Questionnaires and interviews: designing what people can say
By UK GCSE Sociology · independent study · myAudience
Original OpenAI imagegen conceptual teaching illustration, generated 4 October 2026; not observational evidence. · AI-generated original asset for this independent guide; no third-party artwork intentionally reproduced.
Questionnaires and interviews: designing what people can say
A questionnaire supplies questions for respondents to answer, often independently. A structured interview uses a standard set of questions delivered by an interviewer. Semi-structured interviews use a planned topic guide with follow-up questions, while unstructured interviews allow the conversation to develop more flexibly. Closed questions can support comparable categories; open questions can reveal meanings not anticipated by the researcher. No method owns a single data type. A structured interview may produce counts, and a questionnaire may contain rich written accounts. Choose the instrument to fit the inference, population and topic rather than assign automatic superiority to numbers or conversation.
Question wording shapes the evidence. 'How much has the excellent new library improved your reading?' assumes excellence and improvement. 'Are the opening hours convenient and the staff helpful?' combines two questions that can have different answers. Rewrite with neutral, separate questions and a defined period. Response options should cover relevant possibilities without overlap, including not applicable or prefer not to answer when appropriate. 'Often' may mean different frequencies to different people. Precise categories improve comparability but can still fail to capture an experience. A pilot checks how questions are understood, not whether respondents provide the preferred answer.
Standardisation can improve reliability by making wording and order more consistent. It does not guarantee validity: a repeated question about attendance might consistently mismeasure belonging. Self-completion can reduce pressure from an interviewer, yet literacy, language, access and willingness to write may affect who responds and what they say. Interviewers can clarify meaning and follow unexpected issues, but their wording, position and reactions can influence answers. Social desirability means respondents may present themselves in ways they believe are acceptable. It is a possible context-dependent mechanism, not a blanket accusation that all respondents lie.
An interviewer studying school support faces a power problem if learners think criticism could reach teachers. A teacher conducting the interview may have especially strong influence. Privacy, explanation of use and separation from assessment can help, but an assurance alone cannot establish freedom from pressure. Rapport may encourage detail while also producing answers tailored to please the interviewer. Record reflexive notes about the relationship and seek contrary accounts. A flexible conversation can have high interpretive value if its analysis is transparent; it need not be identical in every interview to be informative, though comparing differently prompted answers requires caution.
Coding translates responses into analytical categories. For a fictional open question about a library, 'I cannot arrive before closing' might be coded as timing, while 'the building entrance is difficult to use' could be access. A response can have more than one code. Create definitions, preserve examples and revise a codebook when unexpected themes appear. Count coded responses only with a clear denominator and rules for multiple coding. If ten responses mention two themes each, their percentages may legitimately total more than 100%. Counts of coded comments describe the responses, not necessarily the prevalence of experiences among all users.
Method choice involves practical trade-offs. A postal or online questionnaire may reach many people at lower interview time per response but can require accessible formats and follow-up. In-depth interviews allow exploration of contradictory meanings but take time to arrange, conduct and analyse. Cost savings that exclude relevant participants weaken the conclusion. Mixed methods can pair a broad survey with interviews to investigate how categories are interpreted. The gain is conditional: two poorly designed instruments do not become sound simply because they are combined. Use fictional responses here; do not ask peers for sensitive family or school disclosures.
Worked example: Original hypothetical instrument: 'Do responsible families always help with homework? Yes/No.' A researcher reports that this measures parenting quality. Rewrite the instrument and evaluate a questionnaire versus a semi-structured interview for understanding support without ranking family worth.
Identify the loaded term responsible, the absolute always and the mismatch between homework help and the broad concept of parenting quality.
Ask about defined forms of support in a stated period: discussing a task, providing time or arranging another source of help. Include no assigned homework and prefer-not-to-answer options.
A questionnaire can compare specified responses consistently. A semi-structured interview can explore why support takes different forms and what respondents regard as useful.
Address social desirability, privacy, varied interpretations and practical costs. Report evidence about support indicators rather than a moral score for families.
Answer: The original question cannot validly measure parenting quality. It combines judgment with an unrealistic absolute and assumes one form of support has a single meaning. A neutral instrument could ask which kinds of homework support were available during a specified week, allowing appropriate omissions. Standardised responses could support comparisons within the sample, while semi-structured interviews could explain constraints and unexpected forms of support. Both rely on accounts shaped by context. The evidence should therefore concern defined support practices, with limitations, rather than label families as responsible or irresponsible.
Check your understanding: Transfer task: six fictional respondents answer an open library question. Four mention timing, three mention transport, and two mention both. Can a report present 67% timing and 50% transport? Explain the denominator, rounding and multiple coding; then propose a question that distinguishes access from satisfaction. Indicative rubric: 10 points: calculations 2, coding logic 3, neutral instrument 3, inference limit 2.
Answer: Yes, if six respondents are the denominator and multiple themes can be coded: 4/6 is about 67% and 3/6 is 50%. Two appear in both categories, so totals above 100% do not imply an arithmetic error. Ask separately whether users could visit at a needed time and how satisfied they were with opening hours, including a not-applicable option. These proportions describe six fictional responses, not all library users.
Why: The calculation alone earns limited credit: the explanation must make overlapping categories explicit. A strong question separates opportunity from evaluation and defines a period. Reward an inference limit about the achieved sample or coded themes. The data are internally possible: two both, two timing only, one transport only and one neither account for all six respondents.