The chapter concerns social desirability in personality assessment. After a short review of existing models of social desirability, our own is outlined. The main tenet of the model is that test items themselves may evoke socially desirable responding, particularly when clearly evaluative. After a short review of existing techniques for reducing social desirability in self-rating measures, our own is introduced. The main feature of the technique is the rephrasing of items as to make them evaluatively neutral. The few studies that have examined the method so far are described. Finally it is argued that from a psychometric perspective separate measurement of personality and social desirability would be ideal, and that purer measures would enable more efficient testing of personality models.