Childhood dystonia that does not respond to treatment with levodopa (dopa-nonresponsive dystonia, DND) has an unclear pathogenesis and is notoriously difficult to treat. To test the hypothesis that there may be abnormalities in serotonin turnover in DND we measured cerebrospinal fluid (CSF) concentrations of homovanillic (HVA) and 5-hydroxyindoleacetic (HIAA) acids, metabolites of dopamine and serotonin, respectively, in 18 children with dystonia not responsive to levodopa. These were combined with a reference population of 85 children with neurologic or metabolic disease known not to affect dopamine or serotonin metabolism. Because of the known natural age-related decrement in HVA and HIAA concentrations, the results were analyzed using multiple regression using age and DND as predictors of CSF HIAA and HVA concentrations. DND was a highly significant predictor of CSF HIAA concentration (p < 0.001) but not of CSF HVA concentration (p = 0.59). After fitting a regression model, the geometric mean ratio of CSF HIAA in DND compared with the reference range was 0.53 whereas that for CSF HVA was 0.95. We also analyzed CSF HIAA/HVA ratios. After fitting a regression model, we found no dependence on age, and the mean of CSF HIAA/HVA in DND was 0.28 whereas that for the reference range was 0.49 (p < 0.001). We conclude that a significant number of children with DND have reduced CNS serotonin turnover. Treatment with drugs that increase serotonin concentration in the synaptic cleft should be considered in this group of patients.