This study tested the association between tardive dyskinesia (TD) and polymorphic variations in (a) 2 cytochrome P450 (CYP) genes (CYP2D6 or CYP3A5), (b) 2 DRD2 variants (Ser311Cys and -141C Ins/del) and the Ser9Gly DRD3 variants, (c) 2 glutathione S-transferases (GSTT1 and GSTM1), and (d) variations in the PgP gene, MDR1. The study sample included 516 severely mentally ill patients from Central Kentucky facilities. Logistic regression models that included clinical variables associated with TD were developed. Gene variants were added to these clinical models. The total sample included 31% (162/516) with TD where 30% (49/162) of those had severe TD. Polymorphisms in DRD2, MDR1, and GSTT1 were never significant. Two gene variants appeared to be significant after adding them to the clinical regression models: (1) Ser9Gly DRD3 polymorphism was associated with severe TD (odds ratio for patients with 1 mutant allele when compared with individuals with 2 wild types was 2.5, 95% confidence interval 1.1-5.6, whereas the odds ratio for patients with 2 mutant alleles when compared with individuals with 1 mutant was 2.8, 95% confidence interval 1.0-7.4), and (2) GSTM1 absence was associated with TD (odds ratio 1.7, 95% confidence interval 1.2-2.4) particularly in white women. The CYP2D6 and CYP3A5 absence showed potential for significant associations in larger samples, particularly in white men. New studies need to replicate whether these or other genes could be used conjointly with clinical variables to identify subjects at risk for TD in clinical settings.