Conferences

Why do respondents answer questions in surveys they do not know anything about? Fazio interpreted this behaviour in the context of attitude strength (Fazio 1986, 1989) . Thereby, Fazio defined attitude strength as the strength of the cognitive association between an attitude object and its evaluation. Further, if there is no association between attitude object and its evaluation, the answer given is defined as nonattitude.

The research question of this paper is how nonattitudes can be identified. To answer this research question, a methodological comparison of the identification with response latencies on the one side and with behaviour coding on the other side is undertaken. A German-wide representative CATI-survey, conducted in 2005, serves as data basis. Within this survey, attitudes towards known and unknown (i.e. fictive) attitude objects were measured, assuming that substantial answers to unknown attitude objects represent nonattitudes. Response latencies have been measured for all items of the survey and about 50% of all respondents were additionally tape recorded during the interview. Three basic steps were undertaken to answer the research question: First, substantial answers to attitude questions with fictive attitude objects have been analysed according to their response latencies. Second, behaviour coding was undertaken for these fictive questions. Aim of those first two steps was to identify response latency patterns and behaviour coding patterns of respondents while answering nonattitudes. As a third step, patterns found from response latencies and behavior coding were applied to attitude questions with non-fictive attitude objects. Items found to be a possible nonattitude in reference to response latencies and/or behaviour coding were cross-checked regarding their attitude strength, measured with meta-criteria. Further, attitude stability of possible nonattitudes is checked with multi-group analysis within structural equation modelling.

Results indicate that Nonattitudes can in fact occur with long and short response latencies. Further, respondents with long response latencies can be differentiated through behaviour coding into those with strong attitudes and those holding Nonattitudes. Thus, in the case of long response latencies behaviour coding can add additional information one would not have gained through the analysis of response latencies alone. However, in case of short response latencies behaviour coding is not applicable and response latenices identify more reliable Nonattitudes and strong attitudes.