Ethnography and fieldwork methods

Anthropology (BA / B.Sc) · Prehistory and methods · free study notes for exams, placements and interviews

In short

Ethnography is both a method and a written product: the anthropologist lives among a community for an extended period, learns its language, participates in its daily life, and produces a detailed account of its culture from the inside. It is the discipline's defining method — what separates anthropology from survey-based social science — and rests on long-term participant observation rather than one-shot data collection. University papers on "Fieldwork and Research Methods" (UGC CBCS core course in BA/B.Sc Anthropology, and UGC-NET Anthropology Unit on research methodology) examine the techniques, the Malinowskian tradition, Indian village studies and ethical codes. Professionally, the same toolkit is used by UX and design researchers, market and rural-marketing researchers, development and CSR evaluators, public-health teams and NGOs across India.

Explained

What ethnography is, and where it came from

Definition: systematic, first-hand, long-term study of a living community's culture, written up holistically (*ethnos* = people, *graphein* = to write).
Ethnography / ethnology / anthropology: ethnography is the descriptive account of one group; ethnology compares such accounts across groups; social-cultural anthropology is the theory built on both.
Armchair anthropology: nineteenth-century evolutionists (Morgan, Tylor, Frazer) theorised from travellers', missionaries' and administrators' reports without going to the field.
Verandah anthropology: the researcher stays at the colonial station and has informants brought in — observation without immersion.
Malinowski (Trobriand Islands, 1914-18) set the modern standard in *Argonauts of the Western Pacific*: pitch your tent in the village, stay one to two years, learn the vernacular, record the "imponderabilia of actual life", and distinguish (a) the skeleton of institutions, (b) actual behaviour, (c) what people say and think.
Radcliffe-Brown (Andaman Islanders) brought structural-functional analysis; Boas in America insisted on the native language and cultural relativism; Evans-Pritchard (Nuer, Azande) refined kinship and translation of belief.
India: survey-and-census tradition (Risley, Thurston, Russell) gave way to intensive village and tribe studies — S. C. Roy (Oraon, Munda; "father of Indian ethnography"), N. K. Bose, D. N. Majumdar (Mohana village, Lucknow), M. N. Srinivas (Rampura; *The Remembered Village*, concepts of sanskritisation and dominant caste), S. C. Dube (Shamirpet), Verrier Elwin (Baiga, Gond), L. P. Vidyarthi (Maler; the "sacred complex" at Gaya), Irawati Karve (kinship, Maharashtra).
Phases of a project: (1) problem formulation and literature review → (2) research design, tools, sampling → (3) rapport establishment → (4) data collection → (5) analysis and interpretation → (6) report/monograph writing.

Preparing and entering the field

Review of literature and secondary sources — earlier monographs, District Census Handbooks, Census of India tables, *People of India* volumes (Anthropological Survey of India), Tribal Research Institute reports, revenue and forest records, gazetteers.
Selection of the field — relevance to the problem, manageable size, accessibility, language the researcher can learn, and whether permission is obtainable.
Permissions: for Scheduled Areas and PVTG habitations, clearance from state Tribal Welfare / Tribal Research Institute and district administration; entry into the Andaman & Nicobar tribal reserves is restricted under the Andaman and Nicobar Islands (Protection of Aboriginal Tribes) Regulation, 1956; foreign scholars need a research visa and, for certain areas, a Protected/Restricted Area Permit. Requirements vary by state and change — confirm the current rule with the concerned department.
Preparation: language primer, maps, schedules pre-tested in a pilot study, notebooks, camera/recorder with consent protocol, health precautions, budget (typical student fieldwork of 4-8 weeks costs roughly ₹20,000-₹60,000 depending on distance and stay; UGC/ICSSR and university field-work grants part-fund it).
Rapport establishment — the single most decisive step: approach through a respected intermediary or the headman/panchayat, state your purpose honestly, attend public events, avoid taking sides in factional disputes, reciprocate (medicines, literacy help, transport) without creating dependence.
Role the researcher takes: complete participant (covert), participant-as-observer, observer-as-participant, complete observer. Covert roles are now generally held to be unethical except where no harm and no alternative.
Gatekeepers, key informants and marginal-native position: living "in" yet analysing from "outside"; the anthropologist is a *stranger* whose outsider status is itself a research asset.

The core techniques

Observation — non-participant (watching without involvement; good for public rituals, markets, queues) and participant observation (joining activities: harvesting, fishing, festivals, household work). Structured observation uses a checklist/schedule; unstructured observation is open-ended.
Interviewing — *structured* (fixed wording and order, as in a schedule), *semi-structured* (interview guide; standard in practice), *unstructured/in-depth* (conversational), *focused* (on one event), *group*, and key-informant interviewing (a knowledgeable insider — priest, midwife, headman, schoolteacher — interviewed repeatedly).
Schedule vs questionnaire: a schedule is filled by the investigator during a face-to-face interview (works with non-literate respondents, allows probing, high response rate); a questionnaire is filled by the respondent (cheap, wide coverage, low response rate, needs literacy).
Case study method — exhaustive study of one unit (person, family, lineage, village, dispute, event) in its context; yields depth and hypotheses, not generalisation.
Genealogical method (Rivers) — collecting pedigrees with Ego, standard kin symbols (triangle = male, circle = female, equals sign = marriage, vertical line = descent) and vernacular kin terms; used to map descent, marriage rules, inheritance and residence.
Life history / narrative — a single person's account across the life cycle; good for change, migration, conversion, personal experience of caste or gender.
Village/household survey and census — house-listing, household schedule, demographic data; gives the quantitative frame within which qualitative material sits.
Social and physical mapping — settlement map by caste/clan ward, resource map, sociometry (who interacts with whom), social network analysis.
Focus group discussion (FGD) — 6-12 relatively homogeneous participants, a moderator with a topic guide and a note-taker; 60-90 minutes; reveals shared norms and disagreement, not individual prevalence.
Participatory methods (PRA/RRA/PLA) — transect walk, seasonal calendar, daily activity clock, wealth ranking, Venn/chapati diagram, problem tree, participatory mapping; widely used in Indian rural development work.
Projective and psychological tests — Rorschach, TAT, draw-a-picture — used in culture-and-personality studies; validity across cultures is contested.
Audio-visual and visual anthropology — photography, film, video; ethnobotanical/material-culture collection with provenance records.
Documentary/archival work — land records, temple inscriptions, court and police records, school registers, panchayat minutes.
Digital/online ethnography (netnography) — studying WhatsApp groups, forums and gaming communities; raises fresh consent and privacy issues.
Field notes: jottings in the field → expanded descriptive notes the same day → methodological notes (how data was got) → analytical/theoretical memos → personal diary. Rule: write up every day; memory decays within hours.
Triangulation: cross-check by method (observation + interview + records), by source (different informants), by investigator, and by theory.

Sampling, data types and analysis

Qualitative sampling is purposive, not random: purposive/judgemental, quota, snowball/chain-referral (for hidden populations — sex workers, migrants, drug users), theoretical sampling (continue until *saturation*), extreme-case, typical-case, maximum-variation.
Probability sampling (simple random, systematic, stratified, cluster, multistage) is used for the household survey component where population estimates are needed.
Universe, sampling frame, sampling unit, sample size, sampling error, non-sampling error — define each in the report.
Data types: primary (your observation and interviews) vs secondary (census, records, earlier monographs); qualitative (text, image) vs quantitative (counts, measurements); nominal, ordinal, interval and ratio scales.
Emic vs etic: emic = categories and explanations of the insiders; etic = the analyst's comparative, scientific categories. Good ethnography reports both and keeps them distinct.
Analysis: transcription → coding (open, axial, selective) → themes and categories → thick description (Geertz: describing an act plus the web of meaning that makes it intelligible) → comparison with theory. Software: NVivo, ATLAS.ti, MAXQDA, Dedoose; Excel/SPSS/R for the quantitative part.
Rigour in qualitative work (Lincoln and Guba) — credibility (prolonged engagement, member checking), transferability (thick description), dependability (audit trail), confirmability (reflexive account) — the qualitative counterparts to validity, generalisability, reliability and objectivity.
Reliability and validity: ethnography scores high on validity (you see actual behaviour) and low on reliability/replicability (another worker in another decade meets a changed community). The Redfield-Lewis controversy over Tepoztlán is the classic illustration.
Reflexivity: the researcher's caste, class, gender, religion, language and funding shape access and data, and must be stated, not hidden; auto-ethnography turns that lens fully on the self.

Advantages, limitations and comparisons

Advantages: holistic and contextual; captures what people *do*, not only what they say; discovers the unexpected and the unspoken; works with non-literate populations; strong for meaning, process and local categories; generates hypotheses.
Limitations: time-consuming and costly; small, non-representative units; findings not statistically generalisable; heavy observer effect (Hawthorne effect) and interviewer bias; going native (over-identification, loss of analytic distance) or culture shock at the other extreme; dependence on a few key informants and on interpreters; ethical exposure of a small identifiable community; difficult to replicate.
Common pairings to know: emic/etic, qualitative/quantitative, synchronic (one point in time)/diachronic (through time), re-study (returning after years, as Srinivas's and later Rampura work), longitudinal/cross-sectional, insider (indigenous) /outsider research.

Ethics, law and current Indian practice

Core principles: voluntary informed consent (oral consent properly recorded is acceptable where literacy is low), right to withdraw, anonymity and confidentiality (pseudonyms for persons and often for the village), no harm, no deception, honest reciprocity, feedback of findings to the community, and no misuse of data against the community.
Codes: the ICMR *National Ethical Guidelines for Biomedical and Health Research Involving Human Participants* (2017 revision; separate ICMR guidance covers social and behavioural research and research with tribal populations) governs health-related work and requires Institutional Ethics Committee clearance. Many universities apply it to social research too. International codes commonly cited: AAA Statement on Ethics, ASA (UK) ethical guidelines; India's own body is the Indian Anthropological Association.
Biological samples and genetic data: ICMR rules, and for international collaboration the relevant government clearance, apply; the Biological Diversity Act, 2002 (as amended in 2023) and its Rules govern access to biological resources and associated traditional knowledge, with benefit-sharing through the NBA/State Boards.
Forest and tribal rights that shape field relations: the Scheduled Tribes and Other Traditional Forest Dwellers (Recognition of Forest Rights) Act, 2006; PESA, 1996; the SC/ST (Prevention of Atrocities) Act, 1989 — a researcher must know these before writing about land or disputes.
Personal data: the Digital Personal Data Protection Act, 2023 is enacted and the DPDP Rules, 2025 are notified, but the Data Fiduciary duties (notice, consent, breach intimation, erasure) take effect only from 13 May 2027; until then the Information Technology Act s.43A and the SPDI Rules, 2011 govern sensitive personal data held digitally. Research-related exemptions depend on standards to be prescribed, so handle recordings and identifiers as sensitive personal data.
Institutions you may be examined on: Anthropological Survey of India (*People of India* project), Indian Council of Social Science Research (ICSSR; field-work and doctoral fellowships), Tribal Research Institutes in the states, National Commission for Scheduled Tribes, Ministry of Tribal Affairs, Census of India and NSSO/NSO as secondary-data sources.

Key points

Ethnography - long-term, first-hand, holistic study and written account of a living community's culture.
Participant observation - living and taking part in daily life while systematically observing; anthropology's signature technique.
Malinowski's canon - pitch your tent in the village, learn the vernacular, stay a full annual cycle, record the imponderabilia of everyday life.
Armchair / verandah anthropology - theorising from others' reports, or from the station without immersion; both superseded.
Rapport - the trust that makes honest data possible; built through an intermediary, transparency, participation and reciprocity.
Key informant - a knowledgeable insider interviewed repeatedly; indispensable but a source of bias if relied on alone.
Schedule vs questionnaire - schedule is investigator-filled face to face; questionnaire is respondent-filled.
Genealogical method - Rivers's pedigree technique with kin symbols and vernacular kin terms; the basis of kinship analysis.
Case study - exhaustive study of a single unit; depth and hypothesis generation, not statistical generalisation.
FGD - 6-12 homogeneous participants, moderator plus note-taker; captures shared norms, not individual prevalence.
PRA tools - transect walk, seasonal calendar, wealth ranking, Venn and resource mapping; community-generated data.
Emic vs etic - insider's categories vs analyst's comparative categories; report both separately.
Thick description - Geertz: describe the act plus the layers of meaning that make it intelligible.
Triangulation - cross-checking across methods, sources, investigators and theories to raise credibility.
Purposive and snowball sampling - judgement-based selection; snowball reaches hidden populations; stop at saturation.
Going native and Hawthorne effect - loss of analytic distance, and behaviour changing because it is watched.
Reflexivity - stating how the researcher's own identity and funding shaped access and interpretation.
Ethics - informed (often oral) consent, anonymity, no harm, benefit to the community, IEC clearance under ICMR guidelines.

Formulas and examples

Sample size for a proportion (large population): n = Z² p (1 − p) / e²
where Z = standard normal value for the confidence level (1.96 at 95%), p = expected proportion, e = margin of error (as a decimal), n = number of respondents.
Finite-population correction: n_adj = n / [1 + (n − 1)/N], N = population size.
Sex ratio = (number of females / number of males) × 1,000
Dependency ratio = [(population aged 0-14) + (65 and above)] / (population aged 15-64) × 100
Literacy rate = (literates aged 7 and above / population aged 7 and above) × 100
Inter-coder (inter-observer) agreement, Cohen's kappa: κ = (Pо − Pe) / (1 − Pe), Pо = observed agreement, Pe = agreement expected by chance; κ above about 0.75 is good.
Proportional allocation in stratified sampling: n_h = n × (N_h / N)

Example 1 - how many households to survey. A student studying a block of 4,000 households expects about 50% to report seasonal migration, and wants a 95% confidence level with a 7% margin of error.

n = 1.96² × 0.5 × 0.5 / 0.07² = 3.8416 × 0.25 / 0.0049 = 0.9604 / 0.0049 ≈ 196.
Finite-population correction: n_adj = 196 / [1 + (195/4,000)] = 196 / 1.04875 ≈ 186.9 → 187 households.
At roughly 8 schedules a day, that is about 24 field days for one investigator.

Example 2 - stratified allocation across caste wards. Village of 600 households: SC ward 180, OBC 240, General 150, ST 30. Sample of 120 by proportional allocation.

SC: 120 × 180/600 = 36; OBC: 120 × 240/600 = 48; General: 120 × 150/600 = 30; ST: 120 × 30/600 = 6. Total = 36 + 48 + 30 + 6 = 120.
The ST stratum gets only 6 — too few for separate analysis, so over-sample it (say to 20) and use weights when reporting village-level figures.

Example 3 - demographic ratios from a house-to-house census. A hamlet records 512 persons: 243 females, 269 males; ages 0-14 = 148, 15-64 = 326, 65+ = 38; persons aged 7+ = 449, of whom 287 are literate.

Sex ratio = 243/269 × 1,000 ≈ 903 females per 1,000 males (below the 2011 Census all-India figure of 943).
Dependency ratio = (148 + 38)/326 × 100 = 186/326 × 100 ≈ 57.1%.
Literacy rate = 287/449 × 100 ≈ 63.9%.
Note 148 + 326 + 38 = 512, so the age data reconciles with the headcount — always show this check in the report.

Example 4 - observer agreement. Two students independently code 80 recorded interactions as "ritual" or "secular" and agree on 68. Of the 80, observer A called 50 ritual and B called 46 ritual.

Pо = 68/80 = 0.85.
Pe = (50/80 × 46/80) + (30/80 × 34/80) = (0.625 × 0.575) + (0.375 × 0.425) = 0.359375 + 0.159375 = 0.51875.
κ = (0.85 − 0.51875)/(1 − 0.51875) = 0.33125/0.48125 ≈ 0.688 — moderate-to-good; the codebook needs sharper definitions before full coding.

Example 5 - time budget from a daily activity clock. Ten women record a 24-hour day: wage work 5.2 h, domestic work 6.4 h, fuel and water collection 2.1 h, child care 1.8 h, sleep 7.0 h, leisure 1.5 h.

Total = 5.2 + 6.4 + 2.1 + 1.8 + 7.0 + 1.5 = 24.0 h — the clock closes, so no activity is missing.
Unpaid work = 6.4 + 2.1 + 1.8 = 10.3 h = 10.3/24 × 100 ≈ 42.9% of the day against 5.2 h (21.7%) of paid work — the standard finding that ethnographic time-use data makes visible and a household income survey misses.

Feature	Ethnography (qualitative)	Survey (quantitative)
Unit	One community or a few cases	Large representative sample
Duration	Months to years	Days to weeks
Sampling	Purposive, snowball, saturation	Probability sampling
Tools	Participant observation, in-depth interview, genealogy	Questionnaire, structured schedule
Output	Thick description, monograph	Percentages, tables, significance tests
Strength	Validity, meaning, context, discovery	Reliability, generalisability, measurement
Weakness	Not generalisable, observer effect	Misses meaning, context and the unspoken

Technique	Best for	Main risk
Participant observation	Actual behaviour, ritual, work	Going native, observer effect
In-depth interview	Beliefs, motives, life events	Recall and prestige bias
Key informant	Rules, history, local terms	One person's version taken as the culture
FGD	Shared norms, vocabulary	Dominant speaker, no private views
Household schedule	Counts, comparison, base data	Misses meaning; needs rapport first
PRA tools	Community priorities, local knowledge	Elite capture of the exercise

Where it is used

UX and design research in Indian product firms (Flipkart, Zomato, Paytm, PhonePe, Swiggy, Microsoft and Google India) — contextual inquiry, home visits and diary studies are participant observation renamed; roles titled User Researcher or Design Researcher.
Market and rural-marketing research (Kantar, IPSOS, Nielsen, Hansa Research) — ethnographic "immersions", shop-along studies, FGDs and household panels for FMCG and telecom clients.
Development and CSR monitoring and evaluation — PRA-based baselines and impact studies for NRLM/SHG programmes, Jal Jeevan Mission, nutrition and sanitation projects; employers include PRADAN, BAIF, SEWA, Sambodhi, IPE Global, Deloitte and KPMG social-sector teams.
Public health and nutrition — ICMR institutes, NIHFW, PHFI and district health teams use ethnographic methods for immunisation hesitancy, TB treatment adherence, maternal care practices and tribal health; qualitative work is standard in NFHS-linked and PIP studies.
Government and policy — Anthropological Survey of India, Tribal Research Institutes, Tribal Sub-Plan and PVTG baseline surveys, social-impact assessment for land acquisition and for FRA community forest-rights claims, museum and documentation work.
Academia and civil society — university and ICSSR research projects, journalism and documentary film, heritage and language documentation, litigation support for displacement and rehabilitation cases.

Interview questions

Q: What makes ethnography different from a questionnaire survey? A: Ethnography studies a small community over a long period through participation and open-ended enquiry to understand meaning and actual behaviour; a survey measures stated responses from a large representative sample at one point in time.
Q: Why is Malinowski called the founder of modern fieldwork? A: His Trobriand work (1914-18) established that the anthropologist must live in the community for a long stretch, learn its language and record everyday life directly, replacing armchair and verandah anthropology.
Q: Distinguish emic and etic perspectives. A: Emic is the insiders' own categories and explanations; etic is the analyst's external, comparative framework. A sound ethnography records the emic account faithfully and then analyses it etically.
Q: What is rapport and how would you build it in a tribal hamlet in two weeks? A: Rapport is working trust. Enter through the headman or a respected local worker, state the purpose honestly, attend public events, begin with harmless mapping and household listing, reciprocate practically, avoid factional sides and start sensitive questions only later.
Q: What is the difference between a schedule and a questionnaire? A: A schedule is filled by the investigator in a face-to-face interview, so it suits non-literate respondents and allows probing; a questionnaire is self-administered, cheaper and wider but has low response rates.
Q: How do you sample in qualitative research when random sampling is not possible? A: Purposively — selecting cases for their relevance — supplemented by snowball sampling for hidden groups and maximum-variation sampling for range, continuing until new interviews stop adding themes (saturation).
Q: What is "going native" and why is it a problem? A: Over-identifying with the community to the point of losing analytic distance and critical judgement; the researcher stops questioning what insiders take for granted and the account becomes advocacy rather than analysis.
Q: How do you establish the credibility of findings based on one village? A: Through prolonged engagement, triangulation across methods and informants, member checking with the community, an audit trail of field notes, a reflexive statement, and thick description so readers can judge transferability — not through statistical generalisation.
Q: What does triangulation mean and give an example? A: Cross-checking the same question by different routes — for instance verifying a claimed landholding by interview, by the panchayat land record and by a transect walk with the owner.
Q: What consent process would you follow with a non-literate respondent? A: Explain purpose, use, risks, anonymity and the right to withdraw in the local language, take oral consent witnessed and recorded (audio or by a witness's signature), get separate permission for photographs and recordings, and clear the protocol with the Institutional Ethics Committee.
Q: Which Indian village studies would you cite as classics of ethnographic method? A: Srinivas's Rampura (*The Remembered Village*), S. C. Dube's Shamirpet, D. N. Majumdar's Mohana, L. P. Vidyarthi's Maler and sacred-complex study at Gaya, and S. C. Roy's Oraon and Munda monographs.
Q: How would you use ethnographic methods for a product-research brief in a tier-3 Indian town? A: Spend time in homes and shops observing actual use rather than asking preferences, run a few FGDs and shop-alongs to get local vocabulary, triangulate with a short structured schedule for counts, and report verbatim quotes with a usage-pattern typology.

Common mistakes

Writing that ethnography is "just qualitative" and offering no sampling logic, no triangulation and no stated criteria of rigour.
Confusing schedule with questionnaire, and non-participant with participant observation, in short-answer questions.
Treating a key informant's account as the culture, or generalising one village's findings to a district, state or the country.
Starting sensitive enquiry (land, caste, sex, income, witchcraft) before rapport, which produces polite fiction rather than data.
Postponing field notes — writing from memory days later loses the detail and the verbatim phrasing that make thick description possible.
Naming the village and identifiable persons, publishing photographs without specific consent, and skipping ethics-committee clearance.
Reporting arithmetic that does not reconcile — age groups not summing to the headcount, a time-use clock not closing at 24 hours, sex ratio inverted (it is females per 1,000 males in India).