1.5 - Conducting a self-report practical
Learn to plan, conduct, analyse and reflect on a small questionnaire investigation. The worked records are fictional; completing the written checks does not mean you have conducted the required practical yourself.
A question and a reachable sample
A useful self-report investigation asks people about something they can reasonably report and defines whose answers are being studied.
Our original example asks: Which format do adult library users prefer when reading a short factual article, and why? Preference is the format selected, not reading skill. A questionnaire can access a reason that an observer cannot see, but the reported preference may differ from behaviour when the reader actually has both formats available. This is a descriptive investigation, not a manipulation proving that a screen changes concentration.
Define the target population as adult users of this particular library. With institutional permission, invite available adult users during two agreed afternoon sessions until 20 consenting responses are obtained. This is an opportunity sample. Two sessions are a practical way to avoid recruiting only one immediate friendship group, but afternoon visitors may still differ from evening visitors. A sample of 20 is an example for a manageable class project, not an OCR rule or evidence of representativeness. Record invitations, refusals and usable returns as counts; never replace a refusal with a fabricated answer.
Sampling bias
A systematic tendency for a sample to overrepresent or underrepresent some members or characteristics of the target population.
If a membership list and permission existed, randomly selecting names would give each listed member an equal selection chance. It would not eliminate refusal bias, exclude the need for consent, or make this one library representative of all adults. A voluntary online advertisement instead recruits people who choose to respond, potentially attracting particularly strong preferences.
Questions that measure preference
A questionnaire needs questions and response options that measure the stated aim without pushing the respondent toward a preferred answer.
Use the same short description for everyone: “Imagine reading a short factual article. Both a printed copy and a screen copy are available, with the same words and no price difference.” The first item asks, “Which would you prefer?” with Print, Screen, No preference, and Prefer not to answer. Including no preference avoids forcing a choice. The second asks, “How easy do you usually find reading a factual article on a screen?” on a five-point scale from Very difficult to Very easy, with Not applicable and Prefer not to answer separate from the scale. The third asks, “What is the main reason for your format preference?” with space for a brief answer. Avoid “Why is print better for concentration?”, which assumes both a preference and a benefit.
The chosen format is nominal quantitative data once counted into categories. The ordered ease ratings are ordinal quantitative data: the rank order is meaningful, but the gap between two adjacent labels need not be equal. The written reasons are qualitative data, retaining meaning and detail. An open question can also elicit a number, so response format alone does not determine data type. Coding a written reason into a category creates a numerical summary while the original text remains qualitative evidence.
Pilot the information sheet and questionnaire with two eligible volunteers whose answers will not enter the main dataset. Ask what each question meant to them, whether an option is missing and whether it felt intrusive. If “screen” means a phone to one and a large monitor to another, specify the relevant device or acknowledge that your question intentionally covers both. Keep a dated final version, record changes and restart the main collection with that version; mixing versions changes what the numbers mean.
Safe collection and digital records
Actual collection begins only after the supervisor has approved the protocol and the institution's risk assessment.
List hazards, who might be affected, controls, remaining risk and an action if a control fails. Here, approaching readers could interrupt them or make them feel obliged to cooperate. Use an approved unobstructed invitation area, approach politely once and avoid interrupting private conversations. The information sheet states the purpose, approximate three-minute duration, voluntary participation, freedom to omit items and no effect on access to the library. Do not recruit someone who cannot understand the consent information. Follow the institution's consent requirements; this protocol deliberately uses adults rather than assuming one rule for every age group.
Privacy risk comes from collecting or displaying identifiable answers. Do not request names, membership numbers, emails or sensitive personal information. Give each respondent a random withdrawal code; explain a realistic deadline for removal before anonymous aggregation. Keep consent evidence separately from coded responses. Explain that removal of an individual's data is no longer possible after irreversible anonymisation rather than promising unlimited withdrawal. Offer a debrief explaining the descriptive aim and contact route for concerns. Stop if a participant becomes uncomfortable; permission to begin is not permission to ignore later withdrawal.
A digital form can present one consistent version and export responses to a spreadsheet. Disable collection of email addresses and unnecessary identifying metadata where possible; check the settings rather than assuming a form is anonymous. Use an institution-approved account with access restricted to the project team. Give columns clear names, separate missing answers from “No preference”, and preserve a read-only raw export before coding. Check several entries against the originals and reconcile the number of rows with usable forms. A paper alternative avoids excluding people uncomfortable with a digital form; enter it using the same codes and record its mode. Do not upload identifiable raw responses to a public dashboard or external AI tool.
During real conduct, use the agreed invitation script, obtain consent before the questionnaire, allow private completion, avoid commenting on answers and note deviations without identifying the reader. Document what you personally did and what others did if data were pooled.
Counts, percentages and written reasons
Analyse what respondents actually supplied and make the denominator of every percentage clear.
In a fictional demonstration, 20 usable respondents all chose one format: 12 Print and 8 Screen. The print-to-screen ratio is . Print accounts for of responses and Screen for .
The print percentage is ; the screen percentage is . If two people had omitted this item, use 18 answered items as the denominator for the preference distribution and separately report two missing answers out of 20 respondents. Missing is not “No preference”. Percentages make relative frequencies clear, but do not repair a biased sample.
For reasons, first read all answers, then build a coding frame with operational rules. For example, “Concentration” requires an explicit reference to maintaining attention; “Convenience” includes carrying or accessing a format; “Other” retains reasons outside these definitions. If the question asks for a main reason, code one main theme per response using a stated priority rule when two are given. If you allow multiple themes, totals can exceed the number of respondents; report that rule instead of treating every theme as a separate person. A second coder independently categorises a subset, disagreements are examined, and ambiguous rules are refined before recoding consistently.
Report counts alongside an anonymised short example where permission permits. A frequency table of themes makes common reasons visible, but can hide differences between two people assigned to the same theme. A reason about concentration is a respondent's explanation, not evidence that print objectively improves attention. Keep the original texts securely so another researcher can inspect the link between responses and codes.
Reporting and learning from conduct
A practical record must show both the evidence obtained and what the experience taught you about the method.
Use a concise report: title and aim; a brief rationale; method covering sample, materials, ethics and replicable procedure; results summarising usable responses and coded themes; discussion interpreting the findings and evaluating limitations; references; appendices containing the final questionnaire, instructions and anonymised raw data. Describe actual conduct in past tense only when it occurred. Put the questionnaire and coding rules in the appendices so a later researcher can reconstruct the measures. Keep numerical results separate from explanations: “60% selected Print” is a result; “they may find it less distracting” is an interpretation needing evidence.
Reflection identifies an observed problem, its consequence, a concrete modification and how you would check whether the modification worked. For example, during a real pilot some volunteers might interpret “factual article” as a textbook chapter. That could introduce inconsistent imagined tasks; supplying the same brief article description would reduce variation, and a second pilot could check interpretation. Do not write this as an event you experienced if it is only a fictional possibility. Note useful features too: private completion may reduce interviewer pressure, while it cannot guarantee truthfulness or remove recall error.
Own-practical transfer means using real experience to justify a new design. If your actual project had a confusing response scale, explain the confusion, the change and why it informs the new scale. In a written simulation, explicitly say “the supplied fictional log indicates…” and reason from it. A tidy submitted report or a correct test answer is evidence of written understanding; authentic practical completion requires real supervised conduct and records.