Three versions (22-item, 10-item, and 7-item) of the social support questionnaire (F-SOZU) were psychometrically evaluated in two clinical and three non-clinical Austrian samples. The distribution of sum scores in all three versions was negatively skewed; means on the item-level were in the upper region of the five-point scale (M > 4.0) in all non-clinical samples. Internal consistency estimates were found to be satisfying for the total test scores (alpha > 0.85). The 10-item and the 7-item forms correlated highly with the 22-item form total score (r > 0.90). Principal components analysis supported a one dimensional solution in all forms. The discussion focuses on the problem of the highly skewed test scores. It is also argued that the use of the 7-item version might be preferable and more efficient if the researcher is only interested in obtaining a global score for perceived social support.