Question types in a questionnaire: all types explained

Two questions about the same seminar can produce results you cannot add together. Ask “How satisfied were you with the seminar?” with five labelled options beside it, then ask “What would you change?” with an empty box below it. Only the question format differs, and yet the two answers end up in different worlds.

Question formats are the shapes a question uses to collect an answer: they decide whether respondents pick from a fixed list, place themselves on a graded scale, sort statements into an order or write freely in their own words, and therefore whether you finish with numbers or with text. By the end of this guide you will recognise the common formats, the words used for them in British survey practice, and how to match a format to a research aim.

📌 Key points at a glance

• A question format decides how an answer is collected, not what is being asked.
• UK government guidance splits closed-ended questions into limited choice and multiple choice.
• Response categories are sorted into scalar and non-scalar, not into rating and non-rating.
• Selection questions, scales, grids, ranking and free text cover almost every need.
• Five points is the stated minimum for a Likert scale in UK guidance.

What is a question type?

A question type is the shape of a question in a questionnaire, meaning the combination of a question stem and the answer format offered with it. The type sets out how respondents are allowed to answer, and in what form their answer will reach you once you analyse the results.

Put simply, the stem says what is being asked and the type says how you may answer. Both together make up what a person sees on screen. The same underlying question about satisfaction with campus catering can be asked as a scale from 1 to 5, as a choice between four labelled options, or as an empty text box, and each of the three measures something different.

Several words circulate for the same thing. Question type, question format and question form are interchangeable, and question format is the umbrella term used in UK government guidance. Response format is narrower and covers only the second half, the answer side without the stem. Level of measurement is a separate matter again: it describes which calculations your data will support, and it follows from the question type rather than the other way round.

Why the distinction is more than pedantry becomes obvious at the analysis stage. Anyone who measures satisfaction as a choice between “good”, “average” and “poor” cannot report a mean from it, however obligingly a spreadsheet lines the numbers up. Which calculations your level of measurement permits is decided the moment you pick the question type.

The official UK taxonomy of question formats

British survey practice has its own vocabulary for question formats, and the vocabulary does not map neatly onto the labels used by survey software. Knowing the official terms matters if you write for a UK audience, cite British sources in a dissertation, or hand your data to anyone working in the Civil Service.

The reference point is the questionnaire design guidance published by the Government Analysis Function and owned by the Office for National Statistics Data Quality Hub. Written for social researchers, survey designers and form designers, the guidance sets out the terms government researchers use, and it is freely available, which makes it an easier citation than a paywalled textbook in an undergraduate or postgraduate project.

Open-ended, closed-ended, limited choice and multiple choice

The guidance opens with the familiar split between open-ended questions, which let respondents answer in their own words, and closed-ended questions, which offer a range of the most likely answers to select from. The second branch is where British usage diverges: closed-ended questions are divided into limited choice questions, which allow only one response category, and multiple choice questions, which allow more than one.

Notice which label is missing. Most survey software, and most guides written outside the UK, call a one-answer question a single choice question. UK government guidance never uses that wording; the one-answer format is a limited choice question. Anyone writing a methods chapter for a British institution is better off with the official term and a note of the software label in brackets. The source is worth reading in full: Questionnaire design guidance, Government Analysis Function (2023).

Scalar and non-scalar response categories

Underneath the closed-ended branch, the guidance sorts response categories into scalar and non-scalar. Scalar categories cover a range of opinions on a topic and appear as likelihood, agreement, satisfaction, frequency or quality scales. Non-scalar categories have no scale behind them at all, and the example given is a Census 2021 question asking whether a household owns its home outright, owns with a mortgage, part-owns and part-rents, rents, or lives there rent-free.

German-language survey methodology draws a comparable line but names it differently. Lenzner and Menold (2015) split answer formats into open questions, closed questions with a rating scale and closed questions without one, in the GESIS Survey Guidelines on question wording. Both traditions agree on the substance: a graded list carries meaning in its order, an ungraded list does not, and the two therefore obey different rules.

Where the function of a question comes in

A second axis runs across every one of these formats, namely what a question is for within the running order. A filter question decides the route through the form, a demographic question supplies the characteristics for group comparisons, and an attention check tests whether someone is still reading. Function drives position, not response format, so a filter question can perfectly well be built as a limited choice question. Sequencing is covered in the guide to creating a questionnaire.

Question types in a questionnaire: the overview

Ten formats cover almost everything you will meet in practice, from an undergraduate dissertation to a customer study. The list below is sorted by response format, starting with the selection questions and ending with free input.

Not every questionnaire needs all ten. A good one usually gets by with three to five different formats, and the variety then looks after itself, because different content asks for different shapes.

Two notes on the list. A yes or no question is strictly a limited choice question with exactly two options, and it stands on its own here only because it plays such a distinct part as a filter. The Net Promoter Score question is a rating scale with fixed labelling and a fixed calculation, so it is a standard rather than a new format.

Selection questions: limited choice and multiple choice

Selection questions put a list of possible answers in front of respondents and ask them to pick. The only difference between the two versions is how many ticks are allowed, and that small decision changes the analysis completely.

With a limited choice question the answers add up to 100 per cent, because each person counts exactly once. With a multiple choice question they do not, because one person can tick three boxes and then appears in three categories. Overlook that when you write up your results and you will publish percentages summing to well over 200, which tell the reader nothing.

Limited choice: exactly one answer

A limited choice question fits wherever the categories genuinely exclude one another. The qualification someone is working towards is a good example: an undergraduate degree, a taught master's, a research master's and a doctorate cannot apply at the same moment, so a single tick is right. The list must also cover every case, or respondents will settle for a category that does not describe them. Exhaustive lists are covered in the guide to single choice questions.

Multiple choice: several answers at once

A multiple choice question fits where several answers can apply at the same time, such as modes of transport used or events attended. Convenience for the respondent is exactly what tempts designers into letting the list grow too long. As a rule of thumb, anything past ten options stops being read and starts being skimmed. Analysis and the usual pitfalls are covered in the guide to multiple choice questions.

Why the order of the options counts

Position in the list changes how often an option is chosen, and it does so in opposite directions depending on how the survey is run. Lenzner and Menold (2015) describe both effects: with a visual presentation, on screen or on paper, the earlier alternatives are picked more often, while with questions read aloud over the telephone the later ones are.

UK guidance adds a related instruction, namely to place response categories that are more socially desirable than others at the end of the list, which helps to reduce bias. Randomising the order per respondent is the practical fix in an online questionnaire.

Scale questions: rating, Likert and Net Promoter Score

Scale questions offer a graded answer dimension on which respondents place themselves. Recording not just whether something applies but how strongly, they are the most used format of all in satisfaction and attitude studies.

All scale questions share one property: the order of the steps is fixed and carries meaning in itself. Beyond that they differ in what they measure and in what you are allowed to calculate afterwards.

The rating scale as the basic form

A rating scale asks about a single property across several graded steps, for instance satisfaction running from “highly satisfied” to “highly dissatisfied”. The decisions that matter are the number of steps, the labelling, and whether to offer a neutral middle. All three shape the result more strongly than the wording of the stem does, which is why they get a full treatment in the guide to the rating scale.

The Likert scale for attitudes

A Likert scale puts statements to respondents, who then agree or disagree in graded steps, usually five, from “strongly agree” to “strongly disagree”. Several such statements on one topic together form a measurement instrument. UK guidance is unusually specific here: a minimum of a 5-point Likert scale is recommended, a 7-point scale gives more detailed answers for analysis, and the scale should always include a neutral option in the middle.

British guidance therefore settles a question that many methodology guides leave open. Whether to offer a neutral midpoint is usually presented as a trade-off to be weighed case by case, while the Government Analysis Function (2023) simply prescribes one for every Likert scale. Where the equal-spacing assumption behind such scales came from, and why it does not hold, is set out in the guide to the Likert scale.

The Net Promoter Score question as a fixed standard

The Net Promoter Score question asks a single recommendation question on a scale from 0 to 10 and reduces the answers to one figure by a fixed rule. Comparability across organisations and over time is the advantage. The drawback is the one every single-figure metric carries, namely that it reports where you stand but not why, which is why an open follow-up question almost always sits behind it.

💡 Tip

Stick to one scale direction and one number of steps across a whole questionnaire. Mixing five steps running positive to negative in the first block with seven steps running the other way in the second produces answers that look like data but partly measure how quickly respondents adjusted. UK guidance says the same in a single line: keep the scales in the same order throughout.

Grid questions and ranking questions

Grid and ranking questions bundle several judgements into a single question. Both save a great deal of space and time, and both ask noticeably more concentration of respondents than a plain selection question does.

Each format has its own trap, and in both cases the trap is size. A grid with five rows gets filled in carefully, while a grid with twenty rows gets ticked through.

The grid question for item batteries

A grid question, also called a matrix question, lists several statements underneath each other and rates them all on the same scale. The stem appears once above, the scale once alongside, and respondents work down row by row. For statements that belong together, no format is more efficient, because respondents build the yardstick only once and then keep it for every row.

Why a grid beats “select all that apply”

Lists where several answers can apply come in two builds, and the two are not equivalent. The convenient version has respondents tick everything that applies in one list, while the more demanding version answers each row separately with yes or no. Lenzner and Menold (2015) recommend the second form explicitly, because it avoids the order effects from the previous chapter, and the underlying study by Smyth and colleagues (2006) adds that each option is processed more thoroughly.

In practice: if you genuinely need clean proportions from a multiple-answer question, build it as a grid with yes and no columns rather than a tick list. The price is a few extra seconds per question, the gain is figures that do not depend on how you happened to sort your list. As an aside, UK guidance advises dropping the word please from such instructions, so “Select all that apply” rather than “Please select all that apply”.

The ranking question for real priorities

A ranking question asks respondents to put objects in order instead of rating them one by one. Forcing a decision is exactly what a scale fails to do: someone who marks five criteria as “very important” has said nothing, while someone who ranks them has. The limit sits at roughly five to seven objects, beyond which sorting becomes a chore and the lower places are guessed.

⚠️ Take care

Grid questions are the first thing to break on a narrow screen. Since a large share of responses arrives from mobile phones, treat five answer columns and eight rows as the ceiling. Anything beyond that belongs in two separate questions rather than a smaller font.

Open-ended and semi-open questions

Open-ended questions offer no answer options at all, only the stem and an empty box. As the single format that does so, they return content nobody thought of while the questionnaire was being built, and that is where their value lies.

UK guidance puts their main use narrowly, describing open-ended questions as suitable for exploratory phases when there is not yet enough information to develop appropriate categories. A common prejudice says many respondents are overwhelmed by them. Züll (2015) summarises the research differently in the GESIS Survey Guidelines on open questions: as a rule all respondents are able to answer open questions, and what decides the outcome is interest in the topic rather than ability.

When an open-ended question earns its place

Open-ended questions pay off in four situations: when you are testing knowledge and want to rule out guessing, when you do not yet have an overview of the field, when the list of possible answers would be unmanageable, as with occupations, and when you want to avoid steering respondents with prompts. The price is the coding effort, because text must be categorised before it can be counted. Coding is explained in the guide to open questions.

The semi-open question as a hybrid

A semi-open question offers a list of options and adds a free field, usually labelled “Other”. Standard practice for any content question whose answers cannot be listed exhaustively, the free field prevents the damage that follows from leaving the category out: if an obvious answer is missing, the question is either skipped or something arbitrary is ticked. All three formats are compared in the guide to open and closed questions.

How the size of the box steers the length of the answer

Box size influences how fully people answer, and hardly anyone uses that on purpose. Züll (2015) points to studies in which the size of the field provided in self-completion questionnaires is positively related to the length and detail of the answers. A small box signals that a keyword will do, while a large one invites a paragraph. If you want keywords, make the box small and say what form of answer you expect.

Filter questions and demographic questions

Filter questions and demographic questions belong to the second axis, because both are defined by their function in the questionnaire rather than by their response format. Neither usually measures the thing you actually care about, and hardly any questionnaire works without them.

Take a study on housing. Whether someone can report the service charge on a flat they own depends on whether they own one at all, and that preliminary question decides the route through the rest of the form.

Filter questions steer the route through the form

A filter question decides which questions come next. UK guidance gives the plainest possible instruction: if you ask whether someone was born overseas and the answer is no, do not put the follow-up questions designed for people who were. Lenzner and Menold (2015) add a caveat that is easy to miss when planning, namely that filters are simple to implement in computer-assisted surveys and cause respondents no difficulty, while numerous or complex filters should be avoided on paper.

Demographic questions supply the comparison groups

Demographic questions record characteristics such as age, sex, level of qualification or region. Their purpose is not the value itself but the later analysis by group: without them you cannot show that first-year undergraduates answer differently from finalists. Such questions usually sit at the end, because they motivate least, and they are kept to a minimum, because every extra characteristic weakens anonymity. Wording that has proved itself is collected in the guide to demographic questions.

Sensitive questions and the closing question

Sensitive questions need handling of their own, and the Government Analysis Function (2023) sets out four rules: ask only what the research objectives require, keep such questions away from the start because early drop-out rises, explain why you are asking, and include a “Prefer not to say” option.

A closing question is almost always open and asks whether there is anything else the person would like to add. Züll (2015) describes that question as a motivational device in its own right, because it hands the word back to the respondent after a long run of fixed options.

Which question type suits which aim?

The choice of question type follows from the statement you want to make at the end, not the other way round. Anyone who knows which sentence should appear in the results chapter has usually made the decision already without noticing.

The table below reads from left to right: find your aim in the first column, and the matching format with its reason sits beside it. Where several aims apply, take the row closest to your research question.

One rule sits above every row: analysis first, question type second. Writing down in advance which table or chart should exist at the end shows immediately whether the planned question can produce it. Ten minutes of planning at this stage saves you an entire afternoon later!

Common mistakes when choosing a question type

Most problems in questionnaires come from a question type that does not match the content, not from badly worded stems. Four of them turn up again and again, and all four can be fixed in minutes before the survey goes out.

What the four have in common is timing, because they only become visible during analysis, and by then it is too late. A questionnaire that has already reached two hundred people cannot be corrected.

Mistake 1: the response categories do not fit the question

Lenzner and Menold (2015) give an example you will recognise everywhere afterwards. The stem asks whether someone was an active member of a club in the past twelve months, and the options underneath are “often”, “sometimes”, “rarely” and “never”. The stem calls for a yes or a no, the categories call for a frequency, and respondents have to guess what was meant. UK guidance states the same requirement in one line: response categories must fit the question.

Mistake 2: the categories overlap

Response categories have to be exhaustive and mutually exclusive, or an answer cannot be assigned unambiguously. Time and age bands are the classic case, because someone who gets up at exactly half past six can tick both “between 6:00 and 6:30” and “between 6:30 and 7:00”. UK guidance puts avoiding overlap at the head of its rules for response categories. More rules for clean options are in the guide to closed questions.

Mistake 3: a double-barrelled question dressed up as one

A double-barrelled question packs two questions into one, and no question type can rescue it. UK guidance uses the example “How would you rate the course facilitator and the materials provided?”, which asks for two judgements at once and must be split into two questions with their own scales. Grid questions make the mistake easy to commit, because a row naming two things looks tidy in a table and cannot be analysed.

Mistake 4: the scale changes halfway through

Changing scales are the least conspicuous of the four mistakes and one of the most expensive. If the first block uses four steps, the second uses five and the third reverses the direction, the answers are no longer comparable and some respondents will miss the switch. Settle on one scale before you start writing and see it through, even where a single question would be slightly more precise with seven steps.

Conclusion

The question type is not a formality at the end of the writing, but the first decision about what your questionnaire will be able to say. Anyone who remembers the two axes, response format and function in the form, makes that decision in two small steps instead of one large one. And anyone who writes down the final table in advance has usually found the right format already.

For a British readership a second habit is worth keeping, namely using the official words. Limited choice, multiple choice, scalar and non-scalar are the terms in UK government guidance, and a methods chapter that matches them is easier to defend in a viva than one built on software labels.

Question types are chosen when the analysis is planned, not when the question is written.

Where to go next

Four further guides pick up where this one stops. The first walks through building a questionnaire from the opening question to the last, the second sets open and closed formats against each other in detail, the third collects wording you can adapt, and the fourth clears up the difference between a survey and a questionnaire, which UK guidance treats as two distinct things.

• Building your form right now? Creating a questionnaire: steps, structure and an example
• Weighing open against closed properly? Open and closed questions compared
• Looking for ready-made wording to borrow? Questionnaire templates, examples and samples
• Unsure about survey versus questionnaire? Survey and questionnaire: the differences

Want to try every question type once?

With empirio.ai, an online survey tool from Germany, you can put limited choice, scales, grids, ranking and free text into one questionnaire and watch the analysis update as answers arrive. Create a survey for free