Choosing a published scale is a solved problem. Building the instrument around it is not. Almost every Canadian thesis that collects its own data needs at least one thing that does not exist off the shelf: a demographic block, a set of context items, an interview guide. This page builds those, with the item-writing rules that keep them defensible, a worked survey block and a full interview guide, and the consent elements your research ethics board will check — quoted from the policy that actually governs them rather than from a summary.
One decision comes first. If a validated instrument exists for your construct, use it whole. Building your own means you own the burden of establishing its properties, which is a thesis-sized task in itself. The site’s burnout scale comparison shows what choosing properly looks like. Build only what is genuinely unavailable.
The instrument is a chain, and the weakest link is the item
Your conceptual framework names a construct. Your operationalisation table names an indicator and a recording rule. The instrument is where that rule becomes words a person reads. Every ambiguity you leave in the wording becomes variance you cannot explain later, and no amount of analysis recovers it.
Eight item-writing rules
| Rule | Bad | Better |
|---|---|---|
| One idea per item | “I am satisfied with my supervision and my resources” | Split into two items |
| No negation in the stem | “I do not feel unprepared” | “I feel prepared” |
| Concrete time frame | “How often do you use formative assessment?” | “In the past two teaching weeks, how many times…” |
| Response options exhaustive and exclusive | Age bands 18–25, 25–35 | 18–24, 25–34 |
| Match the scale to the question | Agreement scale on a frequency question | Frequency anchors on a frequency question |
| No assumed behaviour | “How often do you supervise students?” | Filter item first: “Do you supervise students?” |
| Plain language, no internal jargon | “Your unit’s FTE complement” | “The number of full-time staff on your unit” |
| Neutral stem | “How helpful was the new policy?” | “How would you describe the effect of the new policy?” |
The third rule does the most work. “How often” with no window means every respondent answers a different question, and the resulting variable is a mixture of behaviour and memory horizon.
A worked survey block
This is the context block for the running example used across this site’s method guides — a study of assessment practice across departments. Adapt the structure, not the content.
Section A — Your teaching context
A1. Which department are you currently assigned to? (single choice, list supplied by the board, plus “Other, please specify”)
A2. How many years have you taught full time, including this year? (numeric, one decimal place; count a half year as 0.5)
A3. Do you currently teach at least one course with a provincially set examination? (Yes / No / Not sure)
Filter: if No or Not sure, skip to Section C.Section B — Assessment activity in the past two teaching weeks
Instruction shown to respondents: “The next five questions ask about the two most recent weeks in which you taught. If you were away for part of that period, answer for the teaching days you were present.”
B1. In the past two teaching weeks, how many times did you give written feedback on work that was not graded? (Never / Once / 2–3 times / 4–5 times / More than 5 times)
B2–B5. Four further activities, each with the same time frame and the same five frequency anchors, each reported as its own variable and not summed.
Section C — Closing
C1. Is there anything about assessment in your department that these questions did not ask about? (open text, optional, 500 characters)
Four things this block does deliberately. The filter at A3 prevents a whole department from answering a question that does not apply to them. The instruction before section B fixes the reference window in the respondent’s mind rather than leaving it in the item. The five activities share anchors so they are comparable, and are explicitly not summed because they are not interchangeable. And C1 exists because an open closing item is the cheapest instrument-validity check you will ever run: if three respondents name the same missing thing, your instrument has a hole.
A full interview guide
An interview guide is not a list of questions. It is an opening, a small number of main questions each with planned probes, and a close. Seven main questions is a comfortable number for a 45-to-60-minute semi-structured interview.
Opening (2–3 minutes). Thank the participant. Restate the purpose in one sentence. Confirm consent to record. State that there are no right answers and that they can decline any question or stop at any time.
Q1 (grand tour). “Walk me through how assessment works in your department over a typical unit of teaching.” Probes: who decides; where it is written down; what happens when it is not followed.
Q2 (specific instance). “Think of the most recent time you gave feedback that was not going to be graded. What happened?” Probes: what prompted it; how long it took; what the student did with it.
Q3 (contrast). “How is that different in the courses with a provincial examination?” Probes: what changes; when in the year; who notices.
Q4 (the tension). “Some teachers say those two things pull against each other. Does that match your experience?” Probes: in what way; what you do about it; whether it has changed.
Q5 (the department). “How similar is your practice to your colleagues’ in the same department?” Probes: how you know; where you differ; whether it is ever discussed.
Q6 (change). “Has any of this changed in the past two or three years?” Probes: what prompted it; whether it stuck.
Q7 (counterfactual). “If the provincial examination weighting changed tomorrow, what would you do differently?” Probes: immediately; over a year.
Close. “Is there anything I should have asked and did not?” Then: what happens to the recording, when results will be available, and the contact route for questions or withdrawal.
Why this order. Q1 is broad and easy, which builds fluency before anything sensitive. Q2 asks for a specific episode rather than a generalisation, and specific episodes are where the useful data lives. Q4 introduces the study’s tension only after the participant has described their own practice in their own terms, so it does not lead them. Q7 is a counterfactual, which is the safest way to ask about motivation without asking someone to psychoanalyse themselves. And the closing question is not a formality — it is the item that catches what your framework missed.
Two mechanical points. Pilot the guide with at least two people outside your sample and report that you did. And decide before you start how the recording will be handled, because the route you choose has to match what your ethics application says — the options and their implications are set out in the transcription and data-residency guide. What happens to the transcripts afterwards is thematic analysis.
What your REB actually requires, from the policy itself
Research with humans at a Canadian institution is governed by the Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans, TCPS 2 (2022). Its Article 3.2 requires researchers to “provide to prospective participants, or authorized third parties, full disclosure of all information necessary for making an informed decision to participate in a research project”, and the policy then lists the information generally required. Reproduced from that list, the elements are:
- that the individual is being invited to participate in a research project;
- the research purpose in plain language, the identity of the researcher, the identity of the funder or sponsor, the expected duration and nature of participation, a description of procedures, and the participant’s responsibilities;
- a plain-language description of all reasonably foreseeable risks and potential benefits;
- an assurance that participants are under no obligation to participate and are free to withdraw at any time without prejudice to pre-existing entitlements, will receive information relevant to continuing or withdrawing, and will be told of their right to request withdrawal of data, including limits on the feasibility of that withdrawal;
- the possibility of commercialisation, and any real, potential or perceived conflicts of interest;
- how results will be disseminated and whether participants will be identified directly or indirectly;
- the identity and contact information of a qualified representative who can explain scholarly aspects of the research;
- the identity and contact information of someone outside the research team whom participants may contact about ethical issues;
- what information will be collected and for what purposes, who will have access to identifying information, how confidentiality will be protected, the anticipated uses of data, and who may have a duty to disclose information and to whom;
- any payments, incentives, reimbursement of expenses and compensation for injury;
- a statement that consenting does not waive any right to legal recourse in the event of research-related harm.
Three of those are the ones students most often omit. The second contact — someone outside the research team — is mandatory and is usually your institution’s research ethics office, not your supervisor. The limits on withdrawal must be stated up front: TCPS 2 says the consent process “should set out any circumstances that do not allow withdrawal of data”, and where withdrawal is impossible the policy requires a rationale to the REB and protection of participants’ identity throughout. And the duty to disclose line matters for anyone collecting data in schools, health settings or on sensitive topics, where reporting obligations can override confidentiality.
TCPS 2 also states that the policy “neither recommends nor discourages the use of incentives”, placing the onus on the researcher to justify the model and level to the REB, and that incentives “should not be so large or attractive as to encourage reckless disregard of risks”. If you are offering a gift card, that sentence is the standard you will be held to. The application process itself is covered in the REB approval guide.
Before you collect anything: a six-item check
- Every item traces to a row in your operationalisation table. Items that trace to nothing get cut — they cost response rate and answer no question.
- The instrument has been piloted with at least two people outside the sample, and the pilot changed something.
- Skip logic has been tested by walking through every path, including the paths you expect nobody to take.
- Estimated completion time has been measured, not guessed, and appears in the invitation.
- The consent text contains every applicable element above, and the version in your ethics application is the version you will actually use.
- The wording of your recruitment invitation matches your sampling plan — same population, same criteria, same route.
Five mistakes
- A modified published scale. Removing items voids the published reliability and validity evidence for your version. Adopt whole, or justify and re-establish.
- Demographics first. Long demographic blocks at the start depress completion. Put the substantive items first and the demographics at the end, unless you need them for filtering.
- An interview guide of seventeen questions. It is a questionnaire read aloud. Seven main questions with planned probes produces more usable material.
- Consent text written after the instrument. The consent has to describe what the instrument actually does, so write them together and submit them together.
- No open closing item. On both instruments, it is your cheapest check on whether the instrument covers the phenomenon.
Draft the instrument section with the instrument
Tesify drafts the instruments section of your methods chapter alongside the items themselves, so the justification exists at the moment you write the question rather than three months later. Start your thesis with Tesify — 100% written by you.
Frequently asked questions
Should I build my own questionnaire or use a published one?
Use a published one whenever a validated instrument exists for your construct. Build only the parts that genuinely do not exist, usually context and demographic items, and say clearly which parts are yours.
Can I shorten a validated scale?
Only with justification, and the published psychometric evidence no longer applies to your shortened version. You then have to report the properties of the version you used.
How many questions should an interview guide have?
About seven main questions with planned probes for a 45-to-60-minute semi-structured interview. More than ten usually means you have written a spoken questionnaire.
Do I need to pilot my instrument?
Yes, and you should report that you did, with how many people and what changed. A pilot that changes nothing usually means the pilot was too gentle.
What does TCPS 2 require in a consent form?
Article 3.2 requires full disclosure of the information necessary for an informed decision, and the policy lists the elements — purpose, researcher and funder identity, duration and procedures, risks and benefits, the freedom to withdraw and its limits, dissemination, two contacts, data handling and access, payments, and a statement that consenting waives no legal rights.
Who is the second contact on a consent form?
Someone outside the research team whom participants can contact about ethical issues — in practice, your institution’s research ethics office. Your supervisor does not satisfy this requirement.
Can participants always withdraw their data?
No, and you have to say so in advance where they cannot. TCPS 2 requires the consent process to set out any circumstances that do not allow withdrawal, and requires a rationale to the REB for collection methods that make withdrawal impossible.
Can I offer an incentive?
TCPS 2 neither recommends nor discourages incentives, and puts the onus on you to justify the model and level to the REB. The constraint is that an incentive must not be so large or attractive as to encourage reckless disregard of risks.
Where do the instruments go in the thesis?
The full instruments go in appendices, with the methods chapter describing their structure, origin and properties and referring to the appendix.
Should demographic questions come first or last?
Last, unless they are needed to route respondents. Substantive items placed first get answered by more people.