4.2.3.1 - Questionnaire and interview design

From aim to item

A questionnaire is not just a list of questions. In psychology, it is a measuring instrument. Each item should help the researcher collect evidence about the research aim without confusing, pressuring or leading the participant.

The first design move is to translate the aim into measurable questions. If the aim is "to investigate students' stress during exam season", the questionnaire should not ask vague questions such as "Are exams bad?" It should ask items that target the construct of stress more precisely, such as hours of sleep, perceived pressure, physical symptoms, coping strategies or confidence about revision.

Good questionnaire items are:

• relevant to the aim
• clear enough for the target sample to understand
• neutral rather than persuasive
• specific about time, behaviour or attitude
• written so that the response can be analysed

For example, "How often have you felt unable to relax during the last seven days?" is stronger than "Are you stressed?" The improved version gives a clear time frame and focuses on one experience. This helps validity because the response is more likely to measure the intended concept.

The target sample matters. An item about pressure at work may not fit sixth-form students; an item about revision may not fit adults who have no exams. Ask what the answer would actually establish. Hours of sleep can be relevant to stress but do not by themselves measure stress: illness, work or caring responsibilities could also reduce sleep.

Open and closed questions

The central choice in questionnaire construction is whether each item should be open or closed. This choice affects the type of data produced, how easy the data are to analyse, and how much freedom participants have to express their own ideas.

Open questions are useful when the researcher wants participants' own explanations rather than a pre-set category.

An open question might ask, "What is the main reason you feel anxious before exams?" This can produce rich qualitative data because participants can describe experiences the researcher did not predict. The limitation is that answers are harder to compare. The researcher may need to code responses into categories before looking for patterns.

A closed question might ask, "How anxious do you feel before exams? 1 = not anxious at all, 5 = extremely anxious." This produces quantitative data that can be summarised quickly, compared between groups and displayed in graphs or tables. The limitation is that the fixed options may not capture the participant's real view. A participant who feels "panicky only in the final hour before the exam" may feel squeezed into a number that loses important detail.

Closed questions can take several forms:

A questionnaire can mix open and closed items when both kinds of evidence help answer the aim. An exclusively closed questionnaire may be appropriate for a brief frequency survey; an open follow-up may be useful when the researcher needs to know why a response was chosen. More question types do not automatically make a stronger instrument.

Closed responses are not always numerical: selecting a travel method gives categorical data. Categories can be counted, but a code such as bus = 1 and car = 2 does not make car twice as much as bus. Ordered agreement ratings indicate rank without guaranteeing equal gaps between responses. Open responses usually provide qualitative text, which can later be coded and counted.

Writing unbiased questions

Bad wording can make a questionnaire look scientific while quietly damaging the data. A researcher should try to remove wording that confuses respondents or pushes them towards a particular answer.

Common problems include:

Leading questions are a threat to validity because the wording, not the participant's real view, may shape the answer.

Closed questions also need carefully designed response options. The options should usually be mutually exclusive, meaning that each answer fits into only one category, and exhaustive, meaning that all reasonable answers are covered. For example, age bands of "16-18", "18-20" and "20-22" overlap at 18 and 20, so they are not mutually exclusive. A better set would be "16-17", "18-19", "20-21" and "22 or over".

If a response option is missing, participants may give inaccurate answers. "How do you travel to college? Bus / Car / Walk" excludes train, bicycle and "other". If many respondents choose the nearest wrong option, the data may look tidy but be invalid.

Worked response options

Consider "On how many days in the last seven days did you exercise for at least 20 minutes? Tick one: 0 / 1–2 / 2–4 / 5–7." Someone reporting two days fits two boxes. Change the bands to 0 / 1–2 / 3–4 / 5–7. They now cover every possible whole-number answer exactly once. A direct box for a number from 0 to 7 could be simpler still; writing a numerical answer is not the same as an open invitation to explain an experience.

For agreement items, balance the choices around the statement: strongly disagree, disagree, neither agree nor disagree, agree, strongly agree. "Do not know" is different from a neutral opinion. A participant may understand the issue and hold no preference, or may lack enough information to judge it.

Some topics also need a neutral or opt-out option. "Do not know", "prefer not to say" or "not applicable" can be useful when participants may genuinely lack the information or may feel uncomfortable answering. However, too many opt-out options can reduce the amount of usable data, so the researcher should include them when they improve honesty and participant comfort.

Order, instructions and piloting

Questionnaire construction is not finished when the researcher has written the items. The whole questionnaire needs a sensible route through the topic.