Abstract
Survey methodologists have given considerable attention to minimizing the response effects associated with asking sensitive questions. However, this work is largely atheoretical; little attention is paid to issues of definition, rather a common sense understanding of what is sensitive is assumed. Such research largely focuses on how sensitive behaviours are reported and provides little insight into the relationship between responses to attitude items and their sensitivity. Recent work of cognitive psychologists, although drawing attention to the way in which responses are affected by the immediate context of the questionnaire, does not extend to suggesting whether questions that vary in their sensitivity are likely to be differentially affected by this. In the light of these issues five studies are reported that suggest the importance of taking the sensitivity of attitude items into account. Sensitive items are defined in relation to the notion of 'threat' and the context dependence of what is considered to be sensitive is recognised. The sensitivities of a pool of attitude items for a particular population is thus established in advance of being used in further work. The different patterns of responses associated with high and low sensitive items are noted and are seen to bear an orderly relationship with agreement ratings that is not explicable in terms of the orientation of the items. Results also indicate that items that varied in their sensitivity are differentially affected by manipulations in the context of threat within which the questionnaire was embedded. The nature of these effects is explored as is the notion that responses to sensitive items may be used to cope with threat. The implications of these results for research using attitude items within questionnaires is discussed. It is suggested that knowing the sensitivities of the items being used is a necessary first step in being able to determine the extent to which these affect the results.