A meta-analysis was conducted on studies reporting data on associations between candidate genes and human personality. Studies reporting data for psychiatric populations (including organic disease and substance abuse) were excluded. A total of 46 studies contributed to the analysis. Pooled data using a fixed-effects model suggested significant associations between the 5HTT LPR, DRD4 c>t, DRD4 length, DRD2 A1/A2, DRD3 A1/A2 polymorphisms and personality traits. A multivariate analysis using a mixed-effects model and including age, sex and predominant ethnicity as covariates was applied to the analyses of 5HTT LPR and DRD4 length polymorphism data. Only the association between the 5HTT LPR polymorphism and avoidance traits remained significant (P=0.038). However, sensitivity analyses excluding data from studies reporting allele frequencies not in Hardy-Weinberg equilibrium and unpublished data resulted in this association no longer being significant. Implications for the design of future association studies of human personality are discussed, including the likely sample sizes that will be required to achieve sufficient power and the potential role of moderating variables such as sex.