We have assessed in 20 patients the accuracy and precision of an infusion profile for atracurium, which continually set the infusion rate to maintain stable muscle paralysis and a target steady state plasma concentration, when equilibrium between the biophase and plasma had occurred. Muscle paralysis was stable after 20 min, with a mean absolute drift in muscle paralysis in the succeeding 40 min of 0.13 (SD 0.07)% T1/Tc (height of first twitch/height of control twitch) per min. The plasma samples after 30 min, which were assessed empirically as being in equilibrium with the biophase, had an overall mean bias of 8.0 (SEM 3.7)% (P less than 0.05) and an overall mean absolute prediction error of 16.4 (SEM 2.5)% from the target steady state concentration being delivered by the infusion. The profile was then used to estimate the steady state plasma concentration of atracurium required to maintain 90% paralysis (Cpss90), by manually adjusting the delivered target concentration of the infusion until muscle paralysis was stable at 88-92% inhibition of T1/Tc for 15-20 min, with three plasma samples taken over the next 10 min. Measurements were completed within 60-90 min. The mean Cpss90 of atracurium with propofol was 1.039 (SD 0.224) microgram ml-1 (n = 10), with thiopentone 1.334 (0.378) microgram ml-1 (n = 10), and with opioid anaesthesia 0.915 (0.221) microgram ml-1 (n = 10). These differences in the Cpss90 explain some of the variability in response which occurs with neuromuscular blocking drugs. The technique enables the Cpss90 of a myoneural blocker to be determined by a simple model-independent method.