INTRODUCTION : Despite the explosion in genetic association studies over the last decade, clearly identified genetic risk factors for depression remain scarce and replication studies are becoming increasingly important. G-protein-coupled receptor 50 (GPR50) has been implicated in psychiatric disorders in a small number of studies, although not consistently. METHODS : Data were obtained from 1010 elderly men and women from the prospective population-based ESPRIT study. Logistic regression and survival models were used to determine whether three common GPR50 polymorphisms were associated with depression prevalence or the incidence of depression over 12-years. The analyses were adjusted for a range of covariates such as comorbidity and cholesterol levels, to determine independent associations. RESULTS : All three variants showed some evidence of an association with late-life depression in women, although these were not consistent across outcomes, the overall effect sizes were relatively small, and most would not remain significant after correction for multiple testing. Women heterozygous for rs13440581, had a 1.6-fold increased risk of baseline depression, while the odds of depression comorbid with anxiety were increased fourfold for women homozygous for the minor allele of rs2072621. When depressed women at baseline were excluded from the analysis, however, neither variant was associated with the 12-year incidence of depression. In contrast, rs561077 was associated with a 1.8-fold increased risk of incident depression specifically. No significant associations were observed in men. DISCUSSION : Our results thus provide only weak support for the involvement of GPR50 variants in late-life depression, which appear specific to certain subgroups of depressed individuals (i.e., women and those with more severe forms of depression).