Attitudes

Direct Evaluations

Single-item direct measures. Laboratory experiments and attitude surveys frequently use single items to obtain direct evaluations of the attitude object. Confronted with the item ‘Do you approve of the way the President is doing his job?’ respondents may be asked to express their degree of approval on a five-point scale that [Page 111]ranges from ‘approve very much’ to ‘disapprove very much’. Such single items can be remarkably good indicators, especially for well-formed attitudes toward familiar objects. They are sometimes found to have quite high levels of reliability and to correlate well with external criteria. For example, the single item ‘I have high self-esteem’ (attitude toward the self), assessed on a five-point scale ranging from ‘not very true of me’ to ‘very true of me’, was found to have a test-retest reliability of 0.75 over a four-year period, compared to a reliability of 0.88 for the multi-item Rosenberg Self-Esteem Scale (Robins, Hendin & Trzesniewski, 2001, Study 1). Moreover, the single- and multi-item measures correlated highly with each other, and they had comparable correlations with various external criteria (e.g. self-evaluation of physical attractiveness, extraversion, optimism, life satisfaction).

However, single items do not always exhibit such favourable psychometric properties. They often have low reliabilities and can suffer from limited construct validity. Many attitude objects are multidimensional and a single item can be ambiguous with respect to the intended dimension (e.g. ‘religion as an institution’ vs. ‘religious faith’). Furthermore, single items contain nuances of meaning that may inadvertently affect responses to attitudinal inquiries. An item inquiring whether the United States should allow public speech against democracy leads to different conclusions than one asking whether the United States should forbid such speech (see Schuman & Presser, 1981). In addition to such framing effects, research has revealed strong context effects in attitudinal surveys. Respondents tend to interpret a given item in light of the context created by previous questions. Thus, responses to questions about satisfaction with life in general and satisfaction with specific aspects of one's life, such as one's work or romantic relationship, are found to be influenced by the order in which these questions are asked (Schwarz, Strack & Mai, 1991).

...