A survey was conducted of New Zealand personnel consultants. Their beliefs about the validity of various selection tools and their claimed usage of these tools was then compared with the validities in a previously published meta-analysis. The experts claimed to use the predictors they believed to be most valid. However, their beliefs about validity were unrelated to empirically demonstrated validities (Spearman's rho = -0.06). Suggestions were made on the types of research that are needed to improve predictive ability in selection and on the ways in which practitioners can use existing research.