This study evaluated the ability to flexibly shift cognitive set and to consistently maintain a new response preference using the Penn Conditional Exclusion Test (PCET). The relationship of performance errors with catechol-O-methyltransferase (COMT) rs4680 (Val158Met) genotype (Met carriers vs. Val homozygotes) on test performance before and after antipsychotic treatment in 32 first episode psychosis (FEP) patients was examined. After treatment, patients demonstrated a mixture of beneficial and adverse cognitive outcomes that varied in relation to COMT genotype. Met carriers showed decreased perseverative and regressive errors, reflecting improved cognitive flexibility and enhanced stability of behavioral preferences, respectively. In contrast, Val homozygotes exhibited an increase in regressive errors after treatment. These findings suggest that Val homozygotes may be vulnerable to adverse effects of antipsychotic medication on cognitive processes that maintain consistent adaptive response preferences, an ability linked to the striatum in rodent models.