Background: We undertook a study to determine the effects of four routes of administation on the efficacy of midazolam for premedication. Methods: In a randomized double-blind study, 119 unmedicated children, ASA I-II, aged 1.5-5 years, who were scheduled for minor elective surgery and who had been planned to received midazolam as a premedicant drug, were randomly assigned to one of four groups. Group I received intranasal midazolam 0.3 mg·kg-1; group II, oral midazolam 0.5 mg·kg-1; group III, rectal midazolam 0.5 mg·kg-1; and group IV, sublingual midazolam 0.3 mg·kg-1. A blinded observer assessed the children for sedation and anxiolysis every 5 min prior to surgery. Quality of mask acceptance for induction, postanaesthesia care unit behaviour and parents' satisfaction were evaluated. Thirty patients were enrolled in each of groups I, III and IV. Twenty-nine patients were enrolled in group II. Results: There were no significant differences in sedation and anxiety levels among the four groups. Average sedation and anxiolysis increased with time, achieving a maximum at 20 min in group I and at 30 min in groups II-IV. Patient mask acceptance was good for more than 75% of the children. Although the intranasal route provides a faster effect, it causes significant nasal irritation. Seventy-seven percent of the children from this group cried after drug administration. Most parents in all groups (67-73%) were satisfied with the premedication. Conclusions: Intranasal, oral, rectal and sublingual midazolam produces good levels of sedation and anxiolysis. Mask acceptance for inhalation induction was easy in the majority of children, irrespective of the route of drug administration.