Written and maintained by CASRAI Editorial Board
Last updated
Researchers, students and program evaluators searching for good survey questions usually need two things: rules for writing an item that measures what it should, and examples they can adapt. A good survey question is one that every respondent can read the same way, can answer from their own experience, and cannot be nudged toward a particular answer by its wording. This guide gives the rules, five annotated examples, and a table of bad-to-good rewrites, each tied to the published source that states the rule.
What makes a survey question good
Pew Research Center and Nielsen Norman Group (NN/g) independently describe the same core properties: the question asks for one thing, uses neutral and plain wording, and offers answer options that cover every reasonable response without overlapping. Pew adds that small wording changes can move results. In a 2005 survey it cites, 51% favored “making it legal for doctors to give terminally ill patients the means to end their lives” while 44% favored “making it legal for doctors to assist terminally ill patients in committing suicide” (Pew Research Center, “Writing Survey Questions”, read 2026-11-07).
The topic of a question is a separate decision from its quality. Whether a question is “fun”, “interesting” or aimed at students changes what you ask, not the rules for how you ask it.
Five good survey question examples
These are illustrative items written for this guide, not items from a published instrument. Each applies a rule stated in the sources cited below.
- “In the past 7 days, approximately how many times did you use the library’s online catalogue?” It asks about recent behavior rather than a prediction. NN/g advises asking about specific, recent memories instead of “How likely are you to…?” questions.
- “How easy was it to find the information you needed?” followed by a separate item on clarity. It asks about one thing. Both sources warn against double-barreled items such as “How easy and intuitive was this website to use?”
- “Overall, how would you rate the course? Excellent / Good / Neutral / Fair / Poor.” The scale is balanced, with two positive and two negative options around a midpoint, as NN/g recommends.
- “What is your age? Under 20 / 20-29 / 30-39 / 40-49 / 50 or older / Prefer not to answer.” The ranges do not overlap and leave no gap, and the opt-out is its own option, both points made by NN/g. See mutually exclusive survey questions for the underlying rule.
- “Is there anything else you would like to share about your experience?” NN/g recommends mostly closed-ended questions with one broad open-ended question at the end.
Bad-to-good rewrites for research instruments
The wording in the “Bad item” column is quoted from the cited source; the “Better item” column describes the fix the source gives.
| Flaw | Bad item | Better item | Source |
|---|---|---|---|
| Double-barreled | “How much confidence do you have in President Obama to handle domestic and foreign policy?” | Ask separately about domestic policy and foreign policy. | Pew |
| Double-barreled | “How easy and intuitive was this website to use?” | Split into one question on ease and one on intuitiveness. | NN/g |
| Double negative | “Do you favor or oppose not allowing gays and lesbians to legally marry?” | Use a direct, positively framed statement of the policy. | Pew |
| Leading | “We are committed to achieving a 5-star satisfaction rating. How would you rate your satisfaction?” | Drop the first sentence and ask only the rating question. | NN/g |
| Prediction instead of behavior | “How likely are you to use this product?” | “Approximately how many times did you use this product in the past 7 days?” | NN/g |
| Unbalanced scale | Excellent / Very good / Good / Poor / Very poor (three positive, two negative) | Excellent / Good / Neutral / Fair / Poor | NN/g |
| Overlapping and incomplete ranges | 0-20, 20-30, 30-40, 40-50 | Non-overlapping ranges plus an option for values above 50. | NN/g |
| Loaded wording | “Do you suffer from hypertension?” | “Do you have hypertension?” | NN/g |
| Jargon | “ANWR” | “Arctic National Wildlife Refuge” | Pew |
How to write survey questions for research
- Start from the research questions. NN/g advises asking only what your research questions require. Every item should map to one.
- Write one idea per question. Split any item containing “and” that joins two judgments.
- Use neutral, plain words. Pew shows that “assistance to the poor” draws more support than “welfare”, so test whether a term carries an opinion of its own.
- Cover all reasonable answers without overlap. Pew recommends keeping closed-ended options to about four or five, putting unordered options in random order and keeping ordinal scales in order.
- Prefer forced-choice to agree/disagree. Pew notes acquiescence bias, which is stronger among less-educated respondents, and prefers forced-choice formats to select-all-that-apply.
- Handle sensitive topics carefully. Pew’s examples of social desirability are voting (acknowledge circumstances that can prevent a vote) and church attendance (offer several infrequent-attendance options).
- Offer opt-outs separately. NN/g advises distinct options such as “Not applicable”, “I don’t know” and “Prefer not to answer” rather than one merged catch-all, and making most questions optional unless the data justifies the drop-off risk.
- Keep wording identical when measuring change. Pew notes that changing wording or survey mode alters results; see survey mode effects and mixed-mode designs.
Good survey questions for students
For a class project or student survey, use the same rules: pick a narrow research question, ask about specific recent behavior, and keep scales balanced. What changes is the topic and wording level. Avoid jargon (Pew’s ANWR example applies to any audience) and avoid prediction questions. For a worked set of closed-ended items from a real instrument, see quantitative survey question examples, and for yes/no items see dichotomous survey questions.
Checklist before fielding
- Each item maps to a research question.
- Each item asks about one thing.
- No leading preamble, no loaded terms, no double negatives, no jargon.
- Questions ask about recent behavior rather than predictions where possible.
- Options are exhaustive, non-overlapping and, for ratings, balanced.
- Opt-outs are separate options.
- One open-ended item at the end.
- Wording and mode are fixed if you will compare waves.
For the wider picture of how wording errors distort results, see leading questions in surveys and survey bias.
FAQ
What are 5 good survey question examples?
See the five annotated items above: a recent-behavior frequency question, a single-topic ease question, a balanced rating scale, a non-overlapping age range with a separate opt-out, and a closing open-ended question. They are illustrative items written for this guide.
What are some interesting survey questions?
Interest comes from the topic, and the quality rules do not change: one idea per question, neutral wording, balanced options. Choose topics that answer your research question, then apply the checklist.
How do I write survey questions for research?
Map every item to a research question, ask one thing at a time in neutral plain words, and give complete, non-overlapping options. NN/g and Pew set out these steps; see the numbered list above.
What are good survey questions and answers?
A good item pairs a clear question with answer options that are exhaustive and mutually exclusive, with opt-outs offered separately. The bad-to-good table above shows paired fixes.
What are 5 good research questions?
Research questions (what you want to learn) differ from survey questions (what you ask respondents). This guide covers survey items; each survey item should trace back to a research question.
Sources
- Pew Research Center, “Writing Survey Questions”: pewresearch.org/writing-survey-questions (read 2026-11-07).
- Nielsen Norman Group, Maddie Brown, “Writing Good Survey Questions: 10 Best Practices”, 2023-08-20: nngroup.com/articles/survey-best-practices (read 2026-11-07).
