Musk xylene (2,4,6-trinitro-1,3-dimethyl-5-tert.-butylbenzene, MX), a synthetic musk often used in different fragrances and soaps to substitute the natural musk, is a potential contaminant of humans. In this publication, a specific and sensitive detection method for the determination of musk xylene in human blood samples is described. The clean-up of the blood samples includes an extraction step followed by a solid-phase adsorption to separate MX from other plasma components. Separation and detection was carried out by capillary gas chromatography and an electron capture detector (GC-ECD). The results were verified using qualitative capillary gas chromatography and a mass selective detector with electron impact ionisation (GC-EI-MS). epsilon-Hexachlorocyclohexane (epsilon-HCH) is used as internal standard. The reliability of the GC-ECD method has been proved. The relative standard deviations of the within-series imprecision were 12.7% for samples with a concentration of 0.5 microg/l and 2.1% for samples with a concentration of 5.0 microg/l, whereas the relative standard deviations for the between-day imprecision were 14.9% (0.5 microg/l samples) and 3.4% (5.0 microg/l samples). The losses during sample treatment were between 10.1% and 17.8%. No interfering peaks were observed. The absolute detection limit was 0.1 microg/l plasma. A total of 72 human blood samples were analysed to determine the MX concentrations within the general population. In 66 of the 72 human blood samples, the MX concentrations ranged from 0.10 to 1.12 microg/l plasma for the described method. In six samples no MX was detected. The median concentration was 0.24+/-0.23 microg MX/l plasma. The 95 percentile was 0.79 microg/l. No correlation could be found between MX concentrations and smoking habit, broca index, age, sex as well as fish consumption habits. Nevertheless, the results demonstrate the exposure of the general population to MX.