We investigate whether survey answers of trained respondents differ systematically from answers of novice respondents. Focusing on difficult attitudinal or preference questions, we find that novice respondents answer �don�t know� significantly more often. Importantly, however, there is no systematic evidence for a conditioning effect on measured preferences (substantial answers). Overall, our results suggest that panel conditioning does not pose a serious problem, even in the case of rather difficult questions.