Response variation and location of questions within a questionnaire

In a questionnaire administered to a general population sample of 1673 adults, a 14-item scale of social desirability was divided into two groups of questions, almost evenly divided between desirable and undesirable behaviour, the percentage answering 'true' was higher when the question was asked early than when it was asked late. The difference was statistically significant for five of the 11, four of them relating to socially undesirable behaviour. The results indicate that responses can be dependent on the location of questions in a questionnaire, and that the effect of question location may not be the same for groups with different personal characteristics. For this reason, it seems unwise to alter the administration or design of a questionnaire in any way if the results are to be compared directly to those of other studies, unless the altered questionnaire is standardized against the original version.