Context effects in the measurement of attitudes: A comparison of the consistency and framing explanations

This paper investigates two competing social psychological explanations of a question order (context) effect in surveys. Using the topic of the public's understanding of science, the effect of answering knowledge questions on reported interest in science is measured using a split ballot field experiment in a national survey. The consistency explanation predicts that answering prior knowledge questions will increase reported interest for people who do well on these quizzes, but will decrease reported interest in science for those who do badly. Alternatively, the framing explanation attributes the effect to the content of the quizzes, rather than respondents' performance. This content influences how respondents understand what ‘interest in science' means. Here it is predicted that difficult quizzes will lead to lower reported interest regardless of performance. Our findings support the framing explanation. The implications for questionnaire design are discussed.