Psychiatric genetics has been hampered by the fact that initially exciting findings from underpowered studies are so often not replicated in larger, more powerful, data sets. Here we show that the claims of Zhou et al. that neuropeptide Y (NPY) diplotype-predicted expression is correlated with trait anxiety (neuroticism) is not replicated in a data set consisting of phenotypically extreme individuals drawn from a large (n = 88,142) non-clinical population. We found no association between NPY diplotype or diplotype-predicted expression and neuroticism. Our reply to Zhou and colleagues forms part of a larger debate (see, for example, http://www.nature.com/news/2008/080709/full/454154a.html) about the efficacy and replicability of candidate driven versus genome wide approaches to psychiatric genetics.