In 3 studies, I have tested the structure of different phrasing versions of the Private Self-Consciousness Scale (PrSC; Fenigstein, Scheier, & Buss, 1975) using exploratory and confirmatory factor-analytic methods and examined their predictive validity. For the original version of the PrSC, a 2-factorial model similar to Nystedt and Ljungberg's (2002) solution was found to best fit the data. When all scale items included extreme rate of occurrence words such as always, a single-factor solution emerged. Finally, when all words reflecting rate of occurrence were removed, again a 2-factor structure emerged, although different in item composition from that of the original version. In addition, different patterns of association emerged between the PrSC factors and depression and self-esteem for the extreme and neutral versions. I discuss the importance of reconceptualizing self-consciousness and the need for a new, theoretically based scale for self-consciousness.