Background. The purpose of this study is to test precision and systematic bias of a target controlled infusion (TCI) of propofol in human volunteers at two sedative concentrations. Methods. We studied the 'Diprifusor' model (Marsh Pharmacokinetics and a Graseby® 3400 infusion pump) in 18 human volunteers at two sedative target plasma concentrations (0.5 and 1.0 μg ml-1). Twenty minutes after infusion start or change and 20 min after discontinuation of the infusion plasma propofol concentrations were measured using liquid chromatography-mass spectroscopy (LC-MS). Plasma propofol concentrations were compared with concentrations predicted by the TCI system. Agreement of those two measures (precision and bias) was determined using regression analysis. Results. We found little systematic bias but poor precision. When setting the TCI system to deliver a plasma concentration of 1.0 μg ml-1 one can predict the actual plasma concentration with 95% confidence only within a range of 0.44-1.38 μg ml-1. Conclusions. This finding helps to explain differences in responses to propofol sedation; pharmacokinetic variability appears to be an important factor. © The Board of Management and Trustees of the British Journal of Anaesthesia 2005. All rights reserved.