The well-known beneficial health effects of Se have demanded the development of rapid and accurate methods for its analysis. A flow injection (FI) method with inductively coupled plasma mass spectrometry (ICP-MS) as a selenium-selective detector was optimized. Flow injection was carried out using a Knauer 1100 smartline inert series liquid chromatograph coupled with a Perkin Elmer DRC II ICP-mass spectrometer. For sample injection a Perkin Elmer electronic valve equipped with a 25microL sample loop was employed. Before measurement, standards or samples were administered with 1microg/L rhodium as internal standard for correction of changes in detector response according to changes in sample electrolyte concentration. The method characterization parameters are: LOD (3sigma criterion): 26ng/L, LOQ (10sigma criterion): 86ng/L, linearity: 0.05->10microg/L, r(2)=0.9999, serial or day-to-day precision at 2microg/L: 4.48% or 5.6%. Accuracy was determined by (a) recovery experiments (CSF spiked with 2microg/L Se); (b) comparison of FI-ICP-MS measurement with graphite furnace atomic absorption (GFAAS) measurements of 1:10 diluted serum samples; (c) Se determination in urine and serum control materials. Recovery (a) was 101.4%, measurement comparison with GFAAS (b) showed 98.8% (5 serum samples, 1:10 diluted in the range of 0.5-1.3microg/L, compared to GFAAS determination, which was set to 100%), and accuracy was 96.8% or 105.6% for the serum or urine control material. Analysis time per sample was short and typically below 2min for the complete measurement, including sample introduction, sample-line purge and quadruplicate Se determination. This method was used to determine Se in cerebrospinal fluid (CSF) and plasma (here parallel to GFAAS) in 35 paired serum and CSF samples. Se determination gave values in the range of 42-130microg/L for serum and 1.63-6.66microg/L for CSF. The median for Se in 35 individual CSF samples was 3.28microg/L, the mean (+/-SD) was 3.67 (1.35)microg/L, whilst for individual serum samples the median was 81microg/L and the mean (+/-SD) was 85 (26)microg/L. When relating the paired Se concentrations of CSF samples to respective serum samples it turned out that Se-CSF (behind blood brain barrier (BBB)) is independent on Se-serum concentration (before BBB).