PUB - Publikationen an der Universität Bielefeld

Test users assume that all participants use the response scale in the same manner. Research on the occurrence of response styles has shown that this often is not the case. Studies that examine response styles usually collapse the middle category with the next lower category to facilitate interpretation. However, this practice could distort the analysis of response behavior since participants who chose different response categories are treated as if they belonged into one group. In this study, we analyzed the German NEO-PI-R (Ostendorf & Angleitner, 2004) with Mixed Rasch Models to investigate whether merging two response categories is justified. Mixed Rasch Models were computed on a sample of 5862 participants to identify latent classes of different response styles. These latent classes were examined regarding their response behavior and the intraindividual consistency with which participants used the same response style over different personality traits. For most facets of the NEO-PI-R, 3 to 5 classes fit the data best. The three most frequently occurring response styles can be interpreted as 1) participants who avoid extreme responses, 2) participants who prefer extreme responses, and 3) participants who use the response scale as intended. Analyses were conducted once with the NEO-PI-R's five response categories and once with the second and third category merged. Results indicate that collapsing the middle category witha neighboring category misrepresents the participants' response behavior. This practice of combining responses from participants who choose the second and third category is shown to disregard the trait differences between these participants.