Two calibration curves were run using standard solutions of SeIV (injected in triplicate), changing the sensitivity of the detector in order to cover a wide range of concentrations, from 0.3 to 40 ng ml1 (Calibration 1) and from 40 to 1300 ng ml1 (Calibration 2). Table 2 shows the characteristics of both calibration curves, limit of detection and the repeatability of the method expressed as RSD (n7, injected in triplicate), with good results in all instances. The sample throughput was ca. 5 samples h1 for the pre-treatment step and ca. 50 samples h1 for derivatisationdetection.