We have compared the pharmacokinetics of cisatracurium with atracurium when given by bolus dose followed by continuous infusion. Twenty healthy patients were anaesthetised with thiopentone, midazolam, fentanyl and 70% nitrous oxide in oxygen. Ten patients (Group C) were randomly allocated to receive cisatracurium 0.1 mg.kg-1 and 10 patients (Group A) were given atracurium 0.5 mg.kg-1. Neuromuscular block was monitored using a mechanomyograph. When the first twitch of the train-of-four had recovered to 5% of control, an infusion of cisatracurium 3 micrograms.kg-1.min-1 was started in Group C and an infusion of atracurium 10 micrograms.kg-1.min-1 was started in Group A. The infusion rates were adjusted to maintain the first twitch of the train-of-four at 5% of control. The times to 90% and maximum depression of the first twitch of the train-of-four were significantly longer after cisatracurium than atracurium (2.2 and 3.4 min compared with 1.3 and 1.8 min, respectively; p < 0.01 in each instance). No significant differences were found in recovery parameters between the two groups. Blood samples were taken at regular intervals following the bolus, during the infusion and for 8 h thereafter. The plasma samples were analysed using high-performance liquid chromatography for cisatracurium and atracurium (using a method which distinguishes between the three geometric isomer groups), laudanosine and monoquaternary alcohol. The results were analysed using the Non-linear Mixed Effects Model program. A two-compartment model was fitted to the data. The different isomer groups of atracurium have different pharmacokinetics, the trans-trans group having the highest clearance (1440 ml.min-1) and the cis-cis group the lowest (499 ml.min-1). The clearance of cisatracurium (425 ml.min-1) is less than that of cis-cis atracurium and its elimination half-life is longer (34.9 min and 21.9 min, respectively). The plasma concentration of laudanosine after cisatracurium was one-fifth of that after atracurium.