Question answering processes in attitude surveys
Response effects and response latencies

This project aims to investigate the cognitive processes underlying the answers to attitude questions in surveys using reaction time data. A vast body of research shows that the answers to survey questions are influenced by many seemingly irrelevant characteristics of the questionnaire and the context, such as the question wording, the order of the questions or the response options offered (see Schuman & Presser 1981). These so-called response effects are a threat to the validity of surveys. In order to provide researchers with solid advice on how to design their questionnaires, insight is needed into how and why these (con)text characteristics influence the answers.

A good way to come to such an explanation of response effects is to investigate in which stage of the question answering process they emerge (see Tourangeau, Rips & Rasinski 2000). Are the answers 'distorted' in the stage of question comprehension? During the retrieval or formation of the attitude? During the integration of the information retrieved into a judgment? Or when the judgment is translated into one of the answering options?

Holleman (2000) discriminated between these stages by means of correlational experimental research, and was able to show that a well-known response effect, the forbid/allow asymmetry, arises during the last stage: questions worded with the verb forbid measure a similar attitude compared to equivalent questions worded with the verb allow, but the way these answers are expressed on the answering scale differs.

Another way to investigate the process of question answering underlying response effects is to measure response latencies. The general assumption using response latencies is that tasks asking more cognitive effort last longer (see Gernsbacher 1994, Fazio 1990). The question answering process will slow down if the text is more difficult to comprehend, if it refers to an attitudes or to beliefs that are not available or accessible, if an integrative judgment has still to be formed or if it is difficult to translate the judgment into the answering options (Bassili 1996).

The central research goal of this project is to explain response effects by relating them to the underlying cognitive process of question answering.