provides an epistemic advantage: those at the margins must understand both their own perspective and the dominant perspective that governs their lives, while those at the center can afford to treat their own experience as universal. Harding argued that standpoint is not a fixed identity but an achievement: it requires systematic investigation of the social structures that produce different vantage points.

Positionality extends this into methodology. The researcher's position — race, gender, class, sexuality, nationality, institutional affiliation — is not a bias to be eliminated but a condition of knowledge production to be analyzed. Reflexivity, in this framework, is not a confession of limitation but a methodological practice that makes the conditions of knowledge visible. Objectivity, redefined, is not the absence of a standpoint but the disciplined awareness of multiple standpoints.

The performativity of methods

John Law and John Urry have argued that social science methods do not merely describe social reality but help create it. The categories that surveys impose (race, occupation, marital status) become the categories through which people understand themselves and are governed. A survey that asks respondents to classify themselves by race reinforces the idea that race is a fixed, enumerable category. Methods are not neutral instruments; they are performative — they enact the social world they claim to represent.

This argument does not invalidate quantitative methods, but it complicates the positivist assumption that measurement is a transparent window onto preexisting reality. The choice of what to measure, how to categorize, and which differences count as significant are theoretical and political decisions embedded in the research instrument itself.

Case selection logic

Mario Small's influential essay "How Many Cases Do I Need?" (2009) argued that there is no statistical answer to the question of how many cases a qualitative study needs. The number of cases is determined by the research question and the inferential logic, not by a rule of thumb. Case selection should be theoretically motivated: cases are chosen because they illuminate a theoretical problem, not because they are representative of a population in the statistical sense.

Seawright and Gerring (2008) systematized the logic of case selection in small-N research by identifying distinct case types, each serving a different inferential purpose:

• Typical cases confirm that a theory holds in the expected setting.
• Diverse cases test whether a theory holds across varied conditions.
• Extreme cases probe the theory at its limits.
• Deviant cases contradict the theory and force its revision.
• Influential cases are those where, if the theory fails, it fails most damagingly.
• Most-similar systems compare cases that are similar on many dimensions but differ on the outcome, isolating the cause (Mill's method of difference).
• Most-different systems compare cases that differ on many dimensions but share the outcome, isolating what they have in common (Mill's method of agreement).

The choice of case type determines what kind of inference the study can support, and a mismatch between case type and inferential ambition is a common source of weak research design.

Relational sociology

Mustafa Emirbayer's "Manifesto for a Relational Sociology" (1997) argued that sociology should study not substances (individuals, groups, institutions as fixed entities) but transactions — the dynamic, unfolding relations through which social life is constituted. A network is not a collection of nodes but a pattern of ties; an organization is not a thing but a stabilized set of transactions; an identity is not an attribute but a relation.

Relational sociology draws on network analysis, which formalizes social structure as a graph of nodes (actors) and edges (relations), and on process sociology, which treats social life as unfolding over time rather than as a static snapshot. The relational turn pushes against methodological individualism (explaining everything by properties of individuals) and against reification (treating fluid processes as fixed things).

Big data and computational sociology

Matthew Salganik's Bit by Bit (2018) mapped the landscape of computational social science. Digital trace data — search queries, social media posts, mobile phone records, transaction logs — provides behavioral data at unprecedented scale and granularity, often without the researcher needing to ask anyone anything. Online experiments can recruit hundreds of thousands of participants (Salganik, Dodds, and Watts's 2006 music-listening experiment showed how social influence creates inequality in cultural markets). Mass collaboration projects distribute analysis across many researchers. Computational text analysis — topic modeling, word embeddings, machine-learning classifiers — processes cultural material at a scale no human coder could match.

The promise is tempered by three concerns. Ethics: much digital trace data was collected without informed consent for research use; the line between public and private behavior is blurred online, and re-identification of "anonymized" data has repeatedly been demonstrated. Bias: digital traces overrepresent populations with internet access and platform literacy — the global poor, the elderly, and the offline are systematically absent. Representativeness: a Twitter sample is not a population sample; the platform's algorithms shape what is visible, and the population that uses a given platform is not the population of any nation. The danger is that computational convenience displaces sampling theory and that the sheer volume of data masks systematic coverage gaps.

Causal inference in sociology

The causal-inference revolution, synthesized in Morgan and Winship's Counterfactuals and Causal Inference (2007; 3rd ed. 2015), reframed causal questions in terms of potential outcomes (the Neyman-Rubin model). For each individual, there is a potential outcome under treatment and a potential outcome under control; the causal effect is the difference between them. The fundamental problem is that only one outcome is observed — the other is counterfactual. Causal inference is the art of estimating what the unobserved outcome would have been.

When random assignment is impossible, sociologists rely on quasi-experimental designs: propensity score matching (constructing comparison groups by modeling the probability of treatment), instrumental variables (using a variable that affects treatment but not the outcome except through treatment), regression discontinuity designs (exploiting a sharp cutoff in treatment assignment), differences-in-differences (comparing changes over time between treated and untreated groups), and fixed-effects models (using within-unit variation over time to control for unobserved time-invariant confounders).

Directed acyclic graphs (DAGs), from Judea Pearl's framework, provide a visual and formal language for encoding assumptions about causal structure and for determining which variables must be controlled (to block confounding paths) and which must not be controlled (because controlling them opens biasing paths or blocks causal mediation).

Natural experiments in sociology exploit as-if-random variation: Angrist and Evans (1998) used the sex composition of a family's first two children as an instrument for having a third child (parents with two same-sex children are more likely to try again), estimating the effect of childbearing on women's labor supply. Card and Krueger (1994) compared fast-food employment in New Jersey (which raised its minimum wage) and neighboring Pennsylvania (which did not), finding no employment decline — a natural experiment challenging the textbook prediction that minimum wages reduce employment.

Historical-comparative revival and critical realism

The comparative-historical tradition has been revitalized by Mahoney and Rueschemeyer (2003) and by a methodological literature on sequence, timing, and path dependence that goes beyond Mill's methods. Analytic narratives (Bates et al., 1998) combine formal game-theoretic models with detailed historical narrative to explain specific outcomes. Adam Przeworski's work on democracy and development demonstrated that large-N statistical comparison and small-N case analysis can be complementary rather than competing.

Critical realism, developed by Roy Bhaskar and Margaret Archer, offers a philosophical foundation that steers between positivism and interpretivism. It holds that social structures are real — they have causal power whether or not individuals are aware of them — but are not directly observable in the way physical objects are. Social structures are tendencies, not laws; they are mediated through human agency and are therefore historically and culturally specific. Critical realism argues that sociology should seek the mechanisms that generate observed regularities, not merely the regularities themselves. A suicide rate, on this view, is a surface phenomenon; the mechanisms — alienation, integration breakdown, normative confusion — are the deeper causal structures that the rate makes visible.

Research ethics beyond Belmont

The Belmont Report's three principles — respect for persons (informed consent), beneficence (minimize harm, maximize benefit), and justice (fair distribution of burdens) — remain the foundation of human-subjects protection. IRB review operationalizes these principles through prospective evaluation of research protocols.

But Belmont was written for an era of face-to-face research with identifiable subjects. Three challenges strain the framework. First, big data: datasets assembled for commercial purposes (social media, health records, browsing histories) are repurposed for research without the data subjects' knowledge or meaningful consent. Michael Zimmer documented how a dataset that Netflix released as "anonymized" was re-identified by cross-referencing with public IMDb ratings, exposing individuals' viewing histories. The Belmont framework's assumption that subjects can be informed and can opt out breaks down when the subject never encounters the researcher.

Second, deception: some research requires concealment to avoid demand effects. Laud Humphreys's Tearoom Trade (1970), which traced participants in anonymous sexual encounters to their homes and interviewed them under false pretenses, remains the paradigmatic case of ethically untenable deception. The question is whether any knowledge justifies the violation of consent that deception entails, and whether the same knowledge could have been obtained through transparent methods.

Third, research with vulnerable populations: the history of research abuse — the U.S. Public Health Service's Tuskegee Syphilis Study (1932–1972), in which Black men were deliberately denied treatment so researchers could observe the progression of the disease, and the case of Henrietta Lacks, whose cancer cells were cultured and distributed for research without her or her family's knowledge or consent — demonstrates that "justice" in the Belmont sense has been systematically violated when the subjects were racial minorities, the poor, or the institutionalized. The HeLa cell line, derived from Lacks's tissue in 1951, became one of the most important tools in biomedical history; her family learned of it decades later and received no compensation. These cases are not historical curiosities; they shape the trust — or mistrust — that vulnerable communities bring to research today, and they demand that ethics review attend not only to procedural compliance but to the power relations embedded in who studies whom.

Connections Master

• Sociological imagination and research methods 30.01.01. This unit extends the methodological foundations sketched in the introductory unit. The five theoretical paradigms introduced there (functionalism, conflict theory, symbolic interactionism, feminist theory, postcolonial theory) each favor certain methods: functionalism and conflict theory lean toward large-scale survey and comparative-historical analysis; symbolic interactionism toward ethnography; feminist and postcolonial theory toward reflexive, participatory, and standpoint-informed methods.

• Classical and contemporary theory 30.01.03 pending. The research methods surveyed here are the tools through which theoretical claims are tested, refined, and challenged. The next unit covers the classical theoretical frameworks in depth; every theoretical tradition carries methodological commitments, and the choice of method is itself a theoretical act.

• Psychology research methods [29.01.01, 29.01.02]. Psychology and sociology share the experimental method, survey methodology, and the ethics infrastructure (Belmont, IRBs). The disciplines differ in their primary unit of analysis (the individual versus the group or institution) and in their reliance on qualitative methods, which are more central in sociology. Audit studies, survey experiments, and natural experiments are shared tools.

• Cultural anthropology and ethnography 31.02.01. Ethnography originated in anthropology (Malinowski, Boas, Evans-Pritchard) and was adopted by sociology through the Chicago School. The two disciplines' ethnographic traditions have converged and diverged in different periods; thick description, reflexivity, and the ethics of representation are shared concerns.

• Statistics [26, 45]. Survey sampling, sampling error, hypothesis testing, regression, and causal-inference techniques (propensity scores, instrumental variables, differences-in-differences) all draw on the probability and statistical theory covered in the statistics strand. The counterfactual framework and DAG-based causal reasoning formalized by Morgan and Winship are rooted in the statistical theory of potential outcomes.

• World history [32]. Comparative-historical sociology and world history share methods (archival research, process tracing, comparison across cases) and objects of study (revolutions, colonialism, state formation). Skocpol's and Tilly's work sits at the boundary between the two disciplines.

• Philosophy of science and epistemology [20.01]. The debate between positivism, interpretivism, and critical theory maps onto philosophical debates about the nature of knowledge, the possibility of objective inquiry, and the relationship between facts and values. Standpoint epistemology and the performativity of methods have deep roots in philosophy of science and social epistemology.

• Media literacy [36]. Content analysis and computational text analysis are the primary methods for studying media representations, framing, and discourse. The ability to evaluate media research depends on understanding how content is coded, how samples are drawn, and how computational methods introduce their own biases.

Historical and philosophical context Master

The methodenstreit and the founding split

Sociology's methodological identity was forged in a nineteenth-century debate known as the Methodenstreit (struggle over method). The question was whether the study of society should follow the methods of the natural sciences (the positivist position, associated with Comte and Durkheim) or require a distinct interpretive method suited to the fact that human action is meaningful (the verstehen position, associated with Weber and Dilthey).

Durkheim's The Rules of Sociological Method (1895) argued that social facts should be treated "as things" — external, coercive, and amenable to empirical investigation through statistics, comparison, and observation. Suicide (1897) was the demonstration: Durkheim showed that suicide rates, varying systematically by religion, marital status, and political context, were social facts explicable by the degree of social integration and moral regulation, not by individual psychology. This established the positivist program that dominated sociology's first half-century.

Weber's counter-position was not anti-empirical — he conducted massive historical-comparative studies — but methodologically distinct. Sociology, for Weber, is a science concerning itself with the interpretive understanding of social action in order to arrive at a causal explanation of its course and effects. The sociologist must grasp the subjective meaning that actors attach to their behavior, not merely record the behavioral regularity. Weber's ideal type — a constructed analytical model that exaggerates certain features to make them visible — was his primary tool for comparative analysis.

This founding split between explanation (Erklären) and understanding (Verstehen) has never been resolved. It reappears in every generation as a tension between quantitative and qualitative methods, between variable-oriented and case-oriented analysis, and between positivist and interpretive epistemology.

The Chicago School and the ethnographic tradition

American sociology's distinctive contribution was urban ethnography. The Chicago School of the 1910s–1930s — Robert Park, Ernest Burgess, William Foote Whyte, and others — treated the city as a "social laboratory" and developed participant observation as a systematic method. Whyte's Street Corner Society (1943) was a landmark: by living in an Italian-American slum district in Boston for three and a half years, Whyte showed that what outsiders saw as disorganized "slum" life was in fact structured by networks of obligation, status, and patronage. The study established that ethnography could produce rigorous, generalizable insights, not merely vivid description.

The Chicago tradition was complemented by W. E. B. Du Bois's earlier The Philadelphia Negro (1899), which combined door-to-door surveys with ethnographic observation and historical analysis — a mixed-methods design decades ahead of its time. Du Bois's methodological sophistication was matched by his theoretical originality, and his exclusion from the sociological canon (documented by Aldon Morris in The Scholar Denied, 2015) was a loss the discipline has only begun to repair.

The quantitative revolution and Parsonian grand theory

Mid-twentieth-century American sociology was dominated by two forces that pulled in opposite directions. Talcott Parsons's grand theory at Harvard constructed an abstract, system-level account of social order (The Social System, 1951) that critics found empirically untethered. Paul Lazarsfeld's survey methodology at Columbia pioneered the sample survey, statistical analysis, and the panel design, creating the tools of modern quantitative sociology but sometimes producing studies rich in data and thin in theory.

C. Wright Mills's The Sociological Imagination (1959) attacked both: "grand theory" for its abstraction and disconnection from real social problems, and "abstracted empiricism" for its accumulation of data without theoretical purpose. Mills argued that sociology's promise — connecting personal troubles to public issues — required methods adequate to the question, not commitment to a single methodological camp.

The methodological pluralism of the late twentieth century

The 1960s and 1970s saw a methodological opening. The civil rights, feminist, and anti-colonial movements challenged the discipline's assumption that knowledge produced by white male researchers in Western universities was universal. Feminist methodology (Smith, Harding, Oakley) questioned whether the detached, impersonal survey interview reproduced masculine modes of knowing and argued for methods that took women's lived experience as a starting point rather than a deviation from a male norm. Postcolonial sociology (Bhambra, Connell) challenged the Eurocentrism of the canon and the assumption that European and North American social patterns were universal.

Quantitative methods also advanced. The causal-inference framework (Neyman-Rubin potential outcomes, disseminated in sociology by Morgan and Winship) brought new rigor to non-experimental research. The development of multilevel modeling, event-history analysis, and network analysis expanded the range of social phenomena that could be formalized. The mixed-methods movement (Tashakkori and Teddlie; Creswell) argued that the quantitative-qualitative dichotomy was false and that integration, not segregation, was the methodological frontier.

The computational turn

The twenty-first century has brought a further expansion. The availability of massive digital datasets, the computing power to analyze them, and the machine-learning techniques to extract patterns have created computational sociology as a distinct subfield. Salganik's Bit by Bit (2018) codified its methods; the debates it provoked — about ethics, bias, representativeness, and the relationship between prediction and explanation — are the methodological frontier of the present moment. The performativity critique (Law and Urry) acquires new force when algorithms do not merely describe behavior but shape it through recommendation systems, predictive policing, and credit scoring.

Bibliography Master

1. Giddens, A. & Sutton, P. W., Sociology, 8th ed. (Polity, 2017). Comprehensive introductory textbook; Ch. 2 covers sociological research methods systematically.

2. Macionis, J. J., Sociology, 17th ed. (Pearson, 2019). Widely used introductory text; Ch. 1 situates the sociological perspective within research practice.

3. Geertz, C., "Thick Description: Toward an Interpretive Theory of Culture," in The Interpretation of Cultures (Basic Books, 1973), 3–30. The foundational statement of thick description and the interpretive program.

4. Small, M. L., "How Many Cases Do I Need? On Science and the Logic of Case Selection in Field-Based Research," Ethnography 10(1) (2009), 5–38. Argues that case selection in qualitative research should be theoretically motivated, not governed by a statistical rule.

5. Seawright, J. & Gerring, J., "Case Selection Techniques in Case Study Research: A Menu of Qualitative and Quantitative Options," Political Research Quarterly 61(2) (2008), 294–308. Systematizes the typology of case-selection strategies (typical, diverse, extreme, deviant, influential, most-similar, most-different).

6. Whyte, W. F., Street Corner Society: The Social Structure of an Italian Slum (University of Chicago Press, 1943). Landmark urban ethnography of Boston's North End.

7. Duneier, M., Sidewalk (Farrar, Straus and Giroux, 1999). Ethnography of street vendors in Greenwich Village; pairs observation with analysis of race, work, and public space.

8. Anderson, E., Code of the Street: Decency, Violence, and the Moral Life of the Inner City (W. W. Norton, 1999). Ethnography of the negotiation of respect and violence in Philadelphia.

9. Bourgois, P., In Search of Respect: Selling Crack in El Barrio (Cambridge University Press, 1995). Ethnography of structural violence and survival in East Harlem.

10. Desmond, M., Evicted: Poverty and Profit in the American City (Crown, 2016). Ethnographic and quantitative study of housing insecurity and eviction in Milwaukee.

11. Pager, D., "The Mark of a Criminal Record," American Journal of Sociology 108(5) (2003), 937–975. Audit study finding that a criminal record reduces job callbacks by roughly 50 percent, with a larger racial gap.

12. Skocpol, T., States and Social Revolutions: A Comparative Analysis of France, Russia, and China (Cambridge University Press, 1979). Comparative-historical analysis using Mill's methods.

13. Moore, B., Social Origins of Dictatorship and Democracy: Lord and Peasant in the Making of the Modern World (Beacon Press, 1966). Traces how agrarian class structures shaped political trajectories.

14. Tilly, C., Coercion, Capital, and European States, AD 990–1990 (Blackwell, 1990). Analysis of state formation through war, capital extraction, and contention.

15. Emirbayer, M., "Manifesto for a Relational Sociology," American Journal of Sociology 103(2) (1997), 281–317. Argues for a sociology of transactions rather than substances.

16. Salganik, M. J., Bit by Bit: Social Research in the Digital Age (Princeton University Press, 2018). Synthesis of computational social science methods and ethics.

17. Morgan, S. L. & Winship, C., Counterfactuals and Causal Inference: Methods and Principles for Social Research, 2nd ed. (Cambridge University Press, 2015). The standard reference for the potential-outcomes framework in sociology.

18. Angrist, J. D. & Evans, W. N., "Children and Their Parents' Labor Supply: Evidence from Exogenous Variation in Family Size," American Economic Review 88(3) (1998), 450–477. Uses sibling sex composition as an instrument for fertility.

19. Card, D. & Krueger, A. B., "Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania," American Economic Review 84(4) (1994), 772–793. Natural experiment challenging the prediction that minimum wages reduce employment.

20. Mahoney, J. & Rueschemeyer, D. (eds.), Comparative Historical Analysis in the Social Sciences (Cambridge University Press, 2003). Survey and defense of the comparative-historical tradition.

21. Glaser, B. G. & Strauss, A. L., The Discovery of Grounded Theory: Strategies for Qualitative Research (Aldine, 1967). Foundational statement of grounded theory and iterative qualitative coding.

22. Harding, S., The Science Question in Feminism (Cornell University Press, 1986). Develops standpoint epistemology and critiques of objectivity in science.

23. Smith, D. E., The Everyday World as Problematic: A Feminist Sociology (University of Toronto Press, 1987). Standpoint theory applied to sociological method and institutional analysis.

24. Law, J. & Urry, J., "Enacting the Social," Economy and Society 33(3) (2004), 390–410. Argues that social science methods are performative — they help create the realities they describe.

25. Habermas, J., Knowledge and Human Interests (Beacon Press, 1971 [1968]). Distinguishes technical, practical, and emancipatory knowledge-constitutive interests; foundational for critical theory.

26. Bhaskar, R., The Possibility of Naturalism: A Philosophical Critique of the Contemporary Human Sciences (Harvester Wheatsheaf, 1979; 2nd ed. Routledge, 1998). Foundational statement of critical realism.

27. Archer, M. S., Realist Social Theory: The Morphogenetic Approach (Cambridge University Press, 1995). Develops critical realism into a systematic theory of structure and agency.

28. Mills, C. W., The Sociological Imagination (Oxford University Press, 1959). Critique of grand theory and abstracted empiricism; defines the discipline's methodological promise.

29. National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research, The Belmont Report (DHEW Publication No. OS 78-0012, 1979). The foundational ethics framework: respect for persons, beneficence, justice.

30. Zimmer, M., "'But the Data Is Already Public': On the Ethics of Research in Facebook," Ethics and Information Technology 12(4) (2010), 313–325. Examines the ethics of using digital trace data and the limits of anonymization.

31. Skloot, R., The Immortal Life of Henrietta Lacks (Crown, 2010). Documents the HeLa cell line case and its implications for research ethics and consent.

32. Morris, A., The Scholar Denied: W. E. B. Du Bois and the Birth of Modern Sociology (University of California Press, 2015). Documents Du Bois's methodological contributions and his exclusion from the canon.