According to federal laws in Brazil (CONAMA: Res. 20/86, Art. 21), the maximum permissible total concentration of selenium in effluents from polluting sources must be as low as 0,05 mg L-1. In oil refineries, in which crude oil from seleniferous marine shales are processed, wastewater may contain much higher concentrations, thus requiring removal of this element by coprecipitation/ coagulation procedures. The behavior of selenium in such processes depends critically on its chemical form; e.g. selenocyanate is not efficiently removed by the conventional treatment using Fe(OH)3 precipitation. Knowledge on the speciation of selenium in wastewater is therefore important for the development of adequate treatment strategies. In this work, a new method for the speciation analysis of selenite (Se-IV), selenate (Se-VI) e selenocyanate (SeCN-) is described and first results are presented on the distribution of these species in wastewater samples from a Brazilian oil refinery plant. The method is based on their ion chromatographic separation followed by on-line detection of 77Se, 78Se and 82Se using ICPMS. The chromatographic system employed consisted of HPLC pump equipped with a manual syringe loading injector, and an anion exchange column (Metrosep A Supp1, Metrohm); the latter interfaced with the ICPMS detector via a concentric nebulizer/cyclonic spray chamber sample introduction device. Suppressed conductivity detection and constant potential amperometric detection of these selenium species were also evaluated using a commercial IC-system. Several eluants, already described in the literature for the speciation analysis of inorganic selenium, were tested (e.g. carbonatehidrogencarbonate and phthalate buffer, p-hydroxy benzoic acid, etc) permitting in most cases a good separation of Se(IV) and Se(VI), however resulting all in very long residence times (>20min) and associated peak broadening for the polarizable SeCN - ion. This drawback could be effectively avoided using a solution of cyanuric acid (0.003 mol L-1), modified with acetonitrile (2% v/v) and perclorate (0.002 mol L-1), as the mobile phase. Typical residence times for the three analyte species here studied were: selenite (210 s) < selenate (250 s) < selenocyanate (450 s), with repeatabilities in peak position better than 1% and peak area evaluation better than 3 %. Limits of detection (in ng) for these species using an ELAN 5000 instrument and a 500-uL sample injection loop are 0.04, 0.05 and 0.09, respectively. No certified reference materials were available for this study, however, results on spiked effluent samples of complex composition showed acceptable recoveries (85% - 115%) and repeatibilities (RSD < 5 %), thus validating this method for its intended purpose. Once optimized, the method was applied to wastewater samples from a Brazilian oil refinery plant. In all samples so far analyzed, selenocyanate was by far the most abundant selenium species reaching concentrations of up to 90 uL-1, followed by selenite. Selenate concentrations were below the quantification limit in all samples, explainable by the reducing character of these waters. Total concentrations of selenium in most samples, assessed by hydride generation ICPMS and alternatively by ICPOES, exceeded those obtained from speciation analysis, indicating the presence of other selenium species, not observed by the here used methodology. Work is in progress to identify and quantify these species.