Question types

Explanations > Social Research > Measurement > Question types

There are many different types of question you can ask in a survey or interview.

Scale questions

Scale questions ask the subject to select an item or a value from a fixed scale.

How happy were you at work today? (please tick only one)
[ ] Very happy
[ ] Quite happy
[ ] Neither happy nor unhappy
[ ] Quite unhappy
[ ] Very unhappy

This is the basis of the very common Likert scale. These questions are useful for converting choice into a variable score which can be analyzed, for example discovering correlation with another variable.

Bipolar questions

Bipolar questions have a scale with two ends. The subject is asked to show bias towards either end.

What is your balance of preference here?

I like going for walks.

[ ]

I like watching a movie.

This type of question can be use to force choice between very similar items (thus causing a semantic differential choice) or very different questions.

Dichotomous questions

Dichotomous (or binary) questions ask about black-and-white subjects, seeking only one of two answers. They are effectively scale questions with only two positions on the scale.

Did you vote in the last election?
Yes [ ] No [ ]

If you answered No, please go to section 2.

These are usually easy questions to answer and so can be good at the start, to get the respondent used to the idea of answering questions.

These are also useful as screening questions, as the wrong answer allows you to immediately reject the person or jump to another position within the questionnaire. When used in this way, dichotomous questions are also known as filter questions or contingency questions.

Selection questions

Selection questions ask you to make a choice from a list of items given. Single-selection questions ask you to choose only one item. Multiple-choice questions let you choose as many responses as you wish.

Which of the following places have you visited? (check all that apply):

[ ] Paris
[ ] London
[ ] Rome
[ ] Moscow

Ranking questions

Ranking questions ask questions that asks the respondent to sort items into a particular order, given criteria (typically preference):

Please rank the following places in terms of which you would like to visit (1 = most preferred destination, 2 = second most preferred, etc.):

[ ] Paris
[ ] London
[ ] Rome
[ ] Moscow

Ranking questions gather ordinal data, for example allowing you to understand comparative preferences.

Coding considerations

'Coding' involves turning answers into numbers and otherwise creating a set of data that can be analyzed, typically with a statistical package such as SPSS.

One of the trickiest questions is what to do with 'non' answers, such as:

No view: 'don't know', 'no opinion'.
Missing information: for example where no answer is given or it is partially complete.
Partial information: for example where you need a pair of items and you only have one.
Spoiled answers: for example where the respondent has checked more than one box when one was required.

The simplest approach is often to reject the whole respondent where they have not completed the questionnaire. This may be ok when this happens with only a few responses, but when many do this it may be for an important reason or because the questionnaire is badly designed (and may need doing again).

For individual 'no opinion' answers and missing individual responses, this can be included in the coding and statistical packages may handle these smoothly. Be very careful about adding 'no answer' as an actual answer. For example if 'agree' to 'disagree' is scored on a scale of 1 to 5, then 'no opinion' does not have a value of 6! Sometimes it can be reasonable to score these centrally, for example at 3 on a 1 to 5 scale.