A number of studies have reported an association between large CAG/CTG repeats and both schizophrenia and bipolar disorder. Recently, we reported an inverse correlation between CAG/CTG repeat size and age in a health-selected population, raising the possibility that selection of control groups for physical health was a confounding factor in our previous association studies. We investigated this by health-selection of patients with schizophrenia and bipolar disorder. The maximum CAG/CTG repeat size remained significantly larger in probands with functional psychosis compared with control individuals, and in probands with a diagnosis of schizophrenia compared with control individuals. The number of probands in the healthy bipolar group was small, and although on average this group also had longer CAG/CTG repeats than control individuals, this failed to reach statistical significance. Our findings do not support the notion that the original results with psychosis as a whole, and schizophrenia specifically, are attributable to a stratification effect consequent on health selection. Nevertheless, we are unable formally to reject the hypothesis that the previously observed difference between bipolar probands and control individuals is the result of this phenomenon.