A sample of 121 university students completed the NEO-FFI either as part of a norming exercise (honest condition) or under explicit instructions to create a favourable impression of themselves (fake good condition) and subsequently had judges who knew them for at least 12 months rate them using scales based on the NEO-FFI. Mean scores on four of the five NEO-FFI scales (extraversion, neuroticism. agreeableness, and conscientiousness) differed significantly between the two conditions. Correlational analysis indicated that, with the exception of agreeableness, there were statistically significant reductions in the validity of the scales against judges ratings in the fake good as compared with the honest condition. A series of moderated regression analyses, in which scale score and condition were used jointly to predict judges' ratings, confirmed the inferences from the correlational analysis in all cases except for neuroticism. It was concluded that, consistent with the popular view of self-report tests of personality, deliberate attempts to fake seriously compromise the validity of these tests. Agreeableness, however, appears to be an exception to this generalisation.