A randomised, single-blind study to investigate patient-controlled epidural analgesia during labour was conducted using a solution of low-dose bupivacaine-fentanyl. Two groups (n = 25 in each) received a constant infusion supplemented by patient demand boluses, and midwife-administered boluses if required, the size of the infusion and patient increments varying between groups. The quality of analgesia, as assessed by pain scores and patient ratings, was high and participant acceptability very high in both groups. Fourteen per cent of participants were withdrawn due to cephalad extension of block to T5 or supplementary bolus requirements outside the study protocol. Epidural side-effects were not troublesome and drug dose utilisation was low. Significantly greater pain relief (P less than 0.04) three hours after commencing patient-controlled epidural analgesia and a trend to fewer women requiring more than two supplementary midwife-administered boluses (P = 0.11) was seen in the group receiving a higher infusion rate. This group used significantly more bupivacaine (P less than 0.04) and fentanyl (P less than 0.001), but this did not appear to be clinically important with respect to degree of motor block or side-effects. Patient-controlled epidural analgesia appears to be an effective and well accepted method of obstetric analgesia and warrants further investigation.