To determine which of three types of rectal sedation was most effective preoperatively in facilitating parental separation and intravenous cannulation in young children, 100 children 3.0 +/- 1.7 (mean +/- SD) yr of age were randomly assigned to four equal groups. One group (M-K-A) received rectal midazolam (0.5 mg/kg), ketamine (3 mg/kg), and atropine (0.02 mg/kg). The other sedation groups received the same doses of midazolam and atropine (M-A) or ketamine and atropine (K-A) alone, and the control group (A) received only rectal atropine. Most children in either the M-K-A (100%) or M-A (92%) groups separated easily from their parents without struggling or crying, significantly more than in the K-A (60%) or A (64%) groups. However, more children in the M-K-A group (44%) were asleep during separation than in the M-A group (8%; P < 0.05). Only 20% of the children in the M-A or M-K-A groups cried during intravenous catheter placement, significantly less than in the K-A (56%) or A (92%) groups. Intravenous catheter placement was also successful significantly more often in the M-A (80%) and M-K-A (84%) groups than in the K-A (48%) or A (40%) groups. Complications were similar among the groups, but there was evidence that midazolam prolonged recovery time in some patients. Rectal midazolam with or without ketamine is a useful technique when intravenous catheter placement before induction of anesthesia is desired.