The Interview

The conduct of an interview

The stages are examinable and they are practical.

1. Preparation. Know the subject; frame the guide; decide the sample; obtain permission.
2. Rapport. Establishing enough trust and ease for the respondent to speak freely. This is the interviewer's principal skill, and without it the rest is worthless.
3. The opening: explaining who the interviewer is, what the study is for, how the information will be used, and that it is confidential. This is informed consent, and it is also good technique.
4. Questioning: beginning with easy and non-threatening questions, moving to the difficult ones once rapport is established, and keeping sensitive matters for late.
5. Probing: following an incomplete answer without suggesting its content. This is the hardest skill and the one most often done badly.
6. Listening and observing. Hesitation, discomfort and inconsistency are data.
7. Recording: by notes during, by notes immediately after, or by recording with consent. Each has costs, and recording during the interview inhibits some respondents while notes afterwards lose accuracy.
8. Closing, with an opportunity for the respondent to add anything and to ask questions.

Merits

1. It reaches attitudes, motives, opinions and experience, which observation cannot.
2. Flexibility. A question can be repeated, rephrased or explained, so misunderstanding can be corrected on the spot.
3. It reaches the illiterate, which a questionnaire cannot, and this is decisive in Indian conditions.
4. High response rate, far higher than a postal questionnaire.
5. Non-verbal information: hesitation, discomfort, the presence of others, the state of the household.
6. Depth and detail, and the discovery of the unanticipated.
7. Verification on the spot, since an inconsistent answer can be pursued.

Limitations

1. Interviewer bias, treated in the next section, which is the central problem.
2. Cost and time. Interviewing is the most expensive way of collecting data from many people.
3. Small coverage, and therefore weaker generalisation than a survey.
4. It records what the respondent says. Memory fails, and people misreport their own conduct, particularly where it is disapproved.
5. Social desirability. Respondents give the answer they think is acceptable, which is worst on precisely the sensitive subjects.
6. Access and refusal. Those who refuse differ systematically from those who agree.
7. The presence of others. In Indian households an interview is frequently conducted with family members present, and a woman's answers about her own household are not independent of who is listening. This is a serious and under-discussed source of error in Indian survey work.
8. Recording and analysis of unstructured material is laborious and interpretive.

Interviewer bias, in detail

The ways the interviewer changes the answer, and each has a remedy.

By who they are. The respondent adjusts their answer to the interviewer's apparent caste, class, gender, age, language and religion. On questions about caste practice, communal attitudes or domestic violence this effect is large. Remedy: match interviewers to respondents where the subject requires it, and report who conducted the interviews.

By how they ask. Leading questions, tone, emphasis and expression suggest an expected answer. Remedy: standardised wording, training, and pilot testing.

By what they probe. Pursuing answers that fit the expectation and letting others pass. Remedy: written probing rules, and supervision.

By what they record. Selective and interpretive note-taking. Remedy: verbatim recording where possible, and immediate transcription.

By what they expect. The interviewer who believes they know the answer finds it. Remedy: interviewers who do not know the hypothesis, and second coders.

Interview and questionnaire compared

The comparison belongs partly here and partly in the next chapter, and the short form is: an interview reaches depth, the illiterate and the unanticipated, at high cost and with interviewer bias; a questionnaire reaches breadth cheaply and anonymously, and gets shallow answers from literate volunteers only. The full table is in [The Questionnaire and the Schedule].

A worked example

The question: whether payments were made at the respondent's marriage.

Why this is hard. The practice is prohibited by the Dowry Prohibition Act 1961, so an accurate answer is an admission of an offence; it is socially disapproved in the abstract and normal in practice; and the answer differs according to who is in the room.

How the design must respond.

Wording: the question is not asked as "did you give dowry", which invites denial, but through the customary vocabulary the community itself uses for such transfers, and in terms of what is usual in the community before what happened in this household.

Who asks: a woman respondent will answer differently to a man, and differently again if her mother-in-law is present. Privacy is a methodological requirement here and not a courtesy.

Assurance: confidentiality must be explained and must be real, and the study must not record identifying details it does not need.

Triangulation: the interview is checked against other sources, since self-report on this subject is known to understate.

And the limit must be stated in the report. A study of this kind can establish patterns and directions and cannot produce a reliable absolute figure. Saying so is honest research; presenting the number as though it were a count is not, and it is the difference [Social Research: Nature and Purpose] insists on.

Quick revision

1. Interview: purposeful conversation for data collection; reaches attitudes, motives and experience that observation cannot.
2. By structure: structured (fixed wording and order, comparable, rigid); unstructured (a subject rather than questions, deep, uncomparable); semi-structured, the usual compromise.
3. By purpose: focused (Merton, on a known experience), depth, repetitive, diagnostic.
4. Group interviews reach the collective view; their danger is dominance and the suppression of private views.
5. Conduct: preparation, rapport, opening and consent, easy questions first, probing without suggesting, listening for non-verbal data, recording, closing.
6. Merits: attitudes and motives, flexibility, reaches the illiterate, high response, non-verbal data, depth, on-the-spot verification.
7. Limits: interviewer bias, cost, small coverage, reliance on report and memory, social desirability, refusal, and the presence of others, which is acute in Indian households.
8. Interviewer bias operates through who asks, how they ask, what they probe, what they record and what they expect; each has a remedy.

Test yourself

1. Distinguish structured, unstructured and semi-structured interviews. In a structured interview the questions, their wording and their order are fixed in advance and identical for every respondent, which makes answers comparable and countable but permits nothing unanticipated to be pursued. In an unstructured interview the interviewer has a subject rather than a set of questions and follows the respondent's own account, which yields depth and discovery at the cost of comparability and demands considerable skill. A semi-structured interview uses a guide of topics with freedom of wording and order, and is the commonest form in practice.

2. What is a focused interview? Merton's form, in which the respondent is known to have had a particular experience and the interview concentrates on the effect of that experience upon them, using a guide derived from an analysis of the situation the respondent went through. It is the ancestor of the focus group, and it is used where the researcher already knows what the respondents have in common and wishes to learn what it meant to them.

3. State the merits of the interview over the questionnaire. It reaches attitudes, motives and experience; it is flexible, so a misunderstood question can be rephrased or explained on the spot; it reaches the illiterate, which is decisive in Indian conditions; it obtains a far higher response rate; it yields non-verbal information such as hesitation, discomfort and the state of the household; it produces depth and detail, including matters the researcher did not think to ask about; and an inconsistent answer can be pursued immediately.

4. Describe interviewer bias and its remedies. It operates through who the interviewer is, since respondents adjust answers to the interviewer's apparent caste, class, gender and language; through how questions are asked, since tone, emphasis and leading wording suggest expected answers; through what is probed, since answers fitting expectation are pursued and others let pass; through what is recorded, since note-taking is selective; and through what the interviewer expects to find. The remedies are matching interviewers to respondents, standardised wording, training and piloting, written probing rules and supervision, verbatim recording, and keeping interviewers unaware of the hypothesis.

5. Why is the presence of other people a serious problem in Indian survey work? Because interviews in households are frequently conducted with family members present, and answers are not independent of who is listening. A woman asked about the division of domestic work, about her control over money, about violence in the home or about transfers made at her marriage will answer differently in the presence of her husband or her mother-in-law. Privacy is therefore a methodological requirement rather than a courtesy, and a study that does not report the conditions in which interviews were conducted has withheld information necessary to assess it.