A statistical reevaluation of the STAI anxiety questionnaire.

This study examined the capability and efficiency of two commonly used scales, the trait and trait and state inventories (Spielberger et al., 1970), to measure anxiety construct (N = 100). The probabilistic Rasch Model (Wright & Masters, 1982) was applied to athletes' responses to trait and state anxiety items before competition (stress). The analysis indicated that several items of both scales produce misfit responses, share identical locations on the continuum, and do not produce equal units of measurement. The findings question the generalizability of the research on anxiety. Recommendations for improving the anxiety tests are made.