Objective: To test the efficacy of current treatment recommendations for parkinsonism and tardive dyskinesia (TD) severity in patients with severe mental illness (SMI). Methods: We present an 18-year prospective study including all 223 patients with SMI (as defined by the 1987 US National Institute of Mental Health, which were based on DSM-III-R diagnostic criteria) receiving care from the only psychiatric hospital of the former Netherlands Antilles. Eight clinical assessments (1992-2009) focused on movement disorders and medication use. Tardive dyskinesia was measured by the Abnormal Involuntary Movement Scale and parkinsonism by the Unified Parkinson's Disease Rating Scale. Antipsychotics were classified into first-generation antipsychotic (FGA) versus second-generation antipsychotic (SGA) and high versus low dopamine 2 (D-2) affinity categories. The effect that switching has within each category on subsequent movement scores was calculated separately by using time-lagged multilevel logistic regression models. Results: There was a significant association between reduction in TD severity and starting/switching to an FGA (B = -3.54, P Conclusions: The results show that switching from an FGA to an SGA does not necessarily result in a reduction of TD or parkinsonism. Only stopping all antipsychotics reduces the severity of parkinsonism, and starting an FGA or a high D-2 affinity antipsychotic may reduce the severity of TD.