A questionnaire (German Fragebogen) is a text serving the guided production and collection of data that are directly recorded. It consists of a set of items that trigger a response on the part of the subject who is to answer or fill in the questionnaire. This response is recorded. Traditionally, it is written down; but it may also be audio- or video-taped.

The items in the questionnaire may be linguistic or non-linguistic stimuli. Non-linguistic stimuli may, e.g., be pictures. Linguistic stimuli may be questions, but they may also have some other illocutionary force apt to trigger a response.

The questionnaire is introduced by an instruction telling the subject what kind of response is expected from him. The instructions and the questionnaire items themselves may be formulated in order to constrain the kind of response more or less strictly. Anything between the poles of giving the subject full freedom in the form and extent of his reaction and forcing him to mark an item in a multiple-choice question is possible.

If one wants to apply a qualitative analysis to the responses, it may be useful to leave full freedom to the subjects. If one wants to apply a quantitative analysis, it is useful to constrain the kind of response as far as possible. For instance, if one only wants to find out whether subjects know the color of the liberty statue, then the questionnaire item should not read:

What is the color of the liberty statue? 

Instead, it should read:

Mark the color of the liberty statue:☐ white
☐ yellow
☐ red
☐ green
☐ blue
☐ brown
☐ black

Otherwise, one runs the risk of obtaining answers like grean [sic!] or emerald or something between blue and green, which create problems in the evaluation of the answers to the questionnaire.

In analyzing the data thus obtained and evaluating them, there are usually data whose form does not conform to the author's expectations, be it that the subject misunderstood the question or that the author did not succeed in asking what he was actually interested in. To minimize this problem, one applies a draft version of the questionnaire to a pilot group of subjects, analyzes their responses and produces the next version on the basis of the results obtained.