Abstract The main objective of this study was to examine Steel's [Advances in Experimental Social Psychology 21 (1988) 261] self-affirmation theory, by checking the effects of different levels of manipulated control on participants' performance on personality tests. Specifically, we assessed the impact of participants' level of control (i.e. prior familiarity of questionnaire's items, choosing of test content) and anonymous or identified data collection on two different forms of socially desirable responses: impression management and self-deception [Journal of Personality and Social Psychology 46 (1984) 598]. Second, we compared these effects on paper-and-pencil versus computerized testing conditions. In Study 1 ( N=91) we showed that perceived control can be manipulated in computerized tests and that higher level of control is related to more positive attitudes towards the test and to lower levels of anxiety experienced during test taking. In Study 2 ( N=200) we found a significant positive relationship between manipulated control and impression management. The hypothesis claiming a significant relationship between controllability and self-deception has not been corroborated. In addition, no differences were found between the paper-and-pencil and the computerized mode of administration on measures of perceived control, trust, candor, and social desirability.