Response Style

Response styles are distinctive ways of responding to questionnaire surveys that are unrelated to the content of the actual survey items. If the response categories are laid out in terms of amounts of agreement or disagreement, respondents may tend to agree with all the items in a survey, even where the items have meanings that contradict one another. This is known as acquiescence. Alternatively, respondents may favor more extreme responses, agreeing strongly or disagreeing strongly with most items. Finally, some respondents may favor moderate responses to all items. Thus, response styles can also favor either extremity or moderation.

These types of response styles are often referred to as response bias, because it is thought that they reduce the validity of mean scores on a given measure. This will be particularly true when comparisons are made between mean scores obtained from different samples that vary in response style. Response style is less problematic when comparisons are made between scores on different scales that have been completed by the same set of respondents. However, even when several measures are collected from the same set of respondents, the association between scores on each measure can be inflated if there is any tendency for respondents to endorse responses that they see as more socially desirable. Socially desirable responding can occur in two ways. Firstly, respondents may wish to give a good impression of themselves to others, and secondly, they may think of themselves in excessively positive ways. Because what is socially desirable varies between cultural groups, these tendencies can yield acquiescence in some circumstances and extremity or moderation in other circumstances.

Acquiescence is typically measured as the overall mean of responses to a set of survey items that are conceptually unrelated. Extremity is measured by counting the frequency of extreme responses to a set of items, for example, ones and fives on a 5-point scale, whereas moderation would be counted for instance as the frequency of threes on a 5-point scale.

Response styles can vary between individuals of differing personality, between groups in which particular styles of communication are favored, and on the basis of the structure of different types of survey instrument. Variation due to personality is less often a problem, since selection of an adequately sized sample of respondents will enable individual variations in response style to be discounted. However, in some circumstances it is important to test whether any respondents have employed a random response style. This may occur particularly when surveys are being completed for credit or for money. Random responding can be estimated and controlled by inserting a few items with no content into a survey, on each of which respondents are simply instructed to check a different specific response option. Data from respondents who consistently fail to follow these instructions are probably not attending adequately to the other items in the survey. Their data should be discarded.

Group differences and measurement structure differences are more important determinants of response style and each requires more detailed consideration.