The purpose of this study was to investigate whether in practice monitoring of carbamazepine (CBZ) concentrations in saliva rather than serum was advantageous in young intellectually retarded epileptic patients. To this end an EMIT assay was used to measure CBZ concentrations in corresponding serum, serum ultrafiltrate and whole saliva samples from 30 young institutionalised epileptics. Twenty-five of the group were receiving additional drug therapy. Relatively large variations in the serum free fraction (range 16-43%) were observed, suggesting that the monitoring of the free drug concentration rather than total serum concentration might be more appropriate. However, salivary CBZ concentrations correlated only slightly better (r = 0.791) with free concentrations than with total serum concentrations (r = 0.779). These results suggest that salivary CBZ concentrations as measured by EMIT in these patients is a somewhat unreliable indicator of free drug concentration. Due to problems with the handling and collecting of sufficient saliva, blood sampling was generally preferred by the medical staff.