A closed-ended question is a survey or interview item that supplies the respondent with a fixed set of response options to choose from — yes/no, a list of categories, a numbered scale — rather than inviting a free-text answer in the respondent’s own words. The defining feature is not the subject matter but the response format: if the set of possible answers is fixed in advance and the respondent selects from it, the item is closed-ended, regardless of how the question itself is worded.
Closed-ended items are the backbone of quantitative questionnaire design because a fixed response set is what makes an answer codeable as a number and comparable across every respondent who answered it. This guide covers what distinguishes closed-ended from open-ended questions, the six main closed-ended item types with worked examples, when each is the right choice, their real advantages and limitations, how to write ones that produce analyzable data, and how the resulting responses get coded and analyzed.
Closed-Ended vs. Open-Ended Questions
The closed/open-ended distinction is about the response format, not the topic. The same subject can be asked either way:
- Closed-ended: “How satisfied are you with the training program?” — Very dissatisfied / Dissatisfied / Neutral / Satisfied / Very satisfied
- Open-ended: “What did you think of the training program?” — a blank text box
Both ask about the same construct. What differs is what the researcher gets back: a number that slots directly into a frequency table or a mean, versus a string of free text that has to be read, coded, and interpreted before it can be analyzed at all.
| Dimension | Closed-ended | Open-ended |
|---|---|---|
| Response format | Selects from a fixed set of options | Writes a free-text answer |
| Data type produced | Categorical or ordinal, directly codeable | Qualitative text requiring coding/thematic analysis |
| Analysis speed | Fast — frequencies, cross-tabs, means, statistical tests | Slow — requires a coding frame or qualitative analysis |
| Comparability across respondents | High — everyone answers on the same scale | Low — wording, depth, and focus vary by respondent |
| Nuance and unanticipated answers | Limited to the options offered | Captures whatever the respondent considers relevant |
| Respondent burden | Low — a click or tick | Higher — requires composing an answer |
| Risk of forced-choice distortion | Real — a respondent may pick the closest-fit option even if none fits | Low — the respondent frames the answer themselves |
In practice, most instruments use both: closed-ended items for the variables that need to be measured and compared across the sample, with a smaller number of open-ended items — often as a follow-up “please explain” or “other (specify)” field — to capture what a fixed list would miss. This combination sits within the broader questionnaire design decisions covered separately, including question wording, ordering, and pretesting.
The Six Main Types of Closed-Ended Questions
1. Dichotomous questions
A dichotomous question offers exactly two mutually exclusive options — typically yes/no, true/false, or agree/disagree.
Example: “Have you published a peer-reviewed article in the past 12 months?” — Yes / No
Dichotomous items are fast to answer and trivially easy to code (0/1), but they force a binary answer onto questions that may have a genuinely intermediate or conditional true answer, which is their main limitation.
2. Multiple-choice (categorical) questions
A multiple-choice question offers three or more mutually exclusive, unordered categories, and the respondent selects one.
Example: “Which best describes your primary research method?” — Quantitative / Qualitative / Mixed methods / Not applicable
The categories must be both exhaustive (every respondent has a valid option, often ensured with an “other” or “not applicable” catch-all) and mutually exclusive (no respondent’s true answer fits two options at once) — violating either produces missing data or forces an inaccurate answer.
3. Likert-scale questions
A Likert item asks a respondent to rate agreement, frequency, or intensity along an ordered scale, typically five or seven points, anchored with labels at each end (and often at the midpoint).
Example: “The onboarding process for new lab members is well organized.” — Strongly disagree / Disagree / Neutral / Agree / Strongly agree
A single Likert item is one closed-ended question; a Likert scale is a set of such items summed or averaged into a single score for an underlying construct (e.g., job satisfaction). Construction details — number of points, whether to force a choice by omitting the neutral midpoint, and how to check internal consistency — are covered in the questionnaire design guide and in Cronbach’s alpha for scale reliability.
4. Rating-scale questions
A rating scale asks the respondent to assign a numeric or graded value to a single attribute, without the agreement framing a Likert item uses.
Example: “On a scale of 1 to 10, how would you rate the clarity of the informed consent form?”
Rating scales are common for satisfaction and quality measures (Net Promoter Score is a well-known example) and are analyzed as either ordinal or, when the scale is long enough and treated as approximately interval, with means and standard deviations — a decision that should be stated and justified, not assumed.
5. Ranking questions
A ranking question presents a list of items and asks the respondent to order them by preference, importance, or priority, rather than rating each independently.
Example: “Rank the following funding priorities from 1 (highest priority) to 5 (lowest priority): equipment, personnel, travel, publication costs, training.”
Ranking data is ordinal and relative — it shows which option a respondent prefers over another, but not by how much, and one respondent’s “1” is not directly comparable in magnitude to another’s. Ranking questions also become unreliable past roughly seven to eight items, since respondents struggle to meaningfully order long lists.
6. Checklist (multiple-response) questions
A checklist question presents a list of options and allows the respondent to select as many as apply, rather than exactly one.
Example: “Which of the following data repositories has your lab used? (Select all that apply)” — Dryad / Zenodo / figshare / an institutional repository / None of the above
Checklist responses are coded as a separate binary (selected/not selected) variable for each option, not as a single categorical variable — a common analysis error is treating a checklist item as if it were mutually exclusive multiple-choice.
When to Use Closed-Ended Questions
Closed-ended questions are the right choice when:
- The researcher already knows the plausible range of answers (from prior literature, a pilot study, or established categories) well enough to build a meaningful response set.
- The goal is to measure something numerically or compare it across a sample — frequencies, group differences, correlations, trends over time.
- Sample size is large enough that manually coding free text for every respondent is impractical.
- Respondent burden needs to stay low to protect response rate, e.g., in a long instrument or a population with limited time.
Open-ended questions remain the better choice when the researcher is exploring a topic without a settled set of categories, when the research question is about meaning or process rather than magnitude, or when a fixed list would risk missing the actual range of respondent experience — territory covered by qualitative research methods.
Advantages of Closed-Ended Questions
- Quantifiable. Responses convert directly into numbers, enabling frequencies, cross-tabulations, means, and inferential statistics without an intermediate coding step.
- Comparable. Every respondent answers from the same option set, so answers can be aggregated and compared across individuals, groups, or time points on a like-for-like basis.
- Fast to analyze. Analysis is largely automatable — software can tabulate and test closed-ended responses at scale, where open-ended text requires a coding frame and, typically, human judgment.
- Lower respondent burden. Selecting an option takes less time and cognitive effort than composing a written answer, which can improve completion rates, particularly in longer instruments.
- Easier to replicate. A fixed instrument with fixed response categories is straightforward to re-administer to a new sample or at a later time point for comparison.
Limitations of Closed-Ended Questions
- No room for nuance. A respondent whose true answer falls between or outside the offered options has no accurate way to record it.
- Forced-choice distortion. Respondents may select the closest-fitting option rather than leave an item blank, quietly introducing measurement error that looks like clean data.
- Response-option bias. The categories or scale points the researcher chooses shape the answers received — an incomplete or unevenly spaced option set can distort the resulting distribution.
- Can’t capture the unanticipated. A fixed list only reflects what the researcher already thought to include; genuinely new categories, explanations, or concerns go unrecorded.
- Susceptible to response-style bias. Some respondents systematically favor extreme options, the midpoint, or agreement regardless of item content (acquiescence bias), which affects closed-ended scales more than open-ended text.
Writing Closed-Ended Questions That Produce Analyzable Data
The wording failure patterns that corrupt any survey item — double-barrelled, leading, and vague items — are covered in full in the questionnaire design guide; the same fixes apply directly to closed-ended items, with two additions specific to a fixed response set:
- Make the option set exhaustive. Every respondent should have a truthful option available. A question about employment status without a “student” or “retired” category will force inaccurate answers from anyone in those groups; adding “Other (please specify)” is a common safety valve.
- Make the option set mutually exclusive. For single-select items, no respondent’s true answer should fit two categories at once. Overlapping numeric ranges (“0–5 years” and “5–10 years”, where 5 falls in both) are a frequent, easily avoidable error.
- Avoid double-barrelled items even in a closed-ended format. “Rate your satisfaction with the equipment and the training you received” cannot be answered accurately with a single scale point if satisfaction with the two differs — split it into two items.
- Keep scale points balanced and clearly labeled. An unbalanced scale (e.g., three positive options against one negative one) biases responses toward the more heavily represented pole; label every point rather than only the endpoints where precision matters.
- Decide deliberately whether to include a neutral midpoint. Omitting it forces a lean one way or the other; including it lets genuinely neutral respondents answer honestly but can also become a default for anyone who wants to avoid deciding — the right choice depends on the construct being measured.
Coding and Analyzing Closed-Ended Response Data
How a closed-ended response is coded depends on its data level, which in turn determines what analysis is valid:
- Nominal (dichotomous, unordered multiple-choice, checklist items): coded as categories or, for yes/no items, as 0/1. Valid summaries are frequencies and proportions; valid tests include chi-square tests of association. Checklist items are coded as one binary variable per option, not one variable for the whole question.
- Ordinal (Likert items, ranking, many rating scales): coded as ordered integers (e.g., 1–5). The categories have a meaningful order but the intervals between them are not guaranteed to be equal, which is why medians and non-parametric tests are the conservative default, though summed multi-item Likert scales are commonly treated as approximately interval and analyzed with means once internal consistency (Cronbach’s alpha) has been checked.
- Interval/ratio (some numeric rating scales, e.g., 0–100): coded as the number given. Means, standard deviations, and parametric tests (t-tests, ANOVA, regression) are appropriate.
Before analysis, closed-ended data should be checked for the same data-quality issues as any dataset: missing values (distinguish “skipped” from “not applicable”), straight-lining (a respondent selecting the same scale point for every item, a common low-effort response pattern), and, for scales, the reliability check referenced above. Sample-size and power considerations for detecting a real effect in the resulting analysis are covered in power analysis and sample-size calculation.
Frequently Asked Questions
What is an example of a closed-ended question in research?
“Do you currently hold external research funding? Yes / No” is a dichotomous closed-ended question. “How would you rate your satisfaction with the core facility on a scale of 1 to 5?” is a rating-scale closed-ended question. Both restrict the respondent to a fixed set of predetermined answers.
What is the main difference between closed-ended and open-ended questions?
Closed-ended questions restrict the respondent to a predetermined set of response options, producing data that is immediately codeable and comparable across respondents. Open-ended questions let the respondent answer in their own words, producing richer but slower-to-analyze qualitative text.
Can a survey combine closed-ended and open-ended questions?
Yes, and most instruments do. A common pattern uses closed-ended items for the core variables being measured, paired with a small number of open-ended items (often an “other, please specify” field or a final “any additional comments” prompt) to capture what the fixed options might miss.
How many response options should a closed-ended question have?
There is no universal number; it depends on the construct. Likert scales conventionally use five or seven points (enough to detect meaningful variation without overwhelming the respondent); multiple-choice categories should be as many as needed to be exhaustive and mutually exclusive, with an “other” catch-all when the full range can’t be anticipated.
Are closed-ended questions always quantitative?
The response format is what produces quantitative, codeable data, but the underlying research can still sit within a mixed-methods design that also uses open-ended or qualitative components elsewhere in the same study.
What is a Likert scale and how is it different from a rating scale?
A Likert item measures agreement, frequency, or intensity along a labeled, symmetric ordered scale (e.g., strongly disagree to strongly agree), usually as one item within a multi-item scale summed into a single score. A rating scale asks for a numeric or graded judgment of a single attribute (e.g., 1 to 10) without the agreement framing, and is more often used as a standalone item.
