General parameters:

There are no hard and fast rules about how a questionnaire should be constructed, and to some extent you will be guided by the topic under consideration - for example, you may be able to draw on designs used by other researchers in the field, as identified through your literature review. You may, though, find it helpful to bear the following points in mind:

• the flow, structure and length should encourage and maintain the respondents' attention;

• the first few questions set the tone of the questionnaire and should, as far as possible, be directly relevant to the respondent, short, non-controversial and easy to answer;

• longer, more complex questions, such as those designed to gauge attitudes, should be spread throughout the questionnaire, rather than grouped together in one section;

• sensitive or personal questions should be located at the end;

• ensure that all the questions asked are relevant to your research and will provide suitable data for the particular form of analysis which is required;

• check that the questions are clear and unambiguous - avoid overlong or "catch all" questions;

• instructions for completion must be clear, especially if the questionnaire is to be self-administered by the respondent - for example, if some sections are only to be answered by certain respondents, this should be indicated by the use of "filters", such as "If No, go to question 9";

• think carefully about the layout, including font size, clarity, and the use of any graphics; and

• ensure that data analysis is facilitated by clear coding procedures for answers (we will be looking at coding in more detail later in the section).

Gathering data about patterns of behaviour often depends on the ability of respondents to remember and describe events and activities which occurred in the past accurately. As you prepare your questionnaire, think carefully whether the information you are asking for is feasible. For example, you may have no difficulty recalling how many times you have used a train in the last two weeks, but what about the last six months? If respondents cannot remember correctly, the information they provide will be inaccurate. If precise data about past events is required, especially those which may at the time have appeared trivial or mundane to the respondents, it may be more appropriate to use a diary method to obtain it instead.

Sensitive areas of questioning need careful handling in order to obtain honest answers. Sometimes, a series of indirect questions can be used, instead of a direct question which may be unlikely to provoke an accurate response. Alternatively, a series of prepared categories could be offered for responses, to show that the respondent is not alone in engaging in this particular behaviour.

It is important to avoid putting ideas into respondents' minds. For example, a respondent could be "led" by the wording of a question to think that a certain response is sought or may answer in a way which they think will lead the researcher to form a good opinion of them (known as "social desirability" responses).