A comparative study of different chemical modifiers for the direct determination of arsenic in real and in synthetic sea water samples of high (72.8 per thousand ) and low (34.2 per thousand ) salinity was carried out. The use of lanthanum chloride and magnesium nitrate (LOD=3.0 microg l(-1)) offers good recoveries for low salinities, while for high salinities an enhancement higher than 200% was obtained, rendering this method unsuitable for sea water samples. The use of silver nitrate (LOD=2.0 microg l(-1)) produces unfavourable analytical recoveries, around 150 and 300% for synthetic sea water of low and high salinity. Zirconium oxychloride (LOD=1.8 microg l(-1)) and palladium nitrate alone or combined with magnesium nitrate and reducing agents (LOD=1.1-1.3 microg l(-1)) produce analytical recoveries close to 100% for all arsenic concentration tried over all salinities; thus, a single calibration curve using aqueous standard solutions may be applied to the analysis of sea water samples over all salinities. Finally, the addition of reducing agents and magnesium nitrate to palladium does not improve the sensitivity.